1 @c Copyright (C) 1988-2019 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 @hook TARGET_OPTION_DEFAULT_PARAMS
741 @hook TARGET_OPTION_VALIDATE_PARAM
743 @defmac SWITCHABLE_TARGET
744 Some targets need to switch between substantially different subtargets
745 during compilation. For example, the MIPS target has one subtarget for
746 the traditional MIPS architecture and another for MIPS16. Source code
747 can switch between these two subarchitectures using the @code{mips16}
748 and @code{nomips16} attributes.
750 Such subtargets can differ in things like the set of available
751 registers, the set of available instructions, the costs of various
752 operations, and so on. GCC caches a lot of this type of information
753 in global variables, and recomputing them for each subtarget takes a
754 significant amount of time. The compiler therefore provides a facility
755 for maintaining several versions of the global variables and quickly
756 switching between them; see @file{target-globals.h} for details.
758 Define this macro to 1 if your target needs this facility. The default
762 @hook TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P
764 @node Per-Function Data
765 @section Defining data structures for per-function information.
766 @cindex per-function data
767 @cindex data structures
769 If the target needs to store information on a per-function basis, GCC
770 provides a macro and a couple of variables to allow this. Note, just
771 using statics to store the information is a bad idea, since GCC supports
772 nested functions, so you can be halfway through encoding one function
773 when another one comes along.
775 GCC defines a data structure called @code{struct function} which
776 contains all of the data specific to an individual function. This
777 structure contains a field called @code{machine} whose type is
778 @code{struct machine_function *}, which can be used by targets to point
779 to their own specific data.
781 If a target needs per-function specific data it should define the type
782 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
783 This macro should be used to initialize the function pointer
784 @code{init_machine_status}. This pointer is explained below.
786 One typical use of per-function, target specific data is to create an
787 RTX to hold the register containing the function's return address. This
788 RTX can then be used to implement the @code{__builtin_return_address}
789 function, for level 0.
791 Note---earlier implementations of GCC used a single data area to hold
792 all of the per-function information. Thus when processing of a nested
793 function began the old per-function data had to be pushed onto a
794 stack, and when the processing was finished, it had to be popped off the
795 stack. GCC used to provide function pointers called
796 @code{save_machine_status} and @code{restore_machine_status} to handle
797 the saving and restoring of the target specific information. Since the
798 single data area approach is no longer used, these pointers are no
801 @defmac INIT_EXPANDERS
802 Macro called to initialize any target specific information. This macro
803 is called once per function, before generation of any RTL has begun.
804 The intention of this macro is to allow the initialization of the
805 function pointer @code{init_machine_status}.
808 @deftypevar {void (*)(struct function *)} init_machine_status
809 If this function pointer is non-@code{NULL} it will be called once per
810 function, before function compilation starts, in order to allow the
811 target to perform any target specific initialization of the
812 @code{struct function} structure. It is intended that this would be
813 used to initialize the @code{machine} of that structure.
815 @code{struct machine_function} structures are expected to be freed by GC@.
816 Generally, any memory that they reference must be allocated by using
817 GC allocation, including the structure itself.
821 @section Storage Layout
822 @cindex storage layout
824 Note that the definitions of the macros in this table which are sizes or
825 alignments measured in bits do not need to be constant. They can be C
826 expressions that refer to static variables, such as the @code{target_flags}.
827 @xref{Run-time Target}.
829 @defmac BITS_BIG_ENDIAN
830 Define this macro to have the value 1 if the most significant bit in a
831 byte has the lowest number; otherwise define it to have the value zero.
832 This means that bit-field instructions count from the most significant
833 bit. If the machine has no bit-field instructions, then this must still
834 be defined, but it doesn't matter which value it is defined to. This
835 macro need not be a constant.
837 This macro does not affect the way structure fields are packed into
838 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
841 @defmac BYTES_BIG_ENDIAN
842 Define this macro to have the value 1 if the most significant byte in a
843 word has the lowest number. This macro need not be a constant.
846 @defmac WORDS_BIG_ENDIAN
847 Define this macro to have the value 1 if, in a multiword object, the
848 most significant word has the lowest number. This applies to both
849 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
850 order of words in memory is not the same as the order in registers. This
851 macro need not be a constant.
854 @defmac REG_WORDS_BIG_ENDIAN
855 On some machines, the order of words in a multiword object differs between
856 registers in memory. In such a situation, define this macro to describe
857 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
858 the order of words in memory.
861 @defmac FLOAT_WORDS_BIG_ENDIAN
862 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
863 @code{TFmode} floating point numbers are stored in memory with the word
864 containing the sign bit at the lowest address; otherwise define it to
865 have the value 0. This macro need not be a constant.
867 You need not define this macro if the ordering is the same as for
871 @defmac BITS_PER_WORD
872 Number of bits in a word. If you do not define this macro, the default
873 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
876 @defmac MAX_BITS_PER_WORD
877 Maximum number of bits in a word. If this is undefined, the default is
878 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
879 largest value that @code{BITS_PER_WORD} can have at run-time.
882 @defmac UNITS_PER_WORD
883 Number of storage units in a word; normally the size of a general-purpose
884 register, a power of two from 1 or 8.
887 @defmac MIN_UNITS_PER_WORD
888 Minimum number of units in a word. If this is undefined, the default is
889 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
890 smallest value that @code{UNITS_PER_WORD} can have at run-time.
894 Width of a pointer, in bits. You must specify a value no wider than the
895 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
896 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
897 a value the default is @code{BITS_PER_WORD}.
900 @defmac POINTERS_EXTEND_UNSIGNED
901 A C expression that determines how pointers should be extended from
902 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
903 greater than zero if pointers should be zero-extended, zero if they
904 should be sign-extended, and negative if some other sort of conversion
905 is needed. In the last case, the extension is done by the target's
906 @code{ptr_extend} instruction.
908 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
909 and @code{word_mode} are all the same width.
912 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
913 A macro to update @var{m} and @var{unsignedp} when an object whose type
914 is @var{type} and which has the specified mode and signedness is to be
915 stored in a register. This macro is only called when @var{type} is a
918 On most RISC machines, which only have operations that operate on a full
919 register, define this macro to set @var{m} to @code{word_mode} if
920 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
921 cases, only integer modes should be widened because wider-precision
922 floating-point operations are usually more expensive than their narrower
925 For most machines, the macro definition does not change @var{unsignedp}.
926 However, some machines, have instructions that preferentially handle
927 either signed or unsigned quantities of certain modes. For example, on
928 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
929 sign-extend the result to 64 bits. On such machines, set
930 @var{unsignedp} according to which kind of extension is more efficient.
932 Do not define this macro if it would never modify @var{m}.
935 @hook TARGET_C_EXCESS_PRECISION
937 @hook TARGET_PROMOTE_FUNCTION_MODE
939 @defmac PARM_BOUNDARY
940 Normal alignment required for function parameters on the stack, in
941 bits. All stack parameters receive at least this much alignment
942 regardless of data type. On most machines, this is the same as the
946 @defmac STACK_BOUNDARY
947 Define this macro to the minimum alignment enforced by hardware for the
948 stack pointer on this machine. The definition is a C expression for the
949 desired alignment (measured in bits). This value is used as a default
950 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
951 this should be the same as @code{PARM_BOUNDARY}.
954 @defmac PREFERRED_STACK_BOUNDARY
955 Define this macro if you wish to preserve a certain alignment for the
956 stack pointer, greater than what the hardware enforces. The definition
957 is a C expression for the desired alignment (measured in bits). This
958 macro must evaluate to a value equal to or larger than
959 @code{STACK_BOUNDARY}.
962 @defmac INCOMING_STACK_BOUNDARY
963 Define this macro if the incoming stack boundary may be different
964 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
965 to a value equal to or larger than @code{STACK_BOUNDARY}.
968 @defmac FUNCTION_BOUNDARY
969 Alignment required for a function entry point, in bits.
972 @defmac BIGGEST_ALIGNMENT
973 Biggest alignment that any data type can require on this machine, in
974 bits. Note that this is not the biggest alignment that is supported,
975 just the biggest alignment that, when violated, may cause a fault.
978 @hook TARGET_ABSOLUTE_BIGGEST_ALIGNMENT
980 @defmac MALLOC_ABI_ALIGNMENT
981 Alignment, in bits, a C conformant malloc implementation has to
982 provide. If not defined, the default value is @code{BITS_PER_WORD}.
985 @defmac ATTRIBUTE_ALIGNED_VALUE
986 Alignment used by the @code{__attribute__ ((aligned))} construct. If
987 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
990 @defmac MINIMUM_ATOMIC_ALIGNMENT
991 If defined, the smallest alignment, in bits, that can be given to an
992 object that can be referenced in one operation, without disturbing any
993 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
994 on machines that don't have byte or half-word store operations.
997 @defmac BIGGEST_FIELD_ALIGNMENT
998 Biggest alignment that any structure or union field can require on this
999 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1000 structure and union fields only, unless the field alignment has been set
1001 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1004 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{type}, @var{computed})
1005 An expression for the alignment of a structure field @var{field} of
1006 type @var{type} if the alignment computed in the usual way (including
1007 applying of @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1008 alignment) is @var{computed}. It overrides alignment only if the
1009 field alignment has not been set by the
1010 @code{__attribute__ ((aligned (@var{n})))} construct. Note that @var{field}
1011 may be @code{NULL_TREE} in case we just query for the minimum alignment
1012 of a field of type @var{type} in structure context.
1015 @defmac MAX_STACK_ALIGNMENT
1016 Biggest stack alignment guaranteed by the backend. Use this macro
1017 to specify the maximum alignment of a variable on stack.
1019 If not defined, the default value is @code{STACK_BOUNDARY}.
1021 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1022 @c But the fix for PR 32893 indicates that we can only guarantee
1023 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1024 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1027 @defmac MAX_OFILE_ALIGNMENT
1028 Biggest alignment supported by the object file format of this machine.
1029 Use this macro to limit the alignment which can be specified using the
1030 @code{__attribute__ ((aligned (@var{n})))} construct for functions and
1031 objects with static storage duration. The alignment of automatic
1032 objects may exceed the object file format maximum up to the maximum
1033 supported by GCC. If not defined, the default value is
1034 @code{BIGGEST_ALIGNMENT}.
1036 On systems that use ELF, the default (in @file{config/elfos.h}) is
1037 the largest supported 32-bit ELF section alignment representable on
1038 a 32-bit host e.g.@: @samp{(((uint64_t) 1 << 28) * 8)}.
1039 On 32-bit ELF the largest supported section alignment in bits is
1040 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1043 @hook TARGET_STATIC_RTX_ALIGNMENT
1045 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1046 If defined, a C expression to compute the alignment for a variable in
1047 the static store. @var{type} is the data type, and @var{basic-align} is
1048 the alignment that the object would ordinarily have. The value of this
1049 macro is used instead of that alignment to align the object.
1051 If this macro is not defined, then @var{basic-align} is used.
1054 One use of this macro is to increase alignment of medium-size data to
1055 make it all fit in fewer cache lines. Another is to cause character
1056 arrays to be word-aligned so that @code{strcpy} calls that copy
1057 constants to character arrays can be done inline.
1060 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1061 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1062 some alignment increase, instead of optimization only purposes. E.g.@
1063 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1064 must be aligned to 16 byte boundaries.
1066 If this macro is not defined, then @var{basic-align} is used.
1069 @hook TARGET_CONSTANT_ALIGNMENT
1071 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1072 If defined, a C expression to compute the alignment for a variable in
1073 the local store. @var{type} is the data type, and @var{basic-align} is
1074 the alignment that the object would ordinarily have. The value of this
1075 macro is used instead of that alignment to align the object.
1077 If this macro is not defined, then @var{basic-align} is used.
1079 One use of this macro is to increase alignment of medium-size data to
1080 make it all fit in fewer cache lines.
1082 If the value of this macro has a type, it should be an unsigned type.
1085 @hook TARGET_VECTOR_ALIGNMENT
1087 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1088 If defined, a C expression to compute the alignment for stack slot.
1089 @var{type} is the data type, @var{mode} is the widest mode available,
1090 and @var{basic-align} is the alignment that the slot would ordinarily
1091 have. The value of this macro is used instead of that alignment to
1094 If this macro is not defined, then @var{basic-align} is used when
1095 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1098 This macro is to set alignment of stack slot to the maximum alignment
1099 of all possible modes which the slot may have.
1101 If the value of this macro has a type, it should be an unsigned type.
1104 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1105 If defined, a C expression to compute the alignment for a local
1106 variable @var{decl}.
1108 If this macro is not defined, then
1109 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1112 One use of this macro is to increase alignment of medium-size data to
1113 make it all fit in fewer cache lines.
1115 If the value of this macro has a type, it should be an unsigned type.
1118 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1119 If defined, a C expression to compute the minimum required alignment
1120 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1121 @var{mode}, assuming normal alignment @var{align}.
1123 If this macro is not defined, then @var{align} will be used.
1126 @defmac EMPTY_FIELD_BOUNDARY
1127 Alignment in bits to be given to a structure bit-field that follows an
1128 empty field such as @code{int : 0;}.
1130 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1133 @defmac STRUCTURE_SIZE_BOUNDARY
1134 Number of bits which any structure or union's size must be a multiple of.
1135 Each structure or union's size is rounded up to a multiple of this.
1137 If you do not define this macro, the default is the same as
1138 @code{BITS_PER_UNIT}.
1141 @defmac STRICT_ALIGNMENT
1142 Define this macro to be the value 1 if instructions will fail to work
1143 if given data not on the nominal alignment. If instructions will merely
1144 go slower in that case, define this macro as 0.
1147 @defmac PCC_BITFIELD_TYPE_MATTERS
1148 Define this if you wish to imitate the way many other C compilers handle
1149 alignment of bit-fields and the structures that contain them.
1151 The behavior is that the type written for a named bit-field (@code{int},
1152 @code{short}, or other integer type) imposes an alignment for the entire
1153 structure, as if the structure really did contain an ordinary field of
1154 that type. In addition, the bit-field is placed within the structure so
1155 that it would fit within such a field, not crossing a boundary for it.
1157 Thus, on most machines, a named bit-field whose type is written as
1158 @code{int} would not cross a four-byte boundary, and would force
1159 four-byte alignment for the whole structure. (The alignment used may
1160 not be four bytes; it is controlled by the other alignment parameters.)
1162 An unnamed bit-field will not affect the alignment of the containing
1165 If the macro is defined, its definition should be a C expression;
1166 a nonzero value for the expression enables this behavior.
1168 Note that if this macro is not defined, or its value is zero, some
1169 bit-fields may cross more than one alignment boundary. The compiler can
1170 support such references if there are @samp{insv}, @samp{extv}, and
1171 @samp{extzv} insns that can directly reference memory.
1173 The other known way of making bit-fields work is to define
1174 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1175 Then every structure can be accessed with fullwords.
1177 Unless the machine has bit-field instructions or you define
1178 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1179 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1181 If your aim is to make GCC use the same conventions for laying out
1182 bit-fields as are used by another compiler, here is how to investigate
1183 what the other compiler does. Compile and run this program:
1202 printf ("Size of foo1 is %d\n",
1203 sizeof (struct foo1));
1204 printf ("Size of foo2 is %d\n",
1205 sizeof (struct foo2));
1210 If this prints 2 and 5, then the compiler's behavior is what you would
1211 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1214 @defmac BITFIELD_NBYTES_LIMITED
1215 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1216 to aligning a bit-field within the structure.
1219 @hook TARGET_ALIGN_ANON_BITFIELD
1221 @hook TARGET_NARROW_VOLATILE_BITFIELD
1223 @hook TARGET_MEMBER_TYPE_FORCES_BLK
1225 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1226 Define this macro as an expression for the alignment of a type (given
1227 by @var{type} as a tree node) if the alignment computed in the usual
1228 way is @var{computed} and the alignment explicitly specified was
1231 The default is to use @var{specified} if it is larger; otherwise, use
1232 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1235 @defmac MAX_FIXED_MODE_SIZE
1236 An integer expression for the size in bits of the largest integer
1237 machine mode that should actually be used. All integer machine modes of
1238 this size or smaller can be used for structures and unions with the
1239 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1240 (DImode)} is assumed.
1243 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1244 If defined, an expression of type @code{machine_mode} that
1245 specifies the mode of the save area operand of a
1246 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1247 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1248 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1249 having its mode specified.
1251 You need not define this macro if it always returns @code{Pmode}. You
1252 would most commonly define this macro if the
1253 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1257 @defmac STACK_SIZE_MODE
1258 If defined, an expression of type @code{machine_mode} that
1259 specifies the mode of the size increment operand of an
1260 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1262 You need not define this macro if it always returns @code{word_mode}.
1263 You would most commonly define this macro if the @code{allocate_stack}
1264 pattern needs to support both a 32- and a 64-bit mode.
1267 @hook TARGET_LIBGCC_CMP_RETURN_MODE
1269 @hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1271 @hook TARGET_UNWIND_WORD_MODE
1273 @hook TARGET_MS_BITFIELD_LAYOUT_P
1275 @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1277 @hook TARGET_FIXED_POINT_SUPPORTED_P
1279 @hook TARGET_EXPAND_TO_RTL_HOOK
1281 @hook TARGET_INSTANTIATE_DECLS
1283 @hook TARGET_MANGLE_TYPE
1286 @section Layout of Source Language Data Types
1288 These macros define the sizes and other characteristics of the standard
1289 basic data types used in programs being compiled. Unlike the macros in
1290 the previous section, these apply to specific features of C and related
1291 languages, rather than to fundamental aspects of storage layout.
1293 @defmac INT_TYPE_SIZE
1294 A C expression for the size in bits of the type @code{int} on the
1295 target machine. If you don't define this, the default is one word.
1298 @defmac SHORT_TYPE_SIZE
1299 A C expression for the size in bits of the type @code{short} on the
1300 target machine. If you don't define this, the default is half a word.
1301 (If this would be less than one storage unit, it is rounded up to one
1305 @defmac LONG_TYPE_SIZE
1306 A C expression for the size in bits of the type @code{long} on the
1307 target machine. If you don't define this, the default is one word.
1310 @defmac ADA_LONG_TYPE_SIZE
1311 On some machines, the size used for the Ada equivalent of the type
1312 @code{long} by a native Ada compiler differs from that used by C@. In
1313 that situation, define this macro to be a C expression to be used for
1314 the size of that type. If you don't define this, the default is the
1315 value of @code{LONG_TYPE_SIZE}.
1318 @defmac LONG_LONG_TYPE_SIZE
1319 A C expression for the size in bits of the type @code{long long} on the
1320 target machine. If you don't define this, the default is two
1321 words. If you want to support GNU Ada on your machine, the value of this
1322 macro must be at least 64.
1325 @defmac CHAR_TYPE_SIZE
1326 A C expression for the size in bits of the type @code{char} on the
1327 target machine. If you don't define this, the default is
1328 @code{BITS_PER_UNIT}.
1331 @defmac BOOL_TYPE_SIZE
1332 A C expression for the size in bits of the C++ type @code{bool} and
1333 C99 type @code{_Bool} on the target machine. If you don't define
1334 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1337 @defmac FLOAT_TYPE_SIZE
1338 A C expression for the size in bits of the type @code{float} on the
1339 target machine. If you don't define this, the default is one word.
1342 @defmac DOUBLE_TYPE_SIZE
1343 A C expression for the size in bits of the type @code{double} on the
1344 target machine. If you don't define this, the default is two
1348 @defmac LONG_DOUBLE_TYPE_SIZE
1349 A C expression for the size in bits of the type @code{long double} on
1350 the target machine. If you don't define this, the default is two
1354 @defmac SHORT_FRACT_TYPE_SIZE
1355 A C expression for the size in bits of the type @code{short _Fract} on
1356 the target machine. If you don't define this, the default is
1357 @code{BITS_PER_UNIT}.
1360 @defmac FRACT_TYPE_SIZE
1361 A C expression for the size in bits of the type @code{_Fract} on
1362 the target machine. If you don't define this, the default is
1363 @code{BITS_PER_UNIT * 2}.
1366 @defmac LONG_FRACT_TYPE_SIZE
1367 A C expression for the size in bits of the type @code{long _Fract} on
1368 the target machine. If you don't define this, the default is
1369 @code{BITS_PER_UNIT * 4}.
1372 @defmac LONG_LONG_FRACT_TYPE_SIZE
1373 A C expression for the size in bits of the type @code{long long _Fract} on
1374 the target machine. If you don't define this, the default is
1375 @code{BITS_PER_UNIT * 8}.
1378 @defmac SHORT_ACCUM_TYPE_SIZE
1379 A C expression for the size in bits of the type @code{short _Accum} on
1380 the target machine. If you don't define this, the default is
1381 @code{BITS_PER_UNIT * 2}.
1384 @defmac ACCUM_TYPE_SIZE
1385 A C expression for the size in bits of the type @code{_Accum} on
1386 the target machine. If you don't define this, the default is
1387 @code{BITS_PER_UNIT * 4}.
1390 @defmac LONG_ACCUM_TYPE_SIZE
1391 A C expression for the size in bits of the type @code{long _Accum} on
1392 the target machine. If you don't define this, the default is
1393 @code{BITS_PER_UNIT * 8}.
1396 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1397 A C expression for the size in bits of the type @code{long long _Accum} on
1398 the target machine. If you don't define this, the default is
1399 @code{BITS_PER_UNIT * 16}.
1402 @defmac LIBGCC2_GNU_PREFIX
1403 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1404 hook and should be defined if that hook is overriden to be true. It
1405 causes function names in libgcc to be changed to use a @code{__gnu_}
1406 prefix for their name rather than the default @code{__}. A port which
1407 uses this macro should also arrange to use @file{t-gnu-prefix} in
1408 the libgcc @file{config.host}.
1411 @defmac WIDEST_HARDWARE_FP_SIZE
1412 A C expression for the size in bits of the widest floating-point format
1413 supported by the hardware. If you define this macro, you must specify a
1414 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1415 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1419 @defmac DEFAULT_SIGNED_CHAR
1420 An expression whose value is 1 or 0, according to whether the type
1421 @code{char} should be signed or unsigned by default. The user can
1422 always override this default with the options @option{-fsigned-char}
1423 and @option{-funsigned-char}.
1426 @hook TARGET_DEFAULT_SHORT_ENUMS
1429 A C expression for a string describing the name of the data type to use
1430 for size values. The typedef name @code{size_t} is defined using the
1431 contents of the string.
1433 The string can contain more than one keyword. If so, separate them with
1434 spaces, and write first any length keyword, then @code{unsigned} if
1435 appropriate, and finally @code{int}. The string must exactly match one
1436 of the data type names defined in the function
1437 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1438 You may not omit @code{int} or change the order---that would cause the
1439 compiler to crash on startup.
1441 If you don't define this macro, the default is @code{"long unsigned
1446 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1447 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1448 dealing with size. This macro is a C expression for a string describing
1449 the name of the data type from which the precision of @code{sizetype}
1452 The string has the same restrictions as @code{SIZE_TYPE} string.
1454 If you don't define this macro, the default is @code{SIZE_TYPE}.
1457 @defmac PTRDIFF_TYPE
1458 A C expression for a string describing the name of the data type to use
1459 for the result of subtracting two pointers. The typedef name
1460 @code{ptrdiff_t} is defined using the contents of the string. See
1461 @code{SIZE_TYPE} above for more information.
1463 If you don't define this macro, the default is @code{"long int"}.
1467 A C expression for a string describing the name of the data type to use
1468 for wide characters. The typedef name @code{wchar_t} is defined using
1469 the contents of the string. See @code{SIZE_TYPE} above for more
1472 If you don't define this macro, the default is @code{"int"}.
1475 @defmac WCHAR_TYPE_SIZE
1476 A C expression for the size in bits of the data type for wide
1477 characters. This is used in @code{cpp}, which cannot make use of
1482 A C expression for a string describing the name of the data type to
1483 use for wide characters passed to @code{printf} and returned from
1484 @code{getwc}. The typedef name @code{wint_t} is defined using the
1485 contents of the string. See @code{SIZE_TYPE} above for more
1488 If you don't define this macro, the default is @code{"unsigned int"}.
1492 A C expression for a string describing the name of the data type that
1493 can represent any value of any standard or extended signed integer type.
1494 The typedef name @code{intmax_t} is defined using the contents of the
1495 string. See @code{SIZE_TYPE} above for more information.
1497 If you don't define this macro, the default is the first of
1498 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1499 much precision as @code{long long int}.
1502 @defmac UINTMAX_TYPE
1503 A C expression for a string describing the name of the data type that
1504 can represent any value of any standard or extended unsigned integer
1505 type. The typedef name @code{uintmax_t} is defined using the contents
1506 of the string. See @code{SIZE_TYPE} above for more information.
1508 If you don't define this macro, the default is the first of
1509 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1510 unsigned int"} that has as much precision as @code{long long unsigned
1514 @defmac SIG_ATOMIC_TYPE
1520 @defmacx UINT16_TYPE
1521 @defmacx UINT32_TYPE
1522 @defmacx UINT64_TYPE
1523 @defmacx INT_LEAST8_TYPE
1524 @defmacx INT_LEAST16_TYPE
1525 @defmacx INT_LEAST32_TYPE
1526 @defmacx INT_LEAST64_TYPE
1527 @defmacx UINT_LEAST8_TYPE
1528 @defmacx UINT_LEAST16_TYPE
1529 @defmacx UINT_LEAST32_TYPE
1530 @defmacx UINT_LEAST64_TYPE
1531 @defmacx INT_FAST8_TYPE
1532 @defmacx INT_FAST16_TYPE
1533 @defmacx INT_FAST32_TYPE
1534 @defmacx INT_FAST64_TYPE
1535 @defmacx UINT_FAST8_TYPE
1536 @defmacx UINT_FAST16_TYPE
1537 @defmacx UINT_FAST32_TYPE
1538 @defmacx UINT_FAST64_TYPE
1539 @defmacx INTPTR_TYPE
1540 @defmacx UINTPTR_TYPE
1541 C expressions for the standard types @code{sig_atomic_t},
1542 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1543 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1544 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1545 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1546 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1547 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1548 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1549 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1550 @code{SIZE_TYPE} above for more information.
1552 If any of these macros evaluates to a null pointer, the corresponding
1553 type is not supported; if GCC is configured to provide
1554 @code{<stdint.h>} in such a case, the header provided may not conform
1555 to C99, depending on the type in question. The defaults for all of
1556 these macros are null pointers.
1559 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1560 The C++ compiler represents a pointer-to-member-function with a struct
1567 ptrdiff_t vtable_index;
1574 The C++ compiler must use one bit to indicate whether the function that
1575 will be called through a pointer-to-member-function is virtual.
1576 Normally, we assume that the low-order bit of a function pointer must
1577 always be zero. Then, by ensuring that the vtable_index is odd, we can
1578 distinguish which variant of the union is in use. But, on some
1579 platforms function pointers can be odd, and so this doesn't work. In
1580 that case, we use the low-order bit of the @code{delta} field, and shift
1581 the remainder of the @code{delta} field to the left.
1583 GCC will automatically make the right selection about where to store
1584 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1585 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1586 set such that functions always start at even addresses, but the lowest
1587 bit of pointers to functions indicate whether the function at that
1588 address is in ARM or Thumb mode. If this is the case of your
1589 architecture, you should define this macro to
1590 @code{ptrmemfunc_vbit_in_delta}.
1592 In general, you should not have to define this macro. On architectures
1593 in which function addresses are always even, according to
1594 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1595 @code{ptrmemfunc_vbit_in_pfn}.
1598 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1599 Normally, the C++ compiler uses function pointers in vtables. This
1600 macro allows the target to change to use ``function descriptors''
1601 instead. Function descriptors are found on targets for whom a
1602 function pointer is actually a small data structure. Normally the
1603 data structure consists of the actual code address plus a data
1604 pointer to which the function's data is relative.
1606 If vtables are used, the value of this macro should be the number
1607 of words that the function descriptor occupies.
1610 @defmac TARGET_VTABLE_ENTRY_ALIGN
1611 By default, the vtable entries are void pointers, the so the alignment
1612 is the same as pointer alignment. The value of this macro specifies
1613 the alignment of the vtable entry in bits. It should be defined only
1614 when special alignment is necessary. */
1617 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1618 There are a few non-descriptor entries in the vtable at offsets below
1619 zero. If these entries must be padded (say, to preserve the alignment
1620 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1621 of words in each data entry.
1625 @section Register Usage
1626 @cindex register usage
1628 This section explains how to describe what registers the target machine
1629 has, and how (in general) they can be used.
1631 The description of which registers a specific instruction can use is
1632 done with register classes; see @ref{Register Classes}. For information
1633 on using registers to access a stack frame, see @ref{Frame Registers}.
1634 For passing values in registers, see @ref{Register Arguments}.
1635 For returning values in registers, see @ref{Scalar Return}.
1638 * Register Basics:: Number and kinds of registers.
1639 * Allocation Order:: Order in which registers are allocated.
1640 * Values in Registers:: What kinds of values each reg can hold.
1641 * Leaf Functions:: Renumbering registers for leaf functions.
1642 * Stack Registers:: Handling a register stack such as 80387.
1645 @node Register Basics
1646 @subsection Basic Characteristics of Registers
1648 @c prevent bad page break with this line
1649 Registers have various characteristics.
1651 @defmac FIRST_PSEUDO_REGISTER
1652 Number of hardware registers known to the compiler. They receive
1653 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1654 pseudo register's number really is assigned the number
1655 @code{FIRST_PSEUDO_REGISTER}.
1658 @defmac FIXED_REGISTERS
1659 @cindex fixed register
1660 An initializer that says which registers are used for fixed purposes
1661 all throughout the compiled code and are therefore not available for
1662 general allocation. These would include the stack pointer, the frame
1663 pointer (except on machines where that can be used as a general
1664 register when no frame pointer is needed), the program counter on
1665 machines where that is considered one of the addressable registers,
1666 and any other numbered register with a standard use.
1668 This information is expressed as a sequence of numbers, separated by
1669 commas and surrounded by braces. The @var{n}th number is 1 if
1670 register @var{n} is fixed, 0 otherwise.
1672 The table initialized from this macro, and the table initialized by
1673 the following one, may be overridden at run time either automatically,
1674 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1675 the user with the command options @option{-ffixed-@var{reg}},
1676 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1679 @defmac CALL_USED_REGISTERS
1680 @cindex call-used register
1681 @cindex call-clobbered register
1682 @cindex call-saved register
1683 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1684 clobbered (in general) by function calls as well as for fixed
1685 registers. This macro therefore identifies the registers that are not
1686 available for general allocation of values that must live across
1689 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1690 automatically saves it on function entry and restores it on function
1691 exit, if the register is used within the function.
1694 @defmac CALL_REALLY_USED_REGISTERS
1695 @cindex call-used register
1696 @cindex call-clobbered register
1697 @cindex call-saved register
1698 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1699 that the entire set of @code{FIXED_REGISTERS} be included.
1700 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1701 This macro is optional. If not specified, it defaults to the value
1702 of @code{CALL_USED_REGISTERS}.
1705 @cindex call-used register
1706 @cindex call-clobbered register
1707 @cindex call-saved register
1708 @hook TARGET_HARD_REGNO_CALL_PART_CLOBBERED
1710 @hook TARGET_REMOVE_EXTRA_CALL_PRESERVED_REGS
1712 @hook TARGET_RETURN_CALL_WITH_MAX_CLOBBERS
1714 @hook TARGET_GET_MULTILIB_ABI_NAME
1717 @findex call_used_regs
1720 @findex reg_class_contents
1721 @hook TARGET_CONDITIONAL_REGISTER_USAGE
1723 @defmac INCOMING_REGNO (@var{out})
1724 Define this macro if the target machine has register windows. This C
1725 expression returns the register number as seen by the called function
1726 corresponding to the register number @var{out} as seen by the calling
1727 function. Return @var{out} if register number @var{out} is not an
1731 @defmac OUTGOING_REGNO (@var{in})
1732 Define this macro if the target machine has register windows. This C
1733 expression returns the register number as seen by the calling function
1734 corresponding to the register number @var{in} as seen by the called
1735 function. Return @var{in} if register number @var{in} is not an inbound
1739 @defmac LOCAL_REGNO (@var{regno})
1740 Define this macro if the target machine has register windows. This C
1741 expression returns true if the register is call-saved but is in the
1742 register window. Unlike most call-saved registers, such registers
1743 need not be explicitly restored on function exit or during non-local
1748 If the program counter has a register number, define this as that
1749 register number. Otherwise, do not define it.
1752 @node Allocation Order
1753 @subsection Order of Allocation of Registers
1754 @cindex order of register allocation
1755 @cindex register allocation order
1757 @c prevent bad page break with this line
1758 Registers are allocated in order.
1760 @defmac REG_ALLOC_ORDER
1761 If defined, an initializer for a vector of integers, containing the
1762 numbers of hard registers in the order in which GCC should prefer
1763 to use them (from most preferred to least).
1765 If this macro is not defined, registers are used lowest numbered first
1766 (all else being equal).
1768 One use of this macro is on machines where the highest numbered
1769 registers must always be saved and the save-multiple-registers
1770 instruction supports only sequences of consecutive registers. On such
1771 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1772 the highest numbered allocable register first.
1775 @defmac ADJUST_REG_ALLOC_ORDER
1776 A C statement (sans semicolon) to choose the order in which to allocate
1777 hard registers for pseudo-registers local to a basic block.
1779 Store the desired register order in the array @code{reg_alloc_order}.
1780 Element 0 should be the register to allocate first; element 1, the next
1781 register; and so on.
1783 The macro body should not assume anything about the contents of
1784 @code{reg_alloc_order} before execution of the macro.
1786 On most machines, it is not necessary to define this macro.
1789 @defmac HONOR_REG_ALLOC_ORDER
1790 Normally, IRA tries to estimate the costs for saving a register in the
1791 prologue and restoring it in the epilogue. This discourages it from
1792 using call-saved registers. If a machine wants to ensure that IRA
1793 allocates registers in the order given by REG_ALLOC_ORDER even if some
1794 call-saved registers appear earlier than call-used ones, then define this
1795 macro as a C expression to nonzero. Default is 0.
1798 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
1799 In some case register allocation order is not enough for the
1800 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
1801 If this macro is defined, it should return a floating point value
1802 based on @var{regno}. The cost of using @var{regno} for a pseudo will
1803 be increased by approximately the pseudo's usage frequency times the
1804 value returned by this macro. Not defining this macro is equivalent
1805 to having it always return @code{0.0}.
1807 On most machines, it is not necessary to define this macro.
1810 @node Values in Registers
1811 @subsection How Values Fit in Registers
1813 This section discusses the macros that describe which kinds of values
1814 (specifically, which machine modes) each register can hold, and how many
1815 consecutive registers are needed for a given mode.
1817 @hook TARGET_HARD_REGNO_NREGS
1819 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
1820 A C expression that is nonzero if a value of mode @var{mode}, stored
1821 in memory, ends with padding that causes it to take up more space than
1822 in registers starting at register number @var{regno} (as determined by
1823 multiplying GCC's notion of the size of the register when containing
1824 this mode by the number of registers returned by
1825 @code{TARGET_HARD_REGNO_NREGS}). By default this is zero.
1827 For example, if a floating-point value is stored in three 32-bit
1828 registers but takes up 128 bits in memory, then this would be
1831 This macros only needs to be defined if there are cases where
1832 @code{subreg_get_info}
1833 would otherwise wrongly determine that a @code{subreg} can be
1834 represented by an offset to the register number, when in fact such a
1835 @code{subreg} would contain some of the padding not stored in
1836 registers and so not be representable.
1839 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
1840 For values of @var{regno} and @var{mode} for which
1841 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
1842 returning the greater number of registers required to hold the value
1843 including any padding. In the example above, the value would be four.
1846 @defmac REGMODE_NATURAL_SIZE (@var{mode})
1847 Define this macro if the natural size of registers that hold values
1848 of mode @var{mode} is not the word size. It is a C expression that
1849 should give the natural size in bytes for the specified mode. It is
1850 used by the register allocator to try to optimize its results. This
1851 happens for example on SPARC 64-bit where the natural size of
1852 floating-point registers is still 32-bit.
1855 @hook TARGET_HARD_REGNO_MODE_OK
1857 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
1858 A C expression that is nonzero if it is OK to rename a hard register
1859 @var{from} to another hard register @var{to}.
1861 One common use of this macro is to prevent renaming of a register to
1862 another register that is not saved by a prologue in an interrupt
1865 The default is always nonzero.
1868 @hook TARGET_MODES_TIEABLE_P
1870 @hook TARGET_HARD_REGNO_SCRATCH_OK
1872 @defmac AVOID_CCMODE_COPIES
1873 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1874 registers. You should only define this macro if support for copying to/from
1875 @code{CCmode} is incomplete.
1878 @node Leaf Functions
1879 @subsection Handling Leaf Functions
1881 @cindex leaf functions
1882 @cindex functions, leaf
1883 On some machines, a leaf function (i.e., one which makes no calls) can run
1884 more efficiently if it does not make its own register window. Often this
1885 means it is required to receive its arguments in the registers where they
1886 are passed by the caller, instead of the registers where they would
1889 The special treatment for leaf functions generally applies only when
1890 other conditions are met; for example, often they may use only those
1891 registers for its own variables and temporaries. We use the term ``leaf
1892 function'' to mean a function that is suitable for this special
1893 handling, so that functions with no calls are not necessarily ``leaf
1896 GCC assigns register numbers before it knows whether the function is
1897 suitable for leaf function treatment. So it needs to renumber the
1898 registers in order to output a leaf function. The following macros
1901 @defmac LEAF_REGISTERS
1902 Name of a char vector, indexed by hard register number, which
1903 contains 1 for a register that is allowable in a candidate for leaf
1906 If leaf function treatment involves renumbering the registers, then the
1907 registers marked here should be the ones before renumbering---those that
1908 GCC would ordinarily allocate. The registers which will actually be
1909 used in the assembler code, after renumbering, should not be marked with 1
1912 Define this macro only if the target machine offers a way to optimize
1913 the treatment of leaf functions.
1916 @defmac LEAF_REG_REMAP (@var{regno})
1917 A C expression whose value is the register number to which @var{regno}
1918 should be renumbered, when a function is treated as a leaf function.
1920 If @var{regno} is a register number which should not appear in a leaf
1921 function before renumbering, then the expression should yield @minus{}1, which
1922 will cause the compiler to abort.
1924 Define this macro only if the target machine offers a way to optimize the
1925 treatment of leaf functions, and registers need to be renumbered to do
1929 @findex current_function_is_leaf
1930 @findex current_function_uses_only_leaf_regs
1931 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
1932 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
1933 specially. They can test the C variable @code{current_function_is_leaf}
1934 which is nonzero for leaf functions. @code{current_function_is_leaf} is
1935 set prior to local register allocation and is valid for the remaining
1936 compiler passes. They can also test the C variable
1937 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
1938 functions which only use leaf registers.
1939 @code{current_function_uses_only_leaf_regs} is valid after all passes
1940 that modify the instructions have been run and is only useful if
1941 @code{LEAF_REGISTERS} is defined.
1942 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1943 @c of the next paragraph?! --mew 2feb93
1945 @node Stack Registers
1946 @subsection Registers That Form a Stack
1948 There are special features to handle computers where some of the
1949 ``registers'' form a stack. Stack registers are normally written by
1950 pushing onto the stack, and are numbered relative to the top of the
1953 Currently, GCC can only handle one group of stack-like registers, and
1954 they must be consecutively numbered. Furthermore, the existing
1955 support for stack-like registers is specific to the 80387 floating
1956 point coprocessor. If you have a new architecture that uses
1957 stack-like registers, you will need to do substantial work on
1958 @file{reg-stack.c} and write your machine description to cooperate
1959 with it, as well as defining these macros.
1962 Define this if the machine has any stack-like registers.
1965 @defmac STACK_REG_COVER_CLASS
1966 This is a cover class containing the stack registers. Define this if
1967 the machine has any stack-like registers.
1970 @defmac FIRST_STACK_REG
1971 The number of the first stack-like register. This one is the top
1975 @defmac LAST_STACK_REG
1976 The number of the last stack-like register. This one is the bottom of
1980 @node Register Classes
1981 @section Register Classes
1982 @cindex register class definitions
1983 @cindex class definitions, register
1985 On many machines, the numbered registers are not all equivalent.
1986 For example, certain registers may not be allowed for indexed addressing;
1987 certain registers may not be allowed in some instructions. These machine
1988 restrictions are described to the compiler using @dfn{register classes}.
1990 You define a number of register classes, giving each one a name and saying
1991 which of the registers belong to it. Then you can specify register classes
1992 that are allowed as operands to particular instruction patterns.
1996 In general, each register will belong to several classes. In fact, one
1997 class must be named @code{ALL_REGS} and contain all the registers. Another
1998 class must be named @code{NO_REGS} and contain no registers. Often the
1999 union of two classes will be another class; however, this is not required.
2001 @findex GENERAL_REGS
2002 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2003 terribly special about the name, but the operand constraint letters
2004 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2005 the same as @code{ALL_REGS}, just define it as a macro which expands
2008 Order the classes so that if class @var{x} is contained in class @var{y}
2009 then @var{x} has a lower class number than @var{y}.
2011 The way classes other than @code{GENERAL_REGS} are specified in operand
2012 constraints is through machine-dependent operand constraint letters.
2013 You can define such letters to correspond to various classes, then use
2014 them in operand constraints.
2016 You must define the narrowest register classes for allocatable
2017 registers, so that each class either has no subclasses, or that for
2018 some mode, the move cost between registers within the class is
2019 cheaper than moving a register in the class to or from memory
2022 You should define a class for the union of two classes whenever some
2023 instruction allows both classes. For example, if an instruction allows
2024 either a floating point (coprocessor) register or a general register for a
2025 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2026 which includes both of them. Otherwise you will get suboptimal code,
2027 or even internal compiler errors when reload cannot find a register in the
2028 class computed via @code{reg_class_subunion}.
2030 You must also specify certain redundant information about the register
2031 classes: for each class, which classes contain it and which ones are
2032 contained in it; for each pair of classes, the largest class contained
2035 When a value occupying several consecutive registers is expected in a
2036 certain class, all the registers used must belong to that class.
2037 Therefore, register classes cannot be used to enforce a requirement for
2038 a register pair to start with an even-numbered register. The way to
2039 specify this requirement is with @code{TARGET_HARD_REGNO_MODE_OK}.
2041 Register classes used for input-operands of bitwise-and or shift
2042 instructions have a special requirement: each such class must have, for
2043 each fixed-point machine mode, a subclass whose registers can transfer that
2044 mode to or from memory. For example, on some machines, the operations for
2045 single-byte values (@code{QImode}) are limited to certain registers. When
2046 this is so, each register class that is used in a bitwise-and or shift
2047 instruction must have a subclass consisting of registers from which
2048 single-byte values can be loaded or stored. This is so that
2049 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2051 @deftp {Data type} {enum reg_class}
2052 An enumerated type that must be defined with all the register class names
2053 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2054 must be the last register class, followed by one more enumerated value,
2055 @code{LIM_REG_CLASSES}, which is not a register class but rather
2056 tells how many classes there are.
2058 Each register class has a number, which is the value of casting
2059 the class name to type @code{int}. The number serves as an index
2060 in many of the tables described below.
2063 @defmac N_REG_CLASSES
2064 The number of distinct register classes, defined as follows:
2067 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2071 @defmac REG_CLASS_NAMES
2072 An initializer containing the names of the register classes as C string
2073 constants. These names are used in writing some of the debugging dumps.
2076 @defmac REG_CLASS_CONTENTS
2077 An initializer containing the contents of the register classes, as integers
2078 which are bit masks. The @var{n}th integer specifies the contents of class
2079 @var{n}. The way the integer @var{mask} is interpreted is that
2080 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2082 When the machine has more than 32 registers, an integer does not suffice.
2083 Then the integers are replaced by sub-initializers, braced groupings containing
2084 several integers. Each sub-initializer must be suitable as an initializer
2085 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2086 In this situation, the first integer in each sub-initializer corresponds to
2087 registers 0 through 31, the second integer to registers 32 through 63, and
2091 @defmac REGNO_REG_CLASS (@var{regno})
2092 A C expression whose value is a register class containing hard register
2093 @var{regno}. In general there is more than one such class; choose a class
2094 which is @dfn{minimal}, meaning that no smaller class also contains the
2098 @defmac BASE_REG_CLASS
2099 A macro whose definition is the name of the class to which a valid
2100 base register must belong. A base register is one used in an address
2101 which is the register value plus a displacement.
2104 @defmac MODE_BASE_REG_CLASS (@var{mode})
2105 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2106 the selection of a base register in a mode dependent manner. If
2107 @var{mode} is VOIDmode then it should return the same value as
2108 @code{BASE_REG_CLASS}.
2111 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2112 A C expression whose value is the register class to which a valid
2113 base register must belong in order to be used in a base plus index
2114 register address. You should define this macro if base plus index
2115 addresses have different requirements than other base register uses.
2118 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2119 A C expression whose value is the register class to which a valid
2120 base register for a memory reference in mode @var{mode} to address
2121 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2122 define the context in which the base register occurs. @var{outer_code} is
2123 the code of the immediately enclosing expression (@code{MEM} for the top level
2124 of an address, @code{ADDRESS} for something that occurs in an
2125 @code{address_operand}). @var{index_code} is the code of the corresponding
2126 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2129 @defmac INDEX_REG_CLASS
2130 A macro whose definition is the name of the class to which a valid
2131 index register must belong. An index register is one used in an
2132 address where its value is either multiplied by a scale factor or
2133 added to another register (as well as added to a displacement).
2136 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2137 A C expression which is nonzero if register number @var{num} is
2138 suitable for use as a base register in operand addresses.
2141 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2142 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2143 that expression may examine the mode of the memory reference in
2144 @var{mode}. You should define this macro if the mode of the memory
2145 reference affects whether a register may be used as a base register. If
2146 you define this macro, the compiler will use it instead of
2147 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2148 addresses that appear outside a @code{MEM}, i.e., as an
2149 @code{address_operand}.
2152 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2153 A C expression which is nonzero if register number @var{num} is suitable for
2154 use as a base register in base plus index operand addresses, accessing
2155 memory in mode @var{mode}. It may be either a suitable hard register or a
2156 pseudo register that has been allocated such a hard register. You should
2157 define this macro if base plus index addresses have different requirements
2158 than other base register uses.
2160 Use of this macro is deprecated; please use the more general
2161 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2164 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2165 A C expression which is nonzero if register number @var{num} is
2166 suitable for use as a base register in operand addresses, accessing
2167 memory in mode @var{mode} in address space @var{address_space}.
2168 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2169 that that expression may examine the context in which the register
2170 appears in the memory reference. @var{outer_code} is the code of the
2171 immediately enclosing expression (@code{MEM} if at the top level of the
2172 address, @code{ADDRESS} for something that occurs in an
2173 @code{address_operand}). @var{index_code} is the code of the
2174 corresponding index expression if @var{outer_code} is @code{PLUS};
2175 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2176 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2179 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2180 A C expression which is nonzero if register number @var{num} is
2181 suitable for use as an index register in operand addresses. It may be
2182 either a suitable hard register or a pseudo register that has been
2183 allocated such a hard register.
2185 The difference between an index register and a base register is that
2186 the index register may be scaled. If an address involves the sum of
2187 two registers, neither one of them scaled, then either one may be
2188 labeled the ``base'' and the other the ``index''; but whichever
2189 labeling is used must fit the machine's constraints of which registers
2190 may serve in each capacity. The compiler will try both labelings,
2191 looking for one that is valid, and will reload one or both registers
2192 only if neither labeling works.
2195 @hook TARGET_PREFERRED_RENAME_CLASS
2197 @hook TARGET_PREFERRED_RELOAD_CLASS
2199 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2200 A C expression that places additional restrictions on the register class
2201 to use when it is necessary to copy value @var{x} into a register in class
2202 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2203 another, smaller class. On many machines, the following definition is
2207 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2210 Sometimes returning a more restrictive class makes better code. For
2211 example, on the 68000, when @var{x} is an integer constant that is in range
2212 for a @samp{moveq} instruction, the value of this macro is always
2213 @code{DATA_REGS} as long as @var{class} includes the data registers.
2214 Requiring a data register guarantees that a @samp{moveq} will be used.
2216 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2217 @var{class} is if @var{x} is a legitimate constant which cannot be
2218 loaded into some register class. By returning @code{NO_REGS} you can
2219 force @var{x} into a memory location. For example, rs6000 can load
2220 immediate values into general-purpose registers, but does not have an
2221 instruction for loading an immediate value into a floating-point
2222 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2223 @var{x} is a floating-point constant. If the constant cannot be loaded
2224 into any kind of register, code generation will be better if
2225 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2226 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2228 If an insn has pseudos in it after register allocation, reload will go
2229 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2230 to find the best one. Returning @code{NO_REGS}, in this case, makes
2231 reload add a @code{!} in front of the constraint: the x86 back-end uses
2232 this feature to discourage usage of 387 registers when math is done in
2233 the SSE registers (and vice versa).
2236 @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2238 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2239 A C expression that places additional restrictions on the register class
2240 to use when it is necessary to be able to hold a value of mode
2241 @var{mode} in a reload register for which class @var{class} would
2244 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2245 there are certain modes that simply cannot go in certain reload classes.
2247 The value is a register class; perhaps @var{class}, or perhaps another,
2250 Don't define this macro unless the target machine has limitations which
2251 require the macro to do something nontrivial.
2254 @hook TARGET_SECONDARY_RELOAD
2256 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2257 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2258 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2259 These macros are obsolete, new ports should use the target hook
2260 @code{TARGET_SECONDARY_RELOAD} instead.
2262 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2263 target hook. Older ports still define these macros to indicate to the
2264 reload phase that it may
2265 need to allocate at least one register for a reload in addition to the
2266 register to contain the data. Specifically, if copying @var{x} to a
2267 register @var{class} in @var{mode} requires an intermediate register,
2268 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2269 largest register class all of whose registers can be used as
2270 intermediate registers or scratch registers.
2272 If copying a register @var{class} in @var{mode} to @var{x} requires an
2273 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2274 was supposed to be defined be defined to return the largest register
2275 class required. If the
2276 requirements for input and output reloads were the same, the macro
2277 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2280 The values returned by these macros are often @code{GENERAL_REGS}.
2281 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2282 can be directly copied to or from a register of @var{class} in
2283 @var{mode} without requiring a scratch register. Do not define this
2284 macro if it would always return @code{NO_REGS}.
2286 If a scratch register is required (either with or without an
2287 intermediate register), you were supposed to define patterns for
2288 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2289 (@pxref{Standard Names}. These patterns, which were normally
2290 implemented with a @code{define_expand}, should be similar to the
2291 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2294 These patterns need constraints for the reload register and scratch
2296 contain a single register class. If the original reload register (whose
2297 class is @var{class}) can meet the constraint given in the pattern, the
2298 value returned by these macros is used for the class of the scratch
2299 register. Otherwise, two additional reload registers are required.
2300 Their classes are obtained from the constraints in the insn pattern.
2302 @var{x} might be a pseudo-register or a @code{subreg} of a
2303 pseudo-register, which could either be in a hard register or in memory.
2304 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2305 in memory and the hard register number if it is in a register.
2307 These macros should not be used in the case where a particular class of
2308 registers can only be copied to memory and not to another class of
2309 registers. In that case, secondary reload registers are not needed and
2310 would not be helpful. Instead, a stack location must be used to perform
2311 the copy and the @code{mov@var{m}} pattern should use memory as an
2312 intermediate storage. This case often occurs between floating-point and
2316 @hook TARGET_SECONDARY_MEMORY_NEEDED
2318 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2319 Normally when @code{TARGET_SECONDARY_MEMORY_NEEDED} is defined, the compiler
2320 allocates a stack slot for a memory location needed for register copies.
2321 If this macro is defined, the compiler instead uses the memory location
2322 defined by this macro.
2324 Do not define this macro if you do not define
2325 @code{TARGET_SECONDARY_MEMORY_NEEDED}.
2328 @hook TARGET_SECONDARY_MEMORY_NEEDED_MODE
2330 @hook TARGET_SELECT_EARLY_REMAT_MODES
2332 @hook TARGET_CLASS_LIKELY_SPILLED_P
2334 @hook TARGET_CLASS_MAX_NREGS
2336 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2337 A C expression for the maximum number of consecutive registers
2338 of class @var{class} needed to hold a value of mode @var{mode}.
2340 This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}. In fact,
2341 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2342 should be the maximum value of @code{TARGET_HARD_REGNO_NREGS (@var{regno},
2343 @var{mode})} for all @var{regno} values in the class @var{class}.
2345 This macro helps control the handling of multiple-word values
2349 @hook TARGET_CAN_CHANGE_MODE_CLASS
2351 @hook TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS
2355 @hook TARGET_REGISTER_PRIORITY
2357 @hook TARGET_REGISTER_USAGE_LEVELING_P
2359 @hook TARGET_DIFFERENT_ADDR_DISPLACEMENT_P
2361 @hook TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P
2363 @hook TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT
2365 @hook TARGET_SPILL_CLASS
2367 @hook TARGET_ADDITIONAL_ALLOCNO_CLASS_P
2369 @hook TARGET_CSTORE_MODE
2371 @hook TARGET_COMPUTE_PRESSURE_CLASSES
2373 @node Stack and Calling
2374 @section Stack Layout and Calling Conventions
2375 @cindex calling conventions
2377 @c prevent bad page break with this line
2378 This describes the stack layout and calling conventions.
2382 * Exception Handling::
2387 * Register Arguments::
2389 * Aggregate Return::
2394 * Shrink-wrapping separate components::
2395 * Stack Smashing Protection::
2396 * Miscellaneous Register Hooks::
2400 @subsection Basic Stack Layout
2401 @cindex stack frame layout
2402 @cindex frame layout
2404 @c prevent bad page break with this line
2405 Here is the basic stack layout.
2407 @defmac STACK_GROWS_DOWNWARD
2408 Define this macro to be true if pushing a word onto the stack moves the stack
2409 pointer to a smaller address, and false otherwise.
2412 @defmac STACK_PUSH_CODE
2413 This macro defines the operation used when something is pushed
2414 on the stack. In RTL, a push operation will be
2415 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2417 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2418 and @code{POST_INC}. Which of these is correct depends on
2419 the stack direction and on whether the stack pointer points
2420 to the last item on the stack or whether it points to the
2421 space for the next item on the stack.
2423 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2424 true, which is almost always right, and @code{PRE_INC} otherwise,
2425 which is often wrong.
2428 @defmac FRAME_GROWS_DOWNWARD
2429 Define this macro to nonzero value if the addresses of local variable slots
2430 are at negative offsets from the frame pointer.
2433 @defmac ARGS_GROW_DOWNWARD
2434 Define this macro if successive arguments to a function occupy decreasing
2435 addresses on the stack.
2438 @hook TARGET_STARTING_FRAME_OFFSET
2440 @defmac STACK_ALIGNMENT_NEEDED
2441 Define to zero to disable final alignment of the stack during reload.
2442 The nonzero default for this macro is suitable for most ports.
2444 On ports where @code{TARGET_STARTING_FRAME_OFFSET} is nonzero or where there
2445 is a register save block following the local block that doesn't require
2446 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2447 stack alignment and do it in the backend.
2450 @defmac STACK_POINTER_OFFSET
2451 Offset from the stack pointer register to the first location at which
2452 outgoing arguments are placed. If not specified, the default value of
2453 zero is used. This is the proper value for most machines.
2455 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2456 the first location at which outgoing arguments are placed.
2459 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2460 Offset from the argument pointer register to the first argument's
2461 address. On some machines it may depend on the data type of the
2464 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2465 the first argument's address.
2468 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2469 Offset from the stack pointer register to an item dynamically allocated
2470 on the stack, e.g., by @code{alloca}.
2472 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2473 length of the outgoing arguments. The default is correct for most
2474 machines. See @file{function.c} for details.
2477 @defmac INITIAL_FRAME_ADDRESS_RTX
2478 A C expression whose value is RTL representing the address of the initial
2479 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2480 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
2481 default value will be used. Define this macro in order to make frame pointer
2482 elimination work in the presence of @code{__builtin_frame_address (count)} and
2483 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
2486 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2487 A C expression whose value is RTL representing the address in a stack
2488 frame where the pointer to the caller's frame is stored. Assume that
2489 @var{frameaddr} is an RTL expression for the address of the stack frame
2492 If you don't define this macro, the default is to return the value
2493 of @var{frameaddr}---that is, the stack frame address is also the
2494 address of the stack word that points to the previous frame.
2497 @defmac SETUP_FRAME_ADDRESSES
2498 A C expression that produces the machine-specific code to
2499 setup the stack so that arbitrary frames can be accessed. For example,
2500 on the SPARC, we must flush all of the register windows to the stack
2501 before we can access arbitrary stack frames. You will seldom need to
2502 define this macro. The default is to do nothing.
2505 @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
2507 @defmac FRAME_ADDR_RTX (@var{frameaddr})
2508 A C expression whose value is RTL representing the value of the frame
2509 address for the current frame. @var{frameaddr} is the frame pointer
2510 of the current frame. This is used for __builtin_frame_address.
2511 You need only define this macro if the frame address is not the same
2512 as the frame pointer. Most machines do not need to define it.
2515 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2516 A C expression whose value is RTL representing the value of the return
2517 address for the frame @var{count} steps up from the current frame, after
2518 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2519 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2520 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero.
2522 The value of the expression must always be the correct address when
2523 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
2524 determine the return address of other frames.
2527 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2528 Define this macro to nonzero value if the return address of a particular
2529 stack frame is accessed from the frame pointer of the previous stack
2530 frame. The zero default for this macro is suitable for most ports.
2533 @defmac INCOMING_RETURN_ADDR_RTX
2534 A C expression whose value is RTL representing the location of the
2535 incoming return address at the beginning of any function, before the
2536 prologue. This RTL is either a @code{REG}, indicating that the return
2537 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2540 You only need to define this macro if you want to support call frame
2541 debugging information like that provided by DWARF 2.
2543 If this RTL is a @code{REG}, you should also define
2544 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2547 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2548 A C expression whose value is an integer giving a DWARF 2 column
2549 number that may be used as an alternative return column. The column
2550 must not correspond to any gcc hard register (that is, it must not
2551 be in the range of @code{DWARF_FRAME_REGNUM}).
2553 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2554 general register, but an alternative column needs to be used for signal
2555 frames. Some targets have also used different frame return columns
2559 @defmac DWARF_ZERO_REG
2560 A C expression whose value is an integer giving a DWARF 2 register
2561 number that is considered to always have the value zero. This should
2562 only be defined if the target has an architected zero register, and
2563 someone decided it was a good idea to use that register number to
2564 terminate the stack backtrace. New ports should avoid this.
2567 @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
2569 @hook TARGET_DWARF_POLY_INDETERMINATE_VALUE
2571 @defmac INCOMING_FRAME_SP_OFFSET
2572 A C expression whose value is an integer giving the offset, in bytes,
2573 from the value of the stack pointer register to the top of the stack
2574 frame at the beginning of any function, before the prologue. The top of
2575 the frame is defined to be the value of the stack pointer in the
2576 previous frame, just before the call instruction.
2578 You only need to define this macro if you want to support call frame
2579 debugging information like that provided by DWARF 2.
2582 @defmac DEFAULT_INCOMING_FRAME_SP_OFFSET
2583 Like @code{INCOMING_FRAME_SP_OFFSET}, but must be the same for all
2584 functions of the same ABI, and when using GAS @code{.cfi_*} directives
2585 must also agree with the default CFI GAS emits. Define this macro
2586 only if @code{INCOMING_FRAME_SP_OFFSET} can have different values
2587 between different functions of the same ABI or when
2588 @code{INCOMING_FRAME_SP_OFFSET} does not agree with GAS default CFI.
2591 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2592 A C expression whose value is an integer giving the offset, in bytes,
2593 from the argument pointer to the canonical frame address (cfa). The
2594 final value should coincide with that calculated by
2595 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2596 during virtual register instantiation.
2598 The default value for this macro is
2599 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
2600 which is correct for most machines; in general, the arguments are found
2601 immediately before the stack frame. Note that this is not the case on
2602 some targets that save registers into the caller's frame, such as SPARC
2603 and rs6000, and so such targets need to define this macro.
2605 You only need to define this macro if the default is incorrect, and you
2606 want to support call frame debugging information like that provided by
2610 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
2611 If defined, a C expression whose value is an integer giving the offset
2612 in bytes from the frame pointer to the canonical frame address (cfa).
2613 The final value should coincide with that calculated by
2614 @code{INCOMING_FRAME_SP_OFFSET}.
2616 Normally the CFA is calculated as an offset from the argument pointer,
2617 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
2618 variable due to the ABI, this may not be possible. If this macro is
2619 defined, it implies that the virtual register instantiation should be
2620 based on the frame pointer instead of the argument pointer. Only one
2621 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
2625 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
2626 If defined, a C expression whose value is an integer giving the offset
2627 in bytes from the canonical frame address (cfa) to the frame base used
2628 in DWARF 2 debug information. The default is zero. A different value
2629 may reduce the size of debug information on some ports.
2632 @node Exception Handling
2633 @subsection Exception Handling Support
2634 @cindex exception handling
2636 @defmac EH_RETURN_DATA_REGNO (@var{N})
2637 A C expression whose value is the @var{N}th register number used for
2638 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2639 @var{N} registers are usable.
2641 The exception handling library routines communicate with the exception
2642 handlers via a set of agreed upon registers. Ideally these registers
2643 should be call-clobbered; it is possible to use call-saved registers,
2644 but may negatively impact code size. The target must support at least
2645 2 data registers, but should define 4 if there are enough free registers.
2647 You must define this macro if you want to support call frame exception
2648 handling like that provided by DWARF 2.
2651 @defmac EH_RETURN_STACKADJ_RTX
2652 A C expression whose value is RTL representing a location in which
2653 to store a stack adjustment to be applied before function return.
2654 This is used to unwind the stack to an exception handler's call frame.
2655 It will be assigned zero on code paths that return normally.
2657 Typically this is a call-clobbered hard register that is otherwise
2658 untouched by the epilogue, but could also be a stack slot.
2660 Do not define this macro if the stack pointer is saved and restored
2661 by the regular prolog and epilog code in the call frame itself; in
2662 this case, the exception handling library routines will update the
2663 stack location to be restored in place. Otherwise, you must define
2664 this macro if you want to support call frame exception handling like
2665 that provided by DWARF 2.
2668 @defmac EH_RETURN_HANDLER_RTX
2669 A C expression whose value is RTL representing a location in which
2670 to store the address of an exception handler to which we should
2671 return. It will not be assigned on code paths that return normally.
2673 Typically this is the location in the call frame at which the normal
2674 return address is stored. For targets that return by popping an
2675 address off the stack, this might be a memory address just below
2676 the @emph{target} call frame rather than inside the current call
2677 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2678 been assigned, so it may be used to calculate the location of the
2681 Some targets have more complex requirements than storing to an
2682 address calculable during initial code generation. In that case
2683 the @code{eh_return} instruction pattern should be used instead.
2685 If you want to support call frame exception handling, you must
2686 define either this macro or the @code{eh_return} instruction pattern.
2689 @defmac RETURN_ADDR_OFFSET
2690 If defined, an integer-valued C expression for which rtl will be generated
2691 to add it to the exception handler address before it is searched in the
2692 exception handling tables, and to subtract it again from the address before
2693 using it to return to the exception handler.
2696 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
2697 This macro chooses the encoding of pointers embedded in the exception
2698 handling sections. If at all possible, this should be defined such
2699 that the exception handling section will not require dynamic relocations,
2700 and so may be read-only.
2702 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2703 @var{global} is true if the symbol may be affected by dynamic relocations.
2704 The macro should return a combination of the @code{DW_EH_PE_*} defines
2705 as found in @file{dwarf2.h}.
2707 If this macro is not defined, pointers will not be encoded but
2708 represented directly.
2711 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2712 This macro allows the target to emit whatever special magic is required
2713 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2714 Generic code takes care of pc-relative and indirect encodings; this must
2715 be defined if the target uses text-relative or data-relative encodings.
2717 This is a C statement that branches to @var{done} if the format was
2718 handled. @var{encoding} is the format chosen, @var{size} is the number
2719 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2723 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
2724 This macro allows the target to add CPU and operating system specific
2725 code to the call-frame unwinder for use when there is no unwind data
2726 available. The most common reason to implement this macro is to unwind
2727 through signal frames.
2729 This macro is called from @code{uw_frame_state_for} in
2730 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
2731 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2732 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
2733 for the address of the code being executed and @code{context->cfa} for
2734 the stack pointer value. If the frame can be decoded, the register
2735 save addresses should be updated in @var{fs} and the macro should
2736 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
2737 the macro should evaluate to @code{_URC_END_OF_STACK}.
2739 For proper signal handling in Java this macro is accompanied by
2740 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
2743 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
2744 This macro allows the target to add operating system specific code to the
2745 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
2746 usually used for signal or interrupt frames.
2748 This macro is called from @code{uw_update_context} in libgcc's
2749 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2750 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
2751 for the abi and context in the @code{.unwabi} directive. If the
2752 @code{.unwabi} directive can be handled, the register save addresses should
2753 be updated in @var{fs}.
2756 @defmac TARGET_USES_WEAK_UNWIND_INFO
2757 A C expression that evaluates to true if the target requires unwind
2758 info to be given comdat linkage. Define it to be @code{1} if comdat
2759 linkage is necessary. The default is @code{0}.
2762 @node Stack Checking
2763 @subsection Specifying How Stack Checking is Done
2765 GCC will check that stack references are within the boundaries of the
2766 stack, if the option @option{-fstack-check} is specified, in one of
2771 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2772 will assume that you have arranged for full stack checking to be done
2773 at appropriate places in the configuration files. GCC will not do
2774 other special processing.
2777 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
2778 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
2779 that you have arranged for static stack checking (checking of the
2780 static stack frame of functions) to be done at appropriate places
2781 in the configuration files. GCC will only emit code to do dynamic
2782 stack checking (checking on dynamic stack allocations) using the third
2786 If neither of the above are true, GCC will generate code to periodically
2787 ``probe'' the stack pointer using the values of the macros defined below.
2790 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
2791 GCC will change its allocation strategy for large objects if the option
2792 @option{-fstack-check} is specified: they will always be allocated
2793 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
2795 @defmac STACK_CHECK_BUILTIN
2796 A nonzero value if stack checking is done by the configuration files in a
2797 machine-dependent manner. You should define this macro if stack checking
2798 is required by the ABI of your machine or if you would like to do stack
2799 checking in some more efficient way than the generic approach. The default
2800 value of this macro is zero.
2803 @defmac STACK_CHECK_STATIC_BUILTIN
2804 A nonzero value if static stack checking is done by the configuration files
2805 in a machine-dependent manner. You should define this macro if you would
2806 like to do static stack checking in some more efficient way than the generic
2807 approach. The default value of this macro is zero.
2810 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
2811 An integer specifying the interval at which GCC must generate stack probe
2812 instructions, defined as 2 raised to this integer. You will normally
2813 define this macro so that the interval be no larger than the size of
2814 the ``guard pages'' at the end of a stack area. The default value
2815 of 12 (4096-byte interval) is suitable for most systems.
2818 @defmac STACK_CHECK_MOVING_SP
2819 An integer which is nonzero if GCC should move the stack pointer page by page
2820 when doing probes. This can be necessary on systems where the stack pointer
2821 contains the bottom address of the memory area accessible to the executing
2822 thread at any point in time. In this situation an alternate signal stack
2823 is required in order to be able to recover from a stack overflow. The
2824 default value of this macro is zero.
2827 @defmac STACK_CHECK_PROTECT
2828 The number of bytes of stack needed to recover from a stack overflow, for
2829 languages where such a recovery is supported. The default value of 4KB/8KB
2830 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
2831 8KB/12KB with other exception handling mechanisms should be adequate for most
2832 architectures and operating systems.
2835 The following macros are relevant only if neither STACK_CHECK_BUILTIN
2836 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
2837 in the opposite case.
2839 @defmac STACK_CHECK_MAX_FRAME_SIZE
2840 The maximum size of a stack frame, in bytes. GCC will generate probe
2841 instructions in non-leaf functions to ensure at least this many bytes of
2842 stack are available. If a stack frame is larger than this size, stack
2843 checking will not be reliable and GCC will issue a warning. The
2844 default is chosen so that GCC only generates one instruction on most
2845 systems. You should normally not change the default value of this macro.
2848 @defmac STACK_CHECK_FIXED_FRAME_SIZE
2849 GCC uses this value to generate the above warning message. It
2850 represents the amount of fixed frame used by a function, not including
2851 space for any callee-saved registers, temporaries and user variables.
2852 You need only specify an upper bound for this amount and will normally
2853 use the default of four words.
2856 @defmac STACK_CHECK_MAX_VAR_SIZE
2857 The maximum size, in bytes, of an object that GCC will place in the
2858 fixed area of the stack frame when the user specifies
2859 @option{-fstack-check}.
2860 GCC computed the default from the values of the above macros and you will
2861 normally not need to override that default.
2864 @hook TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE
2867 @node Frame Registers
2868 @subsection Registers That Address the Stack Frame
2870 @c prevent bad page break with this line
2871 This discusses registers that address the stack frame.
2873 @defmac STACK_POINTER_REGNUM
2874 The register number of the stack pointer register, which must also be a
2875 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2876 the hardware determines which register this is.
2879 @defmac FRAME_POINTER_REGNUM
2880 The register number of the frame pointer register, which is used to
2881 access automatic variables in the stack frame. On some machines, the
2882 hardware determines which register this is. On other machines, you can
2883 choose any register you wish for this purpose.
2886 @defmac HARD_FRAME_POINTER_REGNUM
2887 On some machines the offset between the frame pointer and starting
2888 offset of the automatic variables is not known until after register
2889 allocation has been done (for example, because the saved registers are
2890 between these two locations). On those machines, define
2891 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2892 be used internally until the offset is known, and define
2893 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2894 used for the frame pointer.
2896 You should define this macro only in the very rare circumstances when it
2897 is not possible to calculate the offset between the frame pointer and
2898 the automatic variables until after register allocation has been
2899 completed. When this macro is defined, you must also indicate in your
2900 definition of @code{ELIMINABLE_REGS} how to eliminate
2901 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2902 or @code{STACK_POINTER_REGNUM}.
2904 Do not define this macro if it would be the same as
2905 @code{FRAME_POINTER_REGNUM}.
2908 @defmac ARG_POINTER_REGNUM
2909 The register number of the arg pointer register, which is used to access
2910 the function's argument list. On some machines, this is the same as the
2911 frame pointer register. On some machines, the hardware determines which
2912 register this is. On other machines, you can choose any register you
2913 wish for this purpose. If this is not the same register as the frame
2914 pointer register, then you must mark it as a fixed register according to
2915 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2916 (@pxref{Elimination}).
2919 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
2920 Define this to a preprocessor constant that is nonzero if
2921 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
2922 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
2923 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
2924 definition is not suitable for use in preprocessor conditionals.
2927 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
2928 Define this to a preprocessor constant that is nonzero if
2929 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
2930 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
2931 ARG_POINTER_REGNUM)}; you only need to define this macro if that
2932 definition is not suitable for use in preprocessor conditionals.
2935 @defmac RETURN_ADDRESS_POINTER_REGNUM
2936 The register number of the return address pointer register, which is used to
2937 access the current function's return address from the stack. On some
2938 machines, the return address is not at a fixed offset from the frame
2939 pointer or stack pointer or argument pointer. This register can be defined
2940 to point to the return address on the stack, and then be converted by
2941 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2943 Do not define this macro unless there is no other way to get the return
2944 address from the stack.
2947 @defmac STATIC_CHAIN_REGNUM
2948 @defmacx STATIC_CHAIN_INCOMING_REGNUM
2949 Register numbers used for passing a function's static chain pointer. If
2950 register windows are used, the register number as seen by the called
2951 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2952 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2953 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2956 The static chain register need not be a fixed register.
2958 If the static chain is passed in memory, these macros should not be
2959 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
2962 @hook TARGET_STATIC_CHAIN
2964 @defmac DWARF_FRAME_REGISTERS
2965 This macro specifies the maximum number of hard registers that can be
2966 saved in a call frame. This is used to size data structures used in
2967 DWARF2 exception handling.
2969 Prior to GCC 3.0, this macro was needed in order to establish a stable
2970 exception handling ABI in the face of adding new hard registers for ISA
2971 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
2972 in the number of hard registers. Nevertheless, this macro can still be
2973 used to reduce the runtime memory requirements of the exception handling
2974 routines, which can be substantial if the ISA contains a lot of
2975 registers that are not call-saved.
2977 If this macro is not defined, it defaults to
2978 @code{FIRST_PSEUDO_REGISTER}.
2981 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
2983 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
2984 for backward compatibility in pre GCC 3.0 compiled code.
2986 If this macro is not defined, it defaults to
2987 @code{DWARF_FRAME_REGISTERS}.
2990 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
2992 Define this macro if the target's representation for dwarf registers
2993 is different than the internal representation for unwind column.
2994 Given a dwarf register, this macro should return the internal unwind
2995 column number to use instead.
2998 @defmac DWARF_FRAME_REGNUM (@var{regno})
3000 Define this macro if the target's representation for dwarf registers
3001 used in .eh_frame or .debug_frame is different from that used in other
3002 debug info sections. Given a GCC hard register number, this macro
3003 should return the .eh_frame register number. The default is
3004 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3008 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3010 Define this macro to map register numbers held in the call frame info
3011 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3012 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3013 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3014 return @code{@var{regno}}.
3018 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3020 Define this macro if the target stores register values as
3021 @code{_Unwind_Word} type in unwind context. It should be defined if
3022 target register size is larger than the size of @code{void *}. The
3023 default is to store register values as @code{void *} type.
3027 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3029 Define this macro to be 1 if the target always uses extended unwind
3030 context with version, args_size and by_value fields. If it is undefined,
3031 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3032 defined and 0 otherwise.
3036 @defmac DWARF_LAZY_REGISTER_VALUE (@var{regno}, @var{value})
3037 Define this macro if the target has pseudo DWARF registers whose
3038 values need to be computed lazily on demand by the unwinder (such as when
3039 referenced in a CFA expression). The macro returns true if @var{regno}
3040 is such a register and stores its value in @samp{*@var{value}} if so.
3044 @subsection Eliminating Frame Pointer and Arg Pointer
3046 @c prevent bad page break with this line
3047 This is about eliminating the frame pointer and arg pointer.
3049 @hook TARGET_FRAME_POINTER_REQUIRED
3051 @defmac ELIMINABLE_REGS
3052 This macro specifies a table of register pairs used to eliminate
3053 unneeded registers that point into the stack frame.
3055 The definition of this macro is a list of structure initializations, each
3056 of which specifies an original and replacement register.
3058 On some machines, the position of the argument pointer is not known until
3059 the compilation is completed. In such a case, a separate hard register
3060 must be used for the argument pointer. This register can be eliminated by
3061 replacing it with either the frame pointer or the argument pointer,
3062 depending on whether or not the frame pointer has been eliminated.
3064 In this case, you might specify:
3066 #define ELIMINABLE_REGS \
3067 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3068 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3069 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3072 Note that the elimination of the argument pointer with the stack pointer is
3073 specified first since that is the preferred elimination.
3076 @hook TARGET_CAN_ELIMINATE
3078 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3079 This macro returns the initial difference between the specified pair
3080 of registers. The value would be computed from information
3081 such as the result of @code{get_frame_size ()} and the tables of
3082 registers @code{df_regs_ever_live_p} and @code{call_used_regs}.
3085 @hook TARGET_COMPUTE_FRAME_LAYOUT
3087 @node Stack Arguments
3088 @subsection Passing Function Arguments on the Stack
3089 @cindex arguments on stack
3090 @cindex stack arguments
3092 The macros in this section control how arguments are passed
3093 on the stack. See the following section for other macros that
3094 control passing certain arguments in registers.
3096 @hook TARGET_PROMOTE_PROTOTYPES
3099 A C expression. If nonzero, push insns will be used to pass
3101 If the target machine does not have a push instruction, set it to zero.
3102 That directs GCC to use an alternate strategy: to
3103 allocate the entire argument block and then store the arguments into
3104 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3107 @defmac PUSH_ARGS_REVERSED
3108 A C expression. If nonzero, function arguments will be evaluated from
3109 last to first, rather than from first to last. If this macro is not
3110 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3111 and args grow in opposite directions, and 0 otherwise.
3114 @defmac PUSH_ROUNDING (@var{npushed})
3115 A C expression that is the number of bytes actually pushed onto the
3116 stack when an instruction attempts to push @var{npushed} bytes.
3118 On some machines, the definition
3121 #define PUSH_ROUNDING(BYTES) (BYTES)
3125 will suffice. But on other machines, instructions that appear
3126 to push one byte actually push two bytes in an attempt to maintain
3127 alignment. Then the definition should be
3130 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3133 If the value of this macro has a type, it should be an unsigned type.
3136 @findex outgoing_args_size
3137 @findex crtl->outgoing_args_size
3138 @defmac ACCUMULATE_OUTGOING_ARGS
3139 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3140 will be computed and placed into
3141 @code{crtl->outgoing_args_size}. No space will be pushed
3142 onto the stack for each call; instead, the function prologue should
3143 increase the stack frame size by this amount.
3145 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3149 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3150 Define this macro if functions should assume that stack space has been
3151 allocated for arguments even when their values are passed in
3154 The value of this macro is the size, in bytes, of the area reserved for
3155 arguments passed in registers for the function represented by @var{fndecl},
3156 which can be zero if GCC is calling a library function.
3157 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3160 This space can be allocated by the caller, or be a part of the
3161 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3164 @c above is overfull. not sure what to do. --mew 5feb93 did
3165 @c something, not sure if it looks good. --mew 10feb93
3167 @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3168 Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3169 Define this macro if space guaranteed when compiling a function body
3170 is different to space required when making a call, a situation that
3171 can arise with K&R style function definitions.
3174 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3175 Define this to a nonzero value if it is the responsibility of the
3176 caller to allocate the area reserved for arguments passed in registers
3177 when calling a function of @var{fntype}. @var{fntype} may be NULL
3178 if the function called is a library function.
3180 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3181 whether the space for these arguments counts in the value of
3182 @code{crtl->outgoing_args_size}.
3185 @defmac STACK_PARMS_IN_REG_PARM_AREA
3186 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3187 stack parameters don't skip the area specified by it.
3188 @c i changed this, makes more sens and it should have taken care of the
3189 @c overfull.. not as specific, tho. --mew 5feb93
3191 Normally, when a parameter is not passed in registers, it is placed on the
3192 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3193 suppresses this behavior and causes the parameter to be passed on the
3194 stack in its natural location.
3197 @hook TARGET_RETURN_POPS_ARGS
3199 @defmac CALL_POPS_ARGS (@var{cum})
3200 A C expression that should indicate the number of bytes a call sequence
3201 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3202 when compiling a function call.
3204 @var{cum} is the variable in which all arguments to the called function
3205 have been accumulated.
3207 On certain architectures, such as the SH5, a call trampoline is used
3208 that pops certain registers off the stack, depending on the arguments
3209 that have been passed to the function. Since this is a property of the
3210 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3214 @node Register Arguments
3215 @subsection Passing Arguments in Registers
3216 @cindex arguments in registers
3217 @cindex registers arguments
3219 This section describes the macros which let you control how various
3220 types of arguments are passed in registers or how they are arranged in
3223 @hook TARGET_FUNCTION_ARG
3225 @hook TARGET_MUST_PASS_IN_STACK
3227 @hook TARGET_FUNCTION_INCOMING_ARG
3229 @hook TARGET_USE_PSEUDO_PIC_REG
3231 @hook TARGET_INIT_PIC_REG
3233 @hook TARGET_ARG_PARTIAL_BYTES
3235 @hook TARGET_PASS_BY_REFERENCE
3237 @hook TARGET_CALLEE_COPIES
3239 @defmac CUMULATIVE_ARGS
3240 A C type for declaring a variable that is used as the first argument
3241 of @code{TARGET_FUNCTION_ARG} and other related values. For some
3242 target machines, the type @code{int} suffices and can hold the number
3243 of bytes of argument so far.
3245 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3246 arguments that have been passed on the stack. The compiler has other
3247 variables to keep track of that. For target machines on which all
3248 arguments are passed on the stack, there is no need to store anything in
3249 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3250 should not be empty, so use @code{int}.
3253 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
3254 If defined, this macro is called before generating any code for a
3255 function, but after the @var{cfun} descriptor for the function has been
3256 created. The back end may use this macro to update @var{cfun} to
3257 reflect an ABI other than that which would normally be used by default.
3258 If the compiler is generating code for a compiler-generated function,
3259 @var{fndecl} may be @code{NULL}.
3262 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3263 A C statement (sans semicolon) for initializing the variable
3264 @var{cum} for the state at the beginning of the argument list. The
3265 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3266 is the tree node for the data type of the function which will receive
3267 the args, or 0 if the args are to a compiler support library function.
3268 For direct calls that are not libcalls, @var{fndecl} contain the
3269 declaration node of the function. @var{fndecl} is also set when
3270 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3271 being compiled. @var{n_named_args} is set to the number of named
3272 arguments, including a structure return address if it is passed as a
3273 parameter, when making a call. When processing incoming arguments,
3274 @var{n_named_args} is set to @minus{}1.
3276 When processing a call to a compiler support library function,
3277 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3278 contains the name of the function, as a string. @var{libname} is 0 when
3279 an ordinary C function call is being processed. Thus, each time this
3280 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3281 never both of them at once.
3284 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3285 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3286 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3287 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3288 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3289 0)} is used instead.
3292 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3293 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3294 finding the arguments for the function being compiled. If this macro is
3295 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3297 The value passed for @var{libname} is always 0, since library routines
3298 with special calling conventions are never compiled with GCC@. The
3299 argument @var{libname} exists for symmetry with
3300 @code{INIT_CUMULATIVE_ARGS}.
3301 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3302 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3305 @hook TARGET_FUNCTION_ARG_ADVANCE
3307 @hook TARGET_FUNCTION_ARG_OFFSET
3309 @hook TARGET_FUNCTION_ARG_PADDING
3311 @defmac PAD_VARARGS_DOWN
3312 If defined, a C expression which determines whether the default
3313 implementation of va_arg will attempt to pad down before reading the
3314 next argument, if that argument is smaller than its aligned space as
3315 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3316 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3319 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3320 Specify padding for the last element of a block move between registers and
3321 memory. @var{first} is nonzero if this is the only element. Defining this
3322 macro allows better control of register function parameters on big-endian
3323 machines, without using @code{PARALLEL} rtl. In particular,
3324 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3325 registers, as there is no longer a "wrong" part of a register; For example,
3326 a three byte aggregate may be passed in the high part of a register if so
3330 @hook TARGET_FUNCTION_ARG_BOUNDARY
3332 @hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY
3334 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3335 A C expression that is nonzero if @var{regno} is the number of a hard
3336 register in which function arguments are sometimes passed. This does
3337 @emph{not} include implicit arguments such as the static chain and
3338 the structure-value address. On many machines, no registers can be
3339 used for this purpose since all function arguments are pushed on the
3343 @hook TARGET_SPLIT_COMPLEX_ARG
3345 @hook TARGET_BUILD_BUILTIN_VA_LIST
3347 @hook TARGET_ENUM_VA_LIST_P
3349 @hook TARGET_FN_ABI_VA_LIST
3351 @hook TARGET_CANONICAL_VA_LIST_TYPE
3353 @hook TARGET_GIMPLIFY_VA_ARG_EXPR
3355 @hook TARGET_VALID_POINTER_MODE
3357 @hook TARGET_REF_MAY_ALIAS_ERRNO
3359 @hook TARGET_TRANSLATE_MODE_ATTRIBUTE
3361 @hook TARGET_SCALAR_MODE_SUPPORTED_P
3363 @hook TARGET_VECTOR_MODE_SUPPORTED_P
3365 @hook TARGET_ARRAY_MODE
3367 @hook TARGET_ARRAY_MODE_SUPPORTED_P
3369 @hook TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P
3371 @hook TARGET_FLOATN_MODE
3373 @hook TARGET_FLOATN_BUILTIN_P
3375 @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
3378 @subsection How Scalar Function Values Are Returned
3379 @cindex return values in registers
3380 @cindex values, returned by functions
3381 @cindex scalars, returned as values
3383 This section discusses the macros that control returning scalars as
3384 values---values that can fit in registers.
3386 @hook TARGET_FUNCTION_VALUE
3388 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3389 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
3390 a new target instead.
3393 @defmac LIBCALL_VALUE (@var{mode})
3394 A C expression to create an RTX representing the place where a library
3395 function returns a value of mode @var{mode}.
3397 Note that ``library function'' in this context means a compiler
3398 support routine, used to perform arithmetic, whose name is known
3399 specially by the compiler and was not mentioned in the C code being
3403 @hook TARGET_LIBCALL_VALUE
3405 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3406 A C expression that is nonzero if @var{regno} is the number of a hard
3407 register in which the values of called function may come back.
3409 A register whose use for returning values is limited to serving as the
3410 second of a pair (for a value of type @code{double}, say) need not be
3411 recognized by this macro. So for most machines, this definition
3415 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3418 If the machine has register windows, so that the caller and the called
3419 function use different registers for the return value, this macro
3420 should recognize only the caller's register numbers.
3422 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
3423 for a new target instead.
3426 @hook TARGET_FUNCTION_VALUE_REGNO_P
3428 @defmac APPLY_RESULT_SIZE
3429 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3430 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3431 saving and restoring an arbitrary return value.
3434 @hook TARGET_OMIT_STRUCT_RETURN_REG
3436 @hook TARGET_RETURN_IN_MSB
3438 @node Aggregate Return
3439 @subsection How Large Values Are Returned
3440 @cindex aggregates as return values
3441 @cindex large return values
3442 @cindex returning aggregate values
3443 @cindex structure value address
3445 When a function value's mode is @code{BLKmode} (and in some other
3446 cases), the value is not returned according to
3447 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
3448 caller passes the address of a block of memory in which the value
3449 should be stored. This address is called the @dfn{structure value
3452 This section describes how to control returning structure values in
3455 @hook TARGET_RETURN_IN_MEMORY
3457 @defmac DEFAULT_PCC_STRUCT_RETURN
3458 Define this macro to be 1 if all structure and union return values must be
3459 in memory. Since this results in slower code, this should be defined
3460 only if needed for compatibility with other compilers or with an ABI@.
3461 If you define this macro to be 0, then the conventions used for structure
3462 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
3465 If not defined, this defaults to the value 1.
3468 @hook TARGET_STRUCT_VALUE_RTX
3470 @defmac PCC_STATIC_STRUCT_RETURN
3471 Define this macro if the usual system convention on the target machine
3472 for returning structures and unions is for the called function to return
3473 the address of a static variable containing the value.
3475 Do not define this if the usual system convention is for the caller to
3476 pass an address to the subroutine.
3478 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3479 nothing when you use @option{-freg-struct-return} mode.
3482 @hook TARGET_GET_RAW_RESULT_MODE
3484 @hook TARGET_GET_RAW_ARG_MODE
3486 @hook TARGET_EMPTY_RECORD_P
3488 @hook TARGET_WARN_PARAMETER_PASSING_ABI
3491 @subsection Caller-Saves Register Allocation
3493 If you enable it, GCC can save registers around function calls. This
3494 makes it possible to use call-clobbered registers to hold variables that
3495 must live across calls.
3497 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3498 A C expression specifying which mode is required for saving @var{nregs}
3499 of a pseudo-register in call-clobbered hard register @var{regno}. If
3500 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3501 returned. For most machines this macro need not be defined since GCC
3502 will select the smallest suitable mode.
3505 @node Function Entry
3506 @subsection Function Entry and Exit
3507 @cindex function entry and exit
3511 This section describes the macros that output function entry
3512 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3514 @hook TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY
3516 @hook TARGET_ASM_FUNCTION_PROLOGUE
3518 @hook TARGET_ASM_FUNCTION_END_PROLOGUE
3520 @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
3522 @hook TARGET_ASM_FUNCTION_EPILOGUE
3526 @findex pretend_args_size
3527 @findex crtl->args.pretend_args_size
3528 A region of @code{crtl->args.pretend_args_size} bytes of
3529 uninitialized space just underneath the first argument arriving on the
3530 stack. (This may not be at the very start of the allocated stack region
3531 if the calling sequence has pushed anything else since pushing the stack
3532 arguments. But usually, on such machines, nothing else has been pushed
3533 yet, because the function prologue itself does all the pushing.) This
3534 region is used on machines where an argument may be passed partly in
3535 registers and partly in memory, and, in some cases to support the
3536 features in @code{<stdarg.h>}.
3539 An area of memory used to save certain registers used by the function.
3540 The size of this area, which may also include space for such things as
3541 the return address and pointers to previous stack frames, is
3542 machine-specific and usually depends on which registers have been used
3543 in the function. Machines with register windows often do not require
3547 A region of at least @var{size} bytes, possibly rounded up to an allocation
3548 boundary, to contain the local variables of the function. On some machines,
3549 this region and the save area may occur in the opposite order, with the
3550 save area closer to the top of the stack.
3553 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3554 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3555 @code{crtl->outgoing_args_size} bytes to be used for outgoing
3556 argument lists of the function. @xref{Stack Arguments}.
3559 @defmac EXIT_IGNORE_STACK
3560 Define this macro as a C expression that is nonzero if the return
3561 instruction or the function epilogue ignores the value of the stack
3562 pointer; in other words, if it is safe to delete an instruction to
3563 adjust the stack pointer before a return from the function. The
3566 Note that this macro's value is relevant only for functions for which
3567 frame pointers are maintained. It is never safe to delete a final
3568 stack adjustment in a function that has no frame pointer, and the
3569 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3572 @defmac EPILOGUE_USES (@var{regno})
3573 Define this macro as a C expression that is nonzero for registers that are
3574 used by the epilogue or the @samp{return} pattern. The stack and frame
3575 pointer registers are already assumed to be used as needed.
3578 @defmac EH_USES (@var{regno})
3579 Define this macro as a C expression that is nonzero for registers that are
3580 used by the exception handling mechanism, and so should be considered live
3581 on entry to an exception edge.
3584 @hook TARGET_ASM_OUTPUT_MI_THUNK
3586 @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
3589 @subsection Generating Code for Profiling
3590 @cindex profiling, code generation
3592 These macros will help you generate code for profiling.
3594 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
3595 A C statement or compound statement to output to @var{file} some
3596 assembler code to call the profiling subroutine @code{mcount}.
3599 The details of how @code{mcount} expects to be called are determined by
3600 your operating system environment, not by GCC@. To figure them out,
3601 compile a small program for profiling using the system's installed C
3602 compiler and look at the assembler code that results.
3604 Older implementations of @code{mcount} expect the address of a counter
3605 variable to be loaded into some register. The name of this variable is
3606 @samp{LP} followed by the number @var{labelno}, so you would generate
3607 the name using @samp{LP%d} in a @code{fprintf}.
3610 @defmac PROFILE_HOOK
3611 A C statement or compound statement to output to @var{file} some assembly
3612 code to call the profiling subroutine @code{mcount} even the target does
3613 not support profiling.
3616 @defmac NO_PROFILE_COUNTERS
3617 Define this macro to be an expression with a nonzero value if the
3618 @code{mcount} subroutine on your system does not need a counter variable
3619 allocated for each function. This is true for almost all modern
3620 implementations. If you define this macro, you must not use the
3621 @var{labelno} argument to @code{FUNCTION_PROFILER}.
3624 @defmac PROFILE_BEFORE_PROLOGUE
3625 Define this macro if the code for function profiling should come before
3626 the function prologue. Normally, the profiling code comes after.
3629 @hook TARGET_KEEP_LEAF_WHEN_PROFILED
3632 @subsection Permitting tail calls
3635 @hook TARGET_FUNCTION_OK_FOR_SIBCALL
3637 @hook TARGET_EXTRA_LIVE_ON_ENTRY
3639 @hook TARGET_SET_UP_BY_PROLOGUE
3641 @hook TARGET_WARN_FUNC_RETURN
3643 @node Shrink-wrapping separate components
3644 @subsection Shrink-wrapping separate components
3645 @cindex shrink-wrapping separate components
3647 The prologue may perform a variety of target dependent tasks such as
3648 saving callee-saved registers, saving the return address, aligning the
3649 stack, creating a stack frame, initializing the PIC register, setting
3650 up the static chain, etc.
3652 On some targets some of these tasks may be independent of others and
3653 thus may be shrink-wrapped separately. These independent tasks are
3654 referred to as components and are handled generically by the target
3655 independent parts of GCC.
3657 Using the following hooks those prologue or epilogue components can be
3658 shrink-wrapped separately, so that the initialization (and possibly
3659 teardown) those components do is not done as frequently on execution
3660 paths where this would unnecessary.
3662 What exactly those components are is up to the target code; the generic
3663 code treats them abstractly, as a bit in an @code{sbitmap}. These
3664 @code{sbitmap}s are allocated by the @code{shrink_wrap.get_separate_components}
3665 and @code{shrink_wrap.components_for_bb} hooks, and deallocated by the
3668 @hook TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS
3670 @hook TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB
3672 @hook TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS
3674 @hook TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS
3676 @hook TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS
3678 @hook TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS
3680 @node Stack Smashing Protection
3681 @subsection Stack smashing protection
3682 @cindex stack smashing protection
3684 @hook TARGET_STACK_PROTECT_GUARD
3686 @hook TARGET_STACK_PROTECT_FAIL
3688 @hook TARGET_STACK_PROTECT_RUNTIME_ENABLED_P
3690 @hook TARGET_SUPPORTS_SPLIT_STACK
3692 @hook TARGET_GET_VALID_OPTION_VALUES
3694 @node Miscellaneous Register Hooks
3695 @subsection Miscellaneous register hooks
3696 @cindex miscellaneous register hooks
3698 @hook TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
3701 @section Implementing the Varargs Macros
3702 @cindex varargs implementation
3704 GCC comes with an implementation of @code{<varargs.h>} and
3705 @code{<stdarg.h>} that work without change on machines that pass arguments
3706 on the stack. Other machines require their own implementations of
3707 varargs, and the two machine independent header files must have
3708 conditionals to include it.
3710 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
3711 the calling convention for @code{va_start}. The traditional
3712 implementation takes just one argument, which is the variable in which
3713 to store the argument pointer. The ISO implementation of
3714 @code{va_start} takes an additional second argument. The user is
3715 supposed to write the last named argument of the function here.
3717 However, @code{va_start} should not use this argument. The way to find
3718 the end of the named arguments is with the built-in functions described
3721 @defmac __builtin_saveregs ()
3722 Use this built-in function to save the argument registers in memory so
3723 that the varargs mechanism can access them. Both ISO and traditional
3724 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3725 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
3727 On some machines, @code{__builtin_saveregs} is open-coded under the
3728 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
3729 other machines, it calls a routine written in assembler language,
3730 found in @file{libgcc2.c}.
3732 Code generated for the call to @code{__builtin_saveregs} appears at the
3733 beginning of the function, as opposed to where the call to
3734 @code{__builtin_saveregs} is written, regardless of what the code is.
3735 This is because the registers must be saved before the function starts
3736 to use them for its own purposes.
3737 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3741 @defmac __builtin_next_arg (@var{lastarg})
3742 This builtin returns the address of the first anonymous stack
3743 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3744 returns the address of the location above the first anonymous stack
3745 argument. Use it in @code{va_start} to initialize the pointer for
3746 fetching arguments from the stack. Also use it in @code{va_start} to
3747 verify that the second parameter @var{lastarg} is the last named argument
3748 of the current function.
3751 @defmac __builtin_classify_type (@var{object})
3752 Since each machine has its own conventions for which data types are
3753 passed in which kind of register, your implementation of @code{va_arg}
3754 has to embody these conventions. The easiest way to categorize the
3755 specified data type is to use @code{__builtin_classify_type} together
3756 with @code{sizeof} and @code{__alignof__}.
3758 @code{__builtin_classify_type} ignores the value of @var{object},
3759 considering only its data type. It returns an integer describing what
3760 kind of type that is---integer, floating, pointer, structure, and so on.
3762 The file @file{typeclass.h} defines an enumeration that you can use to
3763 interpret the values of @code{__builtin_classify_type}.
3766 These machine description macros help implement varargs:
3768 @hook TARGET_EXPAND_BUILTIN_SAVEREGS
3770 @hook TARGET_SETUP_INCOMING_VARARGS
3772 @hook TARGET_STRICT_ARGUMENT_NAMING
3774 @hook TARGET_CALL_ARGS
3776 @hook TARGET_END_CALL_ARGS
3778 @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
3780 @hook TARGET_LOAD_BOUNDS_FOR_ARG
3782 @hook TARGET_STORE_BOUNDS_FOR_ARG
3784 @hook TARGET_LOAD_RETURNED_BOUNDS
3786 @hook TARGET_STORE_RETURNED_BOUNDS
3788 @hook TARGET_SETUP_INCOMING_VARARG_BOUNDS
3791 @section Support for Nested Functions
3792 @cindex support for nested functions
3793 @cindex trampolines for nested functions
3794 @cindex descriptors for nested functions
3795 @cindex nested functions, support for
3797 Taking the address of a nested function requires special compiler
3798 handling to ensure that the static chain register is loaded when
3799 the function is invoked via an indirect call.
3801 GCC has traditionally supported nested functions by creating an
3802 executable @dfn{trampoline} at run time when the address of a nested
3803 function is taken. This is a small piece of code which normally
3804 resides on the stack, in the stack frame of the containing function.
3805 The trampoline loads the static chain register and then jumps to the
3806 real address of the nested function.
3808 The use of trampolines requires an executable stack, which is a
3809 security risk. To avoid this problem, GCC also supports another
3810 strategy: using descriptors for nested functions. Under this model,
3811 taking the address of a nested function results in a pointer to a
3812 non-executable function descriptor object. Initializing the static chain
3813 from the descriptor is handled at indirect call sites.
3815 On some targets, including HPPA and IA-64, function descriptors may be
3816 mandated by the ABI or be otherwise handled in a target-specific way
3817 by the back end in its code generation strategy for indirect calls.
3818 GCC also provides its own generic descriptor implementation to support the
3819 @option{-fno-trampolines} option. In this case runtime detection of
3820 function descriptors at indirect call sites relies on descriptor
3821 pointers being tagged with a bit that is never set in bare function
3822 addresses. Since GCC's generic function descriptors are
3823 not ABI-compliant, this option is typically used only on a
3824 per-language basis (notably by Ada) or when it can otherwise be
3825 applied to the whole program.
3827 Define the following hook if your backend either implements ABI-specified
3828 descriptor support, or can use GCC's generic descriptor implementation
3829 for nested functions.
3831 @hook TARGET_CUSTOM_FUNCTION_DESCRIPTORS
3833 The following macros tell GCC how to generate code to allocate and
3834 initialize an executable trampoline. You can also use this interface
3835 if your back end needs to create ABI-specified non-executable descriptors; in
3836 this case the "trampoline" created is the descriptor containing data only.
3838 The instructions in an executable trampoline must do two things: load
3839 a constant address into the static chain register, and jump to the real
3840 address of the nested function. On CISC machines such as the m68k,
3841 this requires two instructions, a move immediate and a jump. Then the
3842 two addresses exist in the trampoline as word-long immediate operands.
3843 On RISC machines, it is often necessary to load each address into a
3844 register in two parts. Then pieces of each address form separate
3847 The code generated to initialize the trampoline must store the variable
3848 parts---the static chain value and the function address---into the
3849 immediate operands of the instructions. On a CISC machine, this is
3850 simply a matter of copying each address to a memory reference at the
3851 proper offset from the start of the trampoline. On a RISC machine, it
3852 may be necessary to take out pieces of the address and store them
3855 @hook TARGET_ASM_TRAMPOLINE_TEMPLATE
3857 @defmac TRAMPOLINE_SECTION
3858 Return the section into which the trampoline template is to be placed
3859 (@pxref{Sections}). The default value is @code{readonly_data_section}.
3862 @defmac TRAMPOLINE_SIZE
3863 A C expression for the size in bytes of the trampoline, as an integer.
3866 @defmac TRAMPOLINE_ALIGNMENT
3867 Alignment required for trampolines, in bits.
3869 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
3870 is used for aligning trampolines.
3873 @hook TARGET_TRAMPOLINE_INIT
3875 @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
3877 Implementing trampolines is difficult on many machines because they have
3878 separate instruction and data caches. Writing into a stack location
3879 fails to clear the memory in the instruction cache, so when the program
3880 jumps to that location, it executes the old contents.
3882 Here are two possible solutions. One is to clear the relevant parts of
3883 the instruction cache whenever a trampoline is set up. The other is to
3884 make all trampolines identical, by having them jump to a standard
3885 subroutine. The former technique makes trampoline execution faster; the
3886 latter makes initialization faster.
3888 To clear the instruction cache when a trampoline is initialized, define
3889 the following macro.
3891 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
3892 If defined, expands to a C expression clearing the @emph{instruction
3893 cache} in the specified interval. The definition of this macro would
3894 typically be a series of @code{asm} statements. Both @var{beg} and
3895 @var{end} are both pointer expressions.
3898 To use a standard subroutine, define the following macro. In addition,
3899 you must make sure that the instructions in a trampoline fill an entire
3900 cache line with identical instructions, or else ensure that the
3901 beginning of the trampoline code is always aligned at the same point in
3902 its cache line. Look in @file{m68k.h} as a guide.
3904 @defmac TRANSFER_FROM_TRAMPOLINE
3905 Define this macro if trampolines need a special subroutine to do their
3906 work. The macro should expand to a series of @code{asm} statements
3907 which will be compiled with GCC@. They go in a library function named
3908 @code{__transfer_from_trampoline}.
3910 If you need to avoid executing the ordinary prologue code of a compiled
3911 C function when you jump to the subroutine, you can do so by placing a
3912 special label of your own in the assembler code. Use one @code{asm}
3913 statement to generate an assembler label, and another to make the label
3914 global. Then trampolines can use that label to jump directly to your
3915 special assembler code.
3919 @section Implicit Calls to Library Routines
3920 @cindex library subroutine names
3921 @cindex @file{libgcc.a}
3923 @c prevent bad page break with this line
3924 Here is an explanation of implicit calls to library routines.
3926 @defmac DECLARE_LIBRARY_RENAMES
3927 This macro, if defined, should expand to a piece of C code that will get
3928 expanded when compiling functions for libgcc.a. It can be used to
3929 provide alternate names for GCC's internal library functions if there
3930 are ABI-mandated names that the compiler should provide.
3933 @findex set_optab_libfunc
3934 @findex init_one_libfunc
3935 @hook TARGET_INIT_LIBFUNCS
3937 @hook TARGET_LIBFUNC_GNU_PREFIX
3939 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
3940 This macro should return @code{true} if the library routine that
3941 implements the floating point comparison operator @var{comparison} in
3942 mode @var{mode} will return a boolean, and @var{false} if it will
3945 GCC's own floating point libraries return tristates from the
3946 comparison operators, so the default returns false always. Most ports
3947 don't need to define this macro.
3950 @defmac TARGET_LIB_INT_CMP_BIASED
3951 This macro should evaluate to @code{true} if the integer comparison
3952 functions (like @code{__cmpdi2}) return 0 to indicate that the first
3953 operand is smaller than the second, 1 to indicate that they are equal,
3954 and 2 to indicate that the first operand is greater than the second.
3955 If this macro evaluates to @code{false} the comparison functions return
3956 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
3957 in @file{libgcc.a}, you do not need to define this macro.
3960 @defmac TARGET_HAS_NO_HW_DIVIDE
3961 This macro should be defined if the target has no hardware divide
3962 instructions. If this macro is defined, GCC will use an algorithm which
3963 make use of simple logical and arithmetic operations for 64-bit
3964 division. If the macro is not defined, GCC will use an algorithm which
3965 make use of a 64-bit by 32-bit divide primitive.
3968 @cindex @code{EDOM}, implicit usage
3971 The value of @code{EDOM} on the target machine, as a C integer constant
3972 expression. If you don't define this macro, GCC does not attempt to
3973 deposit the value of @code{EDOM} into @code{errno} directly. Look in
3974 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
3977 If you do not define @code{TARGET_EDOM}, then compiled code reports
3978 domain errors by calling the library function and letting it report the
3979 error. If mathematical functions on your system use @code{matherr} when
3980 there is an error, then you should leave @code{TARGET_EDOM} undefined so
3981 that @code{matherr} is used normally.
3984 @cindex @code{errno}, implicit usage
3985 @defmac GEN_ERRNO_RTX
3986 Define this macro as a C expression to create an rtl expression that
3987 refers to the global ``variable'' @code{errno}. (On certain systems,
3988 @code{errno} may not actually be a variable.) If you don't define this
3989 macro, a reasonable default is used.
3992 @hook TARGET_LIBC_HAS_FUNCTION
3994 @hook TARGET_LIBC_HAS_FAST_FUNCTION
3996 @defmac NEXT_OBJC_RUNTIME
3997 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
3998 by default. This calling convention involves passing the object, the selector
3999 and the method arguments all at once to the method-lookup library function.
4000 This is the usual setting when targeting Darwin/Mac OS X systems, which have
4001 the NeXT runtime installed.
4003 If the macro is set to 0, the "GNU" Objective-C message sending convention
4004 will be used by default. This convention passes just the object and the
4005 selector to the method-lookup function, which returns a pointer to the method.
4007 In either case, it remains possible to select code-generation for the alternate
4008 scheme, by means of compiler command line switches.
4011 @node Addressing Modes
4012 @section Addressing Modes
4013 @cindex addressing modes
4015 @c prevent bad page break with this line
4016 This is about addressing modes.
4018 @defmac HAVE_PRE_INCREMENT
4019 @defmacx HAVE_PRE_DECREMENT
4020 @defmacx HAVE_POST_INCREMENT
4021 @defmacx HAVE_POST_DECREMENT
4022 A C expression that is nonzero if the machine supports pre-increment,
4023 pre-decrement, post-increment, or post-decrement addressing respectively.
4026 @defmac HAVE_PRE_MODIFY_DISP
4027 @defmacx HAVE_POST_MODIFY_DISP
4028 A C expression that is nonzero if the machine supports pre- or
4029 post-address side-effect generation involving constants other than
4030 the size of the memory operand.
4033 @defmac HAVE_PRE_MODIFY_REG
4034 @defmacx HAVE_POST_MODIFY_REG
4035 A C expression that is nonzero if the machine supports pre- or
4036 post-address side-effect generation involving a register displacement.
4039 @defmac CONSTANT_ADDRESS_P (@var{x})
4040 A C expression that is 1 if the RTX @var{x} is a constant which
4041 is a valid address. On most machines the default definition of
4042 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
4043 is acceptable, but a few machines are more restrictive as to which
4044 constant addresses are supported.
4047 @defmac CONSTANT_P (@var{x})
4048 @code{CONSTANT_P}, which is defined by target-independent code,
4049 accepts integer-values expressions whose values are not explicitly
4050 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4051 expressions and @code{const} arithmetic expressions, in addition to
4052 @code{const_int} and @code{const_double} expressions.
4055 @defmac MAX_REGS_PER_ADDRESS
4056 A number, the maximum number of registers that can appear in a valid
4057 memory address. Note that it is up to you to specify a value equal to
4058 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
4062 @hook TARGET_LEGITIMATE_ADDRESS_P
4064 @defmac TARGET_MEM_CONSTRAINT
4065 A single character to be used instead of the default @code{'m'}
4066 character for general memory addresses. This defines the constraint
4067 letter which matches the memory addresses accepted by
4068 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
4069 support new address formats in your back end without changing the
4070 semantics of the @code{'m'} constraint. This is necessary in order to
4071 preserve functionality of inline assembly constructs using the
4072 @code{'m'} constraint.
4075 @defmac FIND_BASE_TERM (@var{x})
4076 A C expression to determine the base term of address @var{x},
4077 or to provide a simplified version of @var{x} from which @file{alias.c}
4078 can easily find the base term. This macro is used in only two places:
4079 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
4081 It is always safe for this macro to not be defined. It exists so
4082 that alias analysis can understand machine-dependent addresses.
4084 The typical use of this macro is to handle addresses containing
4085 a label_ref or symbol_ref within an UNSPEC@.
4088 @hook TARGET_LEGITIMIZE_ADDRESS
4090 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4091 A C compound statement that attempts to replace @var{x}, which is an address
4092 that needs reloading, with a valid memory address for an operand of mode
4093 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4094 It is not necessary to define this macro, but it might be useful for
4095 performance reasons.
4097 For example, on the i386, it is sometimes possible to use a single
4098 reload register instead of two by reloading a sum of two pseudo
4099 registers into a register. On the other hand, for number of RISC
4100 processors offsets are limited so that often an intermediate address
4101 needs to be generated in order to address a stack slot. By defining
4102 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4103 generated for adjacent some stack slots can be made identical, and thus
4106 @emph{Note}: This macro should be used with caution. It is necessary
4107 to know something of how reload works in order to effectively use this,
4108 and it is quite easy to produce macros that build in too much knowledge
4109 of reload internals.
4111 @emph{Note}: This macro must be able to reload an address created by a
4112 previous invocation of this macro. If it fails to handle such addresses
4113 then the compiler may generate incorrect code or abort.
4116 The macro definition should use @code{push_reload} to indicate parts that
4117 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4118 suitable to be passed unaltered to @code{push_reload}.
4120 The code generated by this macro must not alter the substructure of
4121 @var{x}. If it transforms @var{x} into a more legitimate form, it
4122 should assign @var{x} (which will always be a C variable) a new value.
4123 This also applies to parts that you change indirectly by calling
4126 @findex strict_memory_address_p
4127 The macro definition may use @code{strict_memory_address_p} to test if
4128 the address has become legitimate.
4131 If you want to change only a part of @var{x}, one standard way of doing
4132 this is to use @code{copy_rtx}. Note, however, that it unshares only a
4133 single level of rtl. Thus, if the part to be changed is not at the
4134 top level, you'll need to replace first the top level.
4135 It is not necessary for this macro to come up with a legitimate
4136 address; but often a machine-dependent strategy can generate better code.
4139 @hook TARGET_MODE_DEPENDENT_ADDRESS_P
4141 @hook TARGET_LEGITIMATE_CONSTANT_P
4143 @hook TARGET_DELEGITIMIZE_ADDRESS
4145 @hook TARGET_CONST_NOT_OK_FOR_DEBUG_P
4147 @hook TARGET_CANNOT_FORCE_CONST_MEM
4149 @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
4151 @hook TARGET_USE_BLOCKS_FOR_DECL_P
4153 @hook TARGET_BUILTIN_RECIPROCAL
4155 @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
4157 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
4159 @hook TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT
4161 @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
4163 @hook TARGET_VECTORIZE_VEC_PERM_CONST
4165 @hook TARGET_VECTORIZE_BUILTIN_CONVERSION
4167 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
4169 @hook TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION
4171 @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
4173 @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
4175 @hook TARGET_VECTORIZE_SPLIT_REDUCTION
4177 @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES
4179 @hook TARGET_VECTORIZE_GET_MASK_MODE
4181 @hook TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE
4183 @hook TARGET_VECTORIZE_INIT_COST
4185 @hook TARGET_VECTORIZE_ADD_STMT_COST
4187 @hook TARGET_VECTORIZE_FINISH_COST
4189 @hook TARGET_VECTORIZE_DESTROY_COST_DATA
4191 @hook TARGET_VECTORIZE_BUILTIN_GATHER
4193 @hook TARGET_VECTORIZE_BUILTIN_SCATTER
4195 @hook TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN
4197 @hook TARGET_SIMD_CLONE_ADJUST
4199 @hook TARGET_SIMD_CLONE_USABLE
4201 @hook TARGET_SIMT_VF
4203 @hook TARGET_GOACC_VALIDATE_DIMS
4205 @hook TARGET_GOACC_DIM_LIMIT
4207 @hook TARGET_GOACC_FORK_JOIN
4209 @hook TARGET_GOACC_REDUCTION
4211 @hook TARGET_PREFERRED_ELSE_VALUE
4213 @node Anchored Addresses
4214 @section Anchored Addresses
4215 @cindex anchored addresses
4216 @cindex @option{-fsection-anchors}
4218 GCC usually addresses every static object as a separate entity.
4219 For example, if we have:
4223 int foo (void) @{ return a + b + c; @}
4226 the code for @code{foo} will usually calculate three separate symbolic
4227 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
4228 it would be better to calculate just one symbolic address and access
4229 the three variables relative to it. The equivalent pseudocode would
4235 register int *xr = &x;
4236 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
4240 (which isn't valid C). We refer to shared addresses like @code{x} as
4241 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
4243 The hooks below describe the target properties that GCC needs to know
4244 in order to make effective use of section anchors. It won't use
4245 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
4246 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
4248 @hook TARGET_MIN_ANCHOR_OFFSET
4250 @hook TARGET_MAX_ANCHOR_OFFSET
4252 @hook TARGET_ASM_OUTPUT_ANCHOR
4254 @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
4256 @node Condition Code
4257 @section Condition Code Status
4258 @cindex condition code status
4260 The macros in this section can be split in two families, according to the
4261 two ways of representing condition codes in GCC.
4263 The first representation is the so called @code{(cc0)} representation
4264 (@pxref{Jump Patterns}), where all instructions can have an implicit
4265 clobber of the condition codes. The second is the condition code
4266 register representation, which provides better schedulability for
4267 architectures that do have a condition code register, but on which
4268 most instructions do not affect it. The latter category includes
4271 The implicit clobbering poses a strong restriction on the placement of
4272 the definition and use of the condition code. In the past the definition
4273 and use were always adjacent. However, recent changes to support trapping
4274 arithmatic may result in the definition and user being in different blocks.
4275 Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally,
4276 the definition may be the source of exception handling edges.
4278 These restrictions can prevent important
4279 optimizations on some machines. For example, on the IBM RS/6000, there
4280 is a delay for taken branches unless the condition code register is set
4281 three instructions earlier than the conditional branch. The instruction
4282 scheduler cannot perform this optimization if it is not permitted to
4283 separate the definition and use of the condition code register.
4285 For this reason, it is possible and suggested to use a register to
4286 represent the condition code for new ports. If there is a specific
4287 condition code register in the machine, use a hard register. If the
4288 condition code or comparison result can be placed in any general register,
4289 or if there are multiple condition registers, use a pseudo register.
4290 Registers used to store the condition code value will usually have a mode
4291 that is in class @code{MODE_CC}.
4293 Alternatively, you can use @code{BImode} if the comparison operator is
4294 specified already in the compare instruction. In this case, you are not
4295 interested in most macros in this section.
4298 * CC0 Condition Codes:: Old style representation of condition codes.
4299 * MODE_CC Condition Codes:: Modern representation of condition codes.
4302 @node CC0 Condition Codes
4303 @subsection Representation of condition codes using @code{(cc0)}
4307 The file @file{conditions.h} defines a variable @code{cc_status} to
4308 describe how the condition code was computed (in case the interpretation of
4309 the condition code depends on the instruction that it was set by). This
4310 variable contains the RTL expressions on which the condition code is
4311 currently based, and several standard flags.
4313 Sometimes additional machine-specific flags must be defined in the machine
4314 description header file. It can also add additional machine-specific
4315 information by defining @code{CC_STATUS_MDEP}.
4317 @defmac CC_STATUS_MDEP
4318 C code for a data type which is used for declaring the @code{mdep}
4319 component of @code{cc_status}. It defaults to @code{int}.
4321 This macro is not used on machines that do not use @code{cc0}.
4324 @defmac CC_STATUS_MDEP_INIT
4325 A C expression to initialize the @code{mdep} field to ``empty''.
4326 The default definition does nothing, since most machines don't use
4327 the field anyway. If you want to use the field, you should probably
4328 define this macro to initialize it.
4330 This macro is not used on machines that do not use @code{cc0}.
4333 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4334 A C compound statement to set the components of @code{cc_status}
4335 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4336 this macro's responsibility to recognize insns that set the condition
4337 code as a byproduct of other activity as well as those that explicitly
4340 This macro is not used on machines that do not use @code{cc0}.
4342 If there are insns that do not set the condition code but do alter
4343 other machine registers, this macro must check to see whether they
4344 invalidate the expressions that the condition code is recorded as
4345 reflecting. For example, on the 68000, insns that store in address
4346 registers do not set the condition code, which means that usually
4347 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4348 insns. But suppose that the previous insn set the condition code
4349 based on location @samp{a4@@(102)} and the current insn stores a new
4350 value in @samp{a4}. Although the condition code is not changed by
4351 this, it will no longer be true that it reflects the contents of
4352 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4353 @code{cc_status} in this case to say that nothing is known about the
4354 condition code value.
4356 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4357 with the results of peephole optimization: insns whose patterns are
4358 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4359 constants which are just the operands. The RTL structure of these
4360 insns is not sufficient to indicate what the insns actually do. What
4361 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4362 @code{CC_STATUS_INIT}.
4364 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4365 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4366 @samp{cc}. This avoids having detailed information about patterns in
4367 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4370 @node MODE_CC Condition Codes
4371 @subsection Representation of condition codes using registers
4375 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4376 On many machines, the condition code may be produced by other instructions
4377 than compares, for example the branch can use directly the condition
4378 code set by a subtract instruction. However, on some machines
4379 when the condition code is set this way some bits (such as the overflow
4380 bit) are not set in the same way as a test instruction, so that a different
4381 branch instruction must be used for some conditional branches. When
4382 this happens, use the machine mode of the condition code register to
4383 record different formats of the condition code register. Modes can
4384 also be used to record which compare instruction (e.g.@: a signed or an
4385 unsigned comparison) produced the condition codes.
4387 If other modes than @code{CCmode} are required, add them to
4388 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
4389 a mode given an operand of a compare. This is needed because the modes
4390 have to be chosen not only during RTL generation but also, for example,
4391 by instruction combination. The result of @code{SELECT_CC_MODE} should
4392 be consistent with the mode used in the patterns; for example to support
4393 the case of the add on the SPARC discussed above, we have the pattern
4399 (plus:SI (match_operand:SI 0 "register_operand" "%r")
4400 (match_operand:SI 1 "arith_operand" "rI"))
4407 together with a @code{SELECT_CC_MODE} that returns @code{CCNZmode}
4408 for comparisons whose argument is a @code{plus}:
4411 #define SELECT_CC_MODE(OP,X,Y) \
4412 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4413 ? ((OP == LT || OP == LE || OP == GT || OP == GE) \
4414 ? CCFPEmode : CCFPmode) \
4415 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4416 || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \
4417 ? CCNZmode : CCmode))
4420 Another reason to use modes is to retain information on which operands
4421 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
4424 You should define this macro if and only if you define extra CC modes
4425 in @file{@var{machine}-modes.def}.
4428 @hook TARGET_CANONICALIZE_COMPARISON
4430 @defmac REVERSIBLE_CC_MODE (@var{mode})
4431 A C expression whose value is one if it is always safe to reverse a
4432 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4433 can ever return @var{mode} for a floating-point inequality comparison,
4434 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4436 You need not define this macro if it would always returns zero or if the
4437 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4438 For example, here is the definition used on the SPARC, where floating-point
4439 inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}:
4442 #define REVERSIBLE_CC_MODE(MODE) \
4443 ((MODE) != CCFPEmode && (MODE) != CCFPmode)
4447 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
4448 A C expression whose value is reversed condition code of the @var{code} for
4449 comparison done in CC_MODE @var{mode}. The macro is used only in case
4450 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
4451 machine has some non-standard way how to reverse certain conditionals. For
4452 instance in case all floating point conditions are non-trapping, compiler may
4453 freely convert unordered compares to ordered ones. Then definition may look
4457 #define REVERSE_CONDITION(CODE, MODE) \
4458 ((MODE) != CCFPmode ? reverse_condition (CODE) \
4459 : reverse_condition_maybe_unordered (CODE))
4463 @hook TARGET_FIXED_CONDITION_CODE_REGS
4465 @hook TARGET_CC_MODES_COMPATIBLE
4467 @hook TARGET_FLAGS_REGNUM
4470 @section Describing Relative Costs of Operations
4471 @cindex costs of instructions
4472 @cindex relative costs
4473 @cindex speed of instructions
4475 These macros let you describe the relative speed of various operations
4476 on the target machine.
4478 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
4479 A C expression for the cost of moving data of mode @var{mode} from a
4480 register in class @var{from} to one in class @var{to}. The classes are
4481 expressed using the enumeration values such as @code{GENERAL_REGS}. A
4482 value of 2 is the default; other values are interpreted relative to
4485 It is not required that the cost always equal 2 when @var{from} is the
4486 same as @var{to}; on some machines it is expensive to move between
4487 registers if they are not general registers.
4489 If reload sees an insn consisting of a single @code{set} between two
4490 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4491 classes returns a value of 2, reload does not check to ensure that the
4492 constraints of the insn are met. Setting a cost of other than 2 will
4493 allow reload to verify that the constraints are met. You should do this
4494 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4496 These macros are obsolete, new ports should use the target hook
4497 @code{TARGET_REGISTER_MOVE_COST} instead.
4500 @hook TARGET_REGISTER_MOVE_COST
4502 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4503 A C expression for the cost of moving data of mode @var{mode} between a
4504 register of class @var{class} and memory; @var{in} is zero if the value
4505 is to be written to memory, nonzero if it is to be read in. This cost
4506 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4507 registers and memory is more expensive than between two registers, you
4508 should define this macro to express the relative cost.
4510 If you do not define this macro, GCC uses a default cost of 4 plus
4511 the cost of copying via a secondary reload register, if one is
4512 needed. If your machine requires a secondary reload register to copy
4513 between memory and a register of @var{class} but the reload mechanism is
4514 more complex than copying via an intermediate, define this macro to
4515 reflect the actual cost of the move.
4517 GCC defines the function @code{memory_move_secondary_cost} if
4518 secondary reloads are needed. It computes the costs due to copying via
4519 a secondary register. If your machine copies from memory using a
4520 secondary register in the conventional way but the default base value of
4521 4 is not correct for your machine, define this macro to add some other
4522 value to the result of that function. The arguments to that function
4523 are the same as to this macro.
4525 These macros are obsolete, new ports should use the target hook
4526 @code{TARGET_MEMORY_MOVE_COST} instead.
4529 @hook TARGET_MEMORY_MOVE_COST
4531 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
4532 A C expression for the cost of a branch instruction. A value of 1 is
4533 the default; other values are interpreted relative to that. Parameter
4534 @var{speed_p} is true when the branch in question should be optimized
4535 for speed. When it is false, @code{BRANCH_COST} should return a value
4536 optimal for code size rather than performance. @var{predictable_p} is
4537 true for well-predicted branches. On many architectures the
4538 @code{BRANCH_COST} can be reduced then.
4541 Here are additional macros which do not specify precise relative costs,
4542 but only that certain actions are more expensive than GCC would
4545 @defmac SLOW_BYTE_ACCESS
4546 Define this macro as a C expression which is nonzero if accessing less
4547 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
4548 faster than accessing a word of memory, i.e., if such access
4549 require more than one instruction or if there is no difference in cost
4550 between byte and (aligned) word loads.
4552 When this macro is not defined, the compiler will access a field by
4553 finding the smallest containing object; when it is defined, a fullword
4554 load will be used if alignment permits. Unless bytes accesses are
4555 faster than word accesses, using word accesses is preferable since it
4556 may eliminate subsequent memory access if subsequent accesses occur to
4557 other fields in the same word of the structure, but to different bytes.
4560 @hook TARGET_SLOW_UNALIGNED_ACCESS
4562 @defmac MOVE_RATIO (@var{speed})
4563 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4564 which a sequence of insns should be generated instead of a
4565 string move insn or a library call. Increasing the value will always
4566 make code faster, but eventually incurs high cost in increased code size.
4568 Note that on machines where the corresponding move insn is a
4569 @code{define_expand} that emits a sequence of insns, this macro counts
4570 the number of such sequences.
4572 The parameter @var{speed} is true if the code is currently being
4573 optimized for speed rather than size.
4575 If you don't define this, a reasonable default is used.
4578 @hook TARGET_USE_BY_PIECES_INFRASTRUCTURE_P
4580 @hook TARGET_COMPARE_BY_PIECES_BRANCH_RATIO
4582 @defmac MOVE_MAX_PIECES
4583 A C expression used by @code{move_by_pieces} to determine the largest unit
4584 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4587 @defmac STORE_MAX_PIECES
4588 A C expression used by @code{store_by_pieces} to determine the largest unit
4589 a store used to memory is. Defaults to @code{MOVE_MAX_PIECES}, or two times
4590 the size of @code{HOST_WIDE_INT}, whichever is smaller.
4593 @defmac COMPARE_MAX_PIECES
4594 A C expression used by @code{compare_by_pieces} to determine the largest unit
4595 a load or store used to compare memory is. Defaults to
4596 @code{MOVE_MAX_PIECES}.
4599 @defmac CLEAR_RATIO (@var{speed})
4600 The threshold of number of scalar move insns, @emph{below} which a sequence
4601 of insns should be generated to clear memory instead of a string clear insn
4602 or a library call. Increasing the value will always make code faster, but
4603 eventually incurs high cost in increased code size.
4605 The parameter @var{speed} is true if the code is currently being
4606 optimized for speed rather than size.
4608 If you don't define this, a reasonable default is used.
4611 @defmac SET_RATIO (@var{speed})
4612 The threshold of number of scalar move insns, @emph{below} which a sequence
4613 of insns should be generated to set memory to a constant value, instead of
4614 a block set insn or a library call.
4615 Increasing the value will always make code faster, but
4616 eventually incurs high cost in increased code size.
4618 The parameter @var{speed} is true if the code is currently being
4619 optimized for speed rather than size.
4621 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
4624 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
4625 A C expression used to determine whether a load postincrement is a good
4626 thing to use for a given mode. Defaults to the value of
4627 @code{HAVE_POST_INCREMENT}.
4630 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
4631 A C expression used to determine whether a load postdecrement is a good
4632 thing to use for a given mode. Defaults to the value of
4633 @code{HAVE_POST_DECREMENT}.
4636 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
4637 A C expression used to determine whether a load preincrement is a good
4638 thing to use for a given mode. Defaults to the value of
4639 @code{HAVE_PRE_INCREMENT}.
4642 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
4643 A C expression used to determine whether a load predecrement is a good
4644 thing to use for a given mode. Defaults to the value of
4645 @code{HAVE_PRE_DECREMENT}.
4648 @defmac USE_STORE_POST_INCREMENT (@var{mode})
4649 A C expression used to determine whether a store postincrement is a good
4650 thing to use for a given mode. Defaults to the value of
4651 @code{HAVE_POST_INCREMENT}.
4654 @defmac USE_STORE_POST_DECREMENT (@var{mode})
4655 A C expression used to determine whether a store postdecrement is a good
4656 thing to use for a given mode. Defaults to the value of
4657 @code{HAVE_POST_DECREMENT}.
4660 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
4661 This macro is used to determine whether a store preincrement is a good
4662 thing to use for a given mode. Defaults to the value of
4663 @code{HAVE_PRE_INCREMENT}.
4666 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
4667 This macro is used to determine whether a store predecrement is a good
4668 thing to use for a given mode. Defaults to the value of
4669 @code{HAVE_PRE_DECREMENT}.
4672 @defmac NO_FUNCTION_CSE
4673 Define this macro to be true if it is as good or better to call a constant
4674 function address than to call an address kept in a register.
4677 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
4678 Define this macro if a non-short-circuit operation produced by
4679 @samp{fold_range_test ()} is optimal. This macro defaults to true if
4680 @code{BRANCH_COST} is greater than or equal to the value 2.
4683 @hook TARGET_OPTAB_SUPPORTED_P
4685 @hook TARGET_RTX_COSTS
4687 @hook TARGET_ADDRESS_COST
4689 @hook TARGET_INSN_COST
4691 @hook TARGET_MAX_NOCE_IFCVT_SEQ_COST
4693 @hook TARGET_NOCE_CONVERSION_PROFITABLE_P
4695 @hook TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P
4697 @hook TARGET_ESTIMATED_POLY_VALUE
4700 @section Adjusting the Instruction Scheduler
4702 The instruction scheduler may need a fair amount of machine-specific
4703 adjustment in order to produce good code. GCC provides several target
4704 hooks for this purpose. It is usually enough to define just a few of
4705 them: try the first ones in this list first.
4707 @hook TARGET_SCHED_ISSUE_RATE
4709 @hook TARGET_SCHED_VARIABLE_ISSUE
4711 @hook TARGET_SCHED_ADJUST_COST
4713 @hook TARGET_SCHED_ADJUST_PRIORITY
4715 @hook TARGET_SCHED_REORDER
4717 @hook TARGET_SCHED_REORDER2
4719 @hook TARGET_SCHED_MACRO_FUSION_P
4721 @hook TARGET_SCHED_MACRO_FUSION_PAIR_P
4723 @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
4725 @hook TARGET_SCHED_INIT
4727 @hook TARGET_SCHED_FINISH
4729 @hook TARGET_SCHED_INIT_GLOBAL
4731 @hook TARGET_SCHED_FINISH_GLOBAL
4733 @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
4735 @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
4737 @hook TARGET_SCHED_DFA_POST_CYCLE_INSN
4739 @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
4741 @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
4743 @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
4745 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
4747 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
4749 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
4751 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
4753 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
4755 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
4757 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
4759 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
4761 @hook TARGET_SCHED_DFA_NEW_CYCLE
4763 @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
4765 @hook TARGET_SCHED_H_I_D_EXTENDED
4767 @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
4769 @hook TARGET_SCHED_INIT_SCHED_CONTEXT
4771 @hook TARGET_SCHED_SET_SCHED_CONTEXT
4773 @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
4775 @hook TARGET_SCHED_FREE_SCHED_CONTEXT
4777 @hook TARGET_SCHED_SPECULATE_INSN
4779 @hook TARGET_SCHED_NEEDS_BLOCK_P
4781 @hook TARGET_SCHED_GEN_SPEC_CHECK
4783 @hook TARGET_SCHED_SET_SCHED_FLAGS
4785 @hook TARGET_SCHED_CAN_SPECULATE_INSN
4787 @hook TARGET_SCHED_SMS_RES_MII
4789 @hook TARGET_SCHED_DISPATCH
4791 @hook TARGET_SCHED_DISPATCH_DO
4793 @hook TARGET_SCHED_EXPOSED_PIPELINE
4795 @hook TARGET_SCHED_REASSOCIATION_WIDTH
4797 @hook TARGET_SCHED_FUSION_PRIORITY
4799 @hook TARGET_EXPAND_DIVMOD_LIBFUNC
4802 @section Dividing the Output into Sections (Texts, Data, @dots{})
4803 @c the above section title is WAY too long. maybe cut the part between
4804 @c the (...)? --mew 10feb93
4806 An object file is divided into sections containing different types of
4807 data. In the most common case, there are three sections: the @dfn{text
4808 section}, which holds instructions and read-only data; the @dfn{data
4809 section}, which holds initialized writable data; and the @dfn{bss
4810 section}, which holds uninitialized data. Some systems have other kinds
4813 @file{varasm.c} provides several well-known sections, such as
4814 @code{text_section}, @code{data_section} and @code{bss_section}.
4815 The normal way of controlling a @code{@var{foo}_section} variable
4816 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
4817 as described below. The macros are only read once, when @file{varasm.c}
4818 initializes itself, so their values must be run-time constants.
4819 They may however depend on command-line flags.
4821 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
4822 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
4823 to be string literals.
4825 Some assemblers require a different string to be written every time a
4826 section is selected. If your assembler falls into this category, you
4827 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
4828 @code{get_unnamed_section} to set up the sections.
4830 You must always create a @code{text_section}, either by defining
4831 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
4832 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
4833 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
4834 create a distinct @code{readonly_data_section}, the default is to
4835 reuse @code{text_section}.
4837 All the other @file{varasm.c} sections are optional, and are null
4838 if the target does not provide them.
4840 @defmac TEXT_SECTION_ASM_OP
4841 A C expression whose value is a string, including spacing, containing the
4842 assembler operation that should precede instructions and read-only data.
4843 Normally @code{"\t.text"} is right.
4846 @defmac HOT_TEXT_SECTION_NAME
4847 If defined, a C string constant for the name of the section containing most
4848 frequently executed functions of the program. If not defined, GCC will provide
4849 a default definition if the target supports named sections.
4852 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
4853 If defined, a C string constant for the name of the section containing unlikely
4854 executed functions in the program.
4857 @defmac DATA_SECTION_ASM_OP
4858 A C expression whose value is a string, including spacing, containing the
4859 assembler operation to identify the following data as writable initialized
4860 data. Normally @code{"\t.data"} is right.
4863 @defmac SDATA_SECTION_ASM_OP
4864 If defined, a C expression whose value is a string, including spacing,
4865 containing the assembler operation to identify the following data as
4866 initialized, writable small data.
4869 @defmac READONLY_DATA_SECTION_ASM_OP
4870 A C expression whose value is a string, including spacing, containing the
4871 assembler operation to identify the following data as read-only initialized
4875 @defmac BSS_SECTION_ASM_OP
4876 If defined, a C expression whose value is a string, including spacing,
4877 containing the assembler operation to identify the following data as
4878 uninitialized global data. If not defined, and
4879 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
4880 uninitialized global data will be output in the data section if
4881 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
4885 @defmac SBSS_SECTION_ASM_OP
4886 If defined, a C expression whose value is a string, including spacing,
4887 containing the assembler operation to identify the following data as
4888 uninitialized, writable small data.
4891 @defmac TLS_COMMON_ASM_OP
4892 If defined, a C expression whose value is a string containing the
4893 assembler operation to identify the following data as thread-local
4894 common data. The default is @code{".tls_common"}.
4897 @defmac TLS_SECTION_ASM_FLAG
4898 If defined, a C expression whose value is a character constant
4899 containing the flag used to mark a section as a TLS section. The
4900 default is @code{'T'}.
4903 @defmac INIT_SECTION_ASM_OP
4904 If defined, a C expression whose value is a string, including spacing,
4905 containing the assembler operation to identify the following data as
4906 initialization code. If not defined, GCC will assume such a section does
4907 not exist. This section has no corresponding @code{init_section}
4908 variable; it is used entirely in runtime code.
4911 @defmac FINI_SECTION_ASM_OP
4912 If defined, a C expression whose value is a string, including spacing,
4913 containing the assembler operation to identify the following data as
4914 finalization code. If not defined, GCC will assume such a section does
4915 not exist. This section has no corresponding @code{fini_section}
4916 variable; it is used entirely in runtime code.
4919 @defmac INIT_ARRAY_SECTION_ASM_OP
4920 If defined, a C expression whose value is a string, including spacing,
4921 containing the assembler operation to identify the following data as
4922 part of the @code{.init_array} (or equivalent) section. If not
4923 defined, GCC will assume such a section does not exist. Do not define
4924 both this macro and @code{INIT_SECTION_ASM_OP}.
4927 @defmac FINI_ARRAY_SECTION_ASM_OP
4928 If defined, a C expression whose value is a string, including spacing,
4929 containing the assembler operation to identify the following data as
4930 part of the @code{.fini_array} (or equivalent) section. If not
4931 defined, GCC will assume such a section does not exist. Do not define
4932 both this macro and @code{FINI_SECTION_ASM_OP}.
4935 @defmac MACH_DEP_SECTION_ASM_FLAG
4936 If defined, a C expression whose value is a character constant
4937 containing the flag used to mark a machine-dependent section. This
4938 corresponds to the @code{SECTION_MACH_DEP} section flag.
4941 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
4942 If defined, an ASM statement that switches to a different section
4943 via @var{section_op}, calls @var{function}, and switches back to
4944 the text section. This is used in @file{crtstuff.c} if
4945 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
4946 to initialization and finalization functions from the init and fini
4947 sections. By default, this macro uses a simple function call. Some
4948 ports need hand-crafted assembly code to avoid dependencies on
4949 registers initialized in the function prologue or to ensure that
4950 constant pools don't end up too far way in the text section.
4953 @defmac TARGET_LIBGCC_SDATA_SECTION
4954 If defined, a string which names the section into which small
4955 variables defined in crtstuff and libgcc should go. This is useful
4956 when the target has options for optimizing access to small data, and
4957 you want the crtstuff and libgcc routines to be conservative in what
4958 they expect of your application yet liberal in what your application
4959 expects. For example, for targets with a @code{.sdata} section (like
4960 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
4961 require small data support from your application, but use this macro
4962 to put small data into @code{.sdata} so that your application can
4963 access these variables whether it uses small data or not.
4966 @defmac FORCE_CODE_SECTION_ALIGN
4967 If defined, an ASM statement that aligns a code section to some
4968 arbitrary boundary. This is used to force all fragments of the
4969 @code{.init} and @code{.fini} sections to have to same alignment
4970 and thus prevent the linker from having to add any padding.
4973 @defmac JUMP_TABLES_IN_TEXT_SECTION
4974 Define this macro to be an expression with a nonzero value if jump
4975 tables (for @code{tablejump} insns) should be output in the text
4976 section, along with the assembler instructions. Otherwise, the
4977 readonly data section is used.
4979 This macro is irrelevant if there is no separate readonly data section.
4982 @hook TARGET_ASM_INIT_SECTIONS
4984 @hook TARGET_ASM_RELOC_RW_MASK
4986 @hook TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC
4988 @hook TARGET_ASM_SELECT_SECTION
4990 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
4991 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
4992 for @code{FUNCTION_DECL}s as well as for variables and constants.
4994 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
4995 function has been determined to be likely to be called, and nonzero if
4996 it is unlikely to be called.
4999 @hook TARGET_ASM_UNIQUE_SECTION
5001 @hook TARGET_ASM_FUNCTION_RODATA_SECTION
5003 @hook TARGET_ASM_MERGEABLE_RODATA_PREFIX
5005 @hook TARGET_ASM_TM_CLONE_TABLE_SECTION
5007 @hook TARGET_ASM_SELECT_RTX_SECTION
5009 @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
5011 @hook TARGET_ENCODE_SECTION_INFO
5013 @hook TARGET_STRIP_NAME_ENCODING
5015 @hook TARGET_IN_SMALL_DATA_P
5017 @hook TARGET_HAVE_SRODATA_SECTION
5019 @hook TARGET_PROFILE_BEFORE_PROLOGUE
5021 @hook TARGET_BINDS_LOCAL_P
5023 @hook TARGET_HAVE_TLS
5027 @section Position Independent Code
5028 @cindex position independent code
5031 This section describes macros that help implement generation of position
5032 independent code. Simply defining these macros is not enough to
5033 generate valid PIC; you must also add support to the hook
5034 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
5035 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
5036 must modify the definition of @samp{movsi} to do something appropriate
5037 when the source operand contains a symbolic address. You may also
5038 need to alter the handling of switch statements so that they use
5040 @c i rearranged the order of the macros above to try to force one of
5041 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5043 @defmac PIC_OFFSET_TABLE_REGNUM
5044 The register number of the register used to address a table of static
5045 data addresses in memory. In some cases this register is defined by a
5046 processor's ``application binary interface'' (ABI)@. When this macro
5047 is defined, RTL is generated for this register once, as with the stack
5048 pointer and frame pointer registers. If this macro is not defined, it
5049 is up to the machine-dependent files to allocate such a register (if
5050 necessary). Note that this register must be fixed when in use (e.g.@:
5051 when @code{flag_pic} is true).
5054 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5055 A C expression that is nonzero if the register defined by
5056 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
5057 the default is zero. Do not define
5058 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5061 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
5062 A C expression that is nonzero if @var{x} is a legitimate immediate
5063 operand on the target machine when generating position independent code.
5064 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5065 check this. You can also assume @var{flag_pic} is true, so you need not
5066 check it either. You need not define this macro if all constants
5067 (including @code{SYMBOL_REF}) can be immediate operands when generating
5068 position independent code.
5071 @node Assembler Format
5072 @section Defining the Output Assembler Language
5074 This section describes macros whose principal purpose is to describe how
5075 to write instructions in assembler language---rather than what the
5079 * File Framework:: Structural information for the assembler file.
5080 * Data Output:: Output of constants (numbers, strings, addresses).
5081 * Uninitialized Data:: Output of uninitialized variables.
5082 * Label Output:: Output and generation of labels.
5083 * Initialization:: General principles of initialization
5084 and termination routines.
5085 * Macros for Initialization::
5086 Specific macros that control the handling of
5087 initialization and termination routines.
5088 * Instruction Output:: Output of actual instructions.
5089 * Dispatch Tables:: Output of jump tables.
5090 * Exception Region Output:: Output of exception region code.
5091 * Alignment Output:: Pseudo ops for alignment and skipping data.
5094 @node File Framework
5095 @subsection The Overall Framework of an Assembler File
5096 @cindex assembler format
5097 @cindex output of assembler code
5099 @c prevent bad page break with this line
5100 This describes the overall framework of an assembly file.
5102 @findex default_file_start
5103 @hook TARGET_ASM_FILE_START
5105 @hook TARGET_ASM_FILE_START_APP_OFF
5107 @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
5109 @hook TARGET_ASM_FILE_END
5111 @deftypefun void file_end_indicate_exec_stack ()
5112 Some systems use a common convention, the @samp{.note.GNU-stack}
5113 special section, to indicate whether or not an object file relies on
5114 the stack being executable. If your system uses this convention, you
5115 should define @code{TARGET_ASM_FILE_END} to this function. If you
5116 need to do other things in that hook, have your hook function call
5120 @hook TARGET_ASM_LTO_START
5122 @hook TARGET_ASM_LTO_END
5124 @hook TARGET_ASM_CODE_END
5126 @defmac ASM_COMMENT_START
5127 A C string constant describing how to begin a comment in the target
5128 assembler language. The compiler assumes that the comment will end at
5129 the end of the line.
5133 A C string constant for text to be output before each @code{asm}
5134 statement or group of consecutive ones. Normally this is
5135 @code{"#APP"}, which is a comment that has no effect on most
5136 assemblers but tells the GNU assembler that it must check the lines
5137 that follow for all valid assembler constructs.
5141 A C string constant for text to be output after each @code{asm}
5142 statement or group of consecutive ones. Normally this is
5143 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5144 time-saving assumptions that are valid for ordinary compiler output.
5147 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5148 A C statement to output COFF information or DWARF debugging information
5149 which indicates that filename @var{name} is the current source file to
5150 the stdio stream @var{stream}.
5152 This macro need not be defined if the standard form of output
5153 for the file format in use is appropriate.
5156 @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
5158 @hook TARGET_ASM_OUTPUT_IDENT
5160 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5161 A C statement to output the string @var{string} to the stdio stream
5162 @var{stream}. If you do not call the function @code{output_quoted_string}
5163 in your config files, GCC will only call it to output filenames to
5164 the assembler source. So you can use it to canonicalize the format
5165 of the filename using this macro.
5168 @hook TARGET_ASM_NAMED_SECTION
5170 @hook TARGET_ASM_ELF_FLAGS_NUMERIC
5172 @hook TARGET_ASM_FUNCTION_SECTION
5174 @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
5176 @hook TARGET_HAVE_NAMED_SECTIONS
5177 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
5178 It must not be modified by command-line option processing.
5181 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
5182 @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
5184 @hook TARGET_SECTION_TYPE_FLAGS
5186 @hook TARGET_ASM_RECORD_GCC_SWITCHES
5188 @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
5192 @subsection Output of Data
5195 @hook TARGET_ASM_BYTE_OP
5197 @hook TARGET_ASM_INTEGER
5199 @hook TARGET_ASM_DECL_END
5201 @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
5203 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5204 A C statement to output to the stdio stream @var{stream} an assembler
5205 instruction to assemble a string constant containing the @var{len}
5206 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5207 @code{char *} and @var{len} a C expression of type @code{int}.
5209 If the assembler has a @code{.ascii} pseudo-op as found in the
5210 Berkeley Unix assembler, do not define the macro
5211 @code{ASM_OUTPUT_ASCII}.
5214 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
5215 A C statement to output word @var{n} of a function descriptor for
5216 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
5217 is defined, and is otherwise unused.
5220 @defmac CONSTANT_POOL_BEFORE_FUNCTION
5221 You may define this macro as a C expression. You should define the
5222 expression to have a nonzero value if GCC should output the constant
5223 pool for a function before the code for the function, or a zero value if
5224 GCC should output the constant pool after the function. If you do
5225 not define this macro, the usual case, GCC will output the constant
5226 pool before the function.
5229 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5230 A C statement to output assembler commands to define the start of the
5231 constant pool for a function. @var{funname} is a string giving
5232 the name of the function. Should the return type of the function
5233 be required, it can be obtained via @var{fundecl}. @var{size}
5234 is the size, in bytes, of the constant pool that will be written
5235 immediately after this call.
5237 If no constant-pool prefix is required, the usual case, this macro need
5241 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5242 A C statement (with or without semicolon) to output a constant in the
5243 constant pool, if it needs special treatment. (This macro need not do
5244 anything for RTL expressions that can be output normally.)
5246 The argument @var{file} is the standard I/O stream to output the
5247 assembler code on. @var{x} is the RTL expression for the constant to
5248 output, and @var{mode} is the machine mode (in case @var{x} is a
5249 @samp{const_int}). @var{align} is the required alignment for the value
5250 @var{x}; you should output an assembler directive to force this much
5253 The argument @var{labelno} is a number to use in an internal label for
5254 the address of this pool entry. The definition of this macro is
5255 responsible for outputting the label definition at the proper place.
5256 Here is how to do this:
5259 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
5262 When you output a pool entry specially, you should end with a
5263 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5264 entry from being output a second time in the usual manner.
5266 You need not define this macro if it would do nothing.
5269 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5270 A C statement to output assembler commands to at the end of the constant
5271 pool for a function. @var{funname} is a string giving the name of the
5272 function. Should the return type of the function be required, you can
5273 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5274 constant pool that GCC wrote immediately before this call.
5276 If no constant-pool epilogue is required, the usual case, you need not
5280 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
5281 Define this macro as a C expression which is nonzero if @var{C} is
5282 used as a logical line separator by the assembler. @var{STR} points
5283 to the position in the string where @var{C} was found; this can be used if
5284 a line separator uses multiple characters.
5286 If you do not define this macro, the default is that only
5287 the character @samp{;} is treated as a logical line separator.
5290 @hook TARGET_ASM_OPEN_PAREN
5292 These macros are provided by @file{real.h} for writing the definitions
5293 of @code{ASM_OUTPUT_DOUBLE} and the like:
5295 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5296 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5297 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5298 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
5299 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
5300 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
5301 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
5302 target's floating point representation, and store its bit pattern in
5303 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
5304 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
5305 simple @code{long int}. For the others, it should be an array of
5306 @code{long int}. The number of elements in this array is determined
5307 by the size of the desired target floating point data type: 32 bits of
5308 it go in each @code{long int} array element. Each array element holds
5309 32 bits of the result, even if @code{long int} is wider than 32 bits
5310 on the host machine.
5312 The array element values are designed so that you can print them out
5313 using @code{fprintf} in the order they should appear in the target
5317 @node Uninitialized Data
5318 @subsection Output of Uninitialized Variables
5320 Each of the macros in this section is used to do the whole job of
5321 outputting a single uninitialized variable.
5323 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5324 A C statement (sans semicolon) to output to the stdio stream
5325 @var{stream} the assembler definition of a common-label named
5326 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5327 is the size rounded up to whatever alignment the caller wants. It is
5328 possible that @var{size} may be zero, for instance if a struct with no
5329 other member than a zero-length array is defined. In this case, the
5330 backend must output a symbol definition that allocates at least one
5331 byte, both so that the address of the resulting object does not compare
5332 equal to any other, and because some object formats cannot even express
5333 the concept of a zero-sized common symbol, as that is how they represent
5334 an ordinary undefined external.
5336 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5337 output the name itself; before and after that, output the additional
5338 assembler syntax for defining the name, and a newline.
5340 This macro controls how the assembler definitions of uninitialized
5341 common global variables are output.
5344 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5345 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5346 separate, explicit argument. If you define this macro, it is used in
5347 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5348 handling the required alignment of the variable. The alignment is specified
5349 as the number of bits.
5352 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5353 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5354 variable to be output, if there is one, or @code{NULL_TREE} if there
5355 is no corresponding variable. If you define this macro, GCC will use it
5356 in place of both @code{ASM_OUTPUT_COMMON} and
5357 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5358 the variable's decl in order to chose what to output.
5361 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5362 A C statement (sans semicolon) to output to the stdio stream
5363 @var{stream} the assembler definition of uninitialized global @var{decl} named
5364 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
5365 is the alignment specified as the number of bits.
5367 Try to use function @code{asm_output_aligned_bss} defined in file
5368 @file{varasm.c} when defining this macro. If unable, use the expression
5369 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5370 before and after that, output the additional assembler syntax for defining
5371 the name, and a newline.
5373 There are two ways of handling global BSS@. One is to define this macro.
5374 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
5375 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
5376 You do not need to do both.
5378 Some languages do not have @code{common} data, and require a
5379 non-common form of global BSS in order to handle uninitialized globals
5380 efficiently. C++ is one example of this. However, if the target does
5381 not support global BSS, the front end may choose to make globals
5382 common in order to save space in the object file.
5385 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5386 A C statement (sans semicolon) to output to the stdio stream
5387 @var{stream} the assembler definition of a local-common-label named
5388 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5389 is the size rounded up to whatever alignment the caller wants.
5391 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5392 output the name itself; before and after that, output the additional
5393 assembler syntax for defining the name, and a newline.
5395 This macro controls how the assembler definitions of uninitialized
5396 static variables are output.
5399 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5400 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5401 separate, explicit argument. If you define this macro, it is used in
5402 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5403 handling the required alignment of the variable. The alignment is specified
5404 as the number of bits.
5407 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5408 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5409 variable to be output, if there is one, or @code{NULL_TREE} if there
5410 is no corresponding variable. If you define this macro, GCC will use it
5411 in place of both @code{ASM_OUTPUT_DECL} and
5412 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5413 the variable's decl in order to chose what to output.
5417 @subsection Output and Generation of Labels
5419 @c prevent bad page break with this line
5420 This is about outputting labels.
5422 @findex assemble_name
5423 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5424 A C statement (sans semicolon) to output to the stdio stream
5425 @var{stream} the assembler definition of a label named @var{name}.
5426 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5427 output the name itself; before and after that, output the additional
5428 assembler syntax for defining the name, and a newline. A default
5429 definition of this macro is provided which is correct for most systems.
5432 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
5433 A C statement (sans semicolon) to output to the stdio stream
5434 @var{stream} the assembler definition of a label named @var{name} of
5436 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5437 output the name itself; before and after that, output the additional
5438 assembler syntax for defining the name, and a newline. A default
5439 definition of this macro is provided which is correct for most systems.
5441 If this macro is not defined, then the function name is defined in the
5442 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5445 @findex assemble_name_raw
5446 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
5447 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
5448 to refer to a compiler-generated label. The default definition uses
5449 @code{assemble_name_raw}, which is like @code{assemble_name} except
5450 that it is more efficient.
5454 A C string containing the appropriate assembler directive to specify the
5455 size of a symbol, without any arguments. On systems that use ELF, the
5456 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
5457 systems, the default is not to define this macro.
5459 Define this macro only if it is correct to use the default definitions
5460 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
5461 for your system. If you need your own custom definitions of those
5462 macros, or if you do not need explicit symbol sizes at all, do not
5466 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
5467 A C statement (sans semicolon) to output to the stdio stream
5468 @var{stream} a directive telling the assembler that the size of the
5469 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
5470 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5474 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
5475 A C statement (sans semicolon) to output to the stdio stream
5476 @var{stream} a directive telling the assembler to calculate the size of
5477 the symbol @var{name} by subtracting its address from the current
5480 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5481 provided. The default assumes that the assembler recognizes a special
5482 @samp{.} symbol as referring to the current address, and can calculate
5483 the difference between this and another symbol. If your assembler does
5484 not recognize @samp{.} or cannot do calculations with it, you will need
5485 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
5488 @defmac NO_DOLLAR_IN_LABEL
5489 Define this macro if the assembler does not accept the character
5490 @samp{$} in label names. By default constructors and destructors in
5491 G++ have @samp{$} in the identifiers. If this macro is defined,
5492 @samp{.} is used instead.
5495 @defmac NO_DOT_IN_LABEL
5496 Define this macro if the assembler does not accept the character
5497 @samp{.} in label names. By default constructors and destructors in G++
5498 have names that use @samp{.}. If this macro is defined, these names
5499 are rewritten to avoid @samp{.}.
5503 A C string containing the appropriate assembler directive to specify the
5504 type of a symbol, without any arguments. On systems that use ELF, the
5505 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
5506 systems, the default is not to define this macro.
5508 Define this macro only if it is correct to use the default definition of
5509 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5510 custom definition of this macro, or if you do not need explicit symbol
5511 types at all, do not define this macro.
5514 @defmac TYPE_OPERAND_FMT
5515 A C string which specifies (using @code{printf} syntax) the format of
5516 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
5517 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
5518 the default is not to define this macro.
5520 Define this macro only if it is correct to use the default definition of
5521 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5522 custom definition of this macro, or if you do not need explicit symbol
5523 types at all, do not define this macro.
5526 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
5527 A C statement (sans semicolon) to output to the stdio stream
5528 @var{stream} a directive telling the assembler that the type of the
5529 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
5530 that string is always either @samp{"function"} or @samp{"object"}, but
5531 you should not count on this.
5533 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
5534 definition of this macro is provided.
5537 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5538 A C statement (sans semicolon) to output to the stdio stream
5539 @var{stream} any text necessary for declaring the name @var{name} of a
5540 function which is being defined. This macro is responsible for
5541 outputting the label definition (perhaps using
5542 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5543 @code{FUNCTION_DECL} tree node representing the function.
5545 If this macro is not defined, then the function name is defined in the
5546 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
5548 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5552 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5553 A C statement (sans semicolon) to output to the stdio stream
5554 @var{stream} any text necessary for declaring the size of a function
5555 which is being defined. The argument @var{name} is the name of the
5556 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5557 representing the function.
5559 If this macro is not defined, then the function size is not defined.
5561 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5565 @defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5566 A C statement (sans semicolon) to output to the stdio stream
5567 @var{stream} any text necessary for declaring the name @var{name} of a
5568 cold function partition which is being defined. This macro is responsible
5569 for outputting the label definition (perhaps using
5570 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5571 @code{FUNCTION_DECL} tree node representing the function.
5573 If this macro is not defined, then the cold partition name is defined in the
5574 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5576 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5580 @defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5581 A C statement (sans semicolon) to output to the stdio stream
5582 @var{stream} any text necessary for declaring the size of a cold function
5583 partition which is being defined. The argument @var{name} is the name of the
5584 cold partition of the function. The argument @var{decl} is the
5585 @code{FUNCTION_DECL} tree node representing the function.
5587 If this macro is not defined, then the partition size is not defined.
5589 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5593 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5594 A C statement (sans semicolon) to output to the stdio stream
5595 @var{stream} any text necessary for declaring the name @var{name} of an
5596 initialized variable which is being defined. This macro must output the
5597 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5598 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5600 If this macro is not defined, then the variable name is defined in the
5601 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5603 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
5604 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
5607 @hook TARGET_ASM_DECLARE_CONSTANT_NAME
5609 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5610 A C statement (sans semicolon) to output to the stdio stream
5611 @var{stream} any text necessary for claiming a register @var{regno}
5612 for a global variable @var{decl} with name @var{name}.
5614 If you don't define this macro, that is equivalent to defining it to do
5618 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5619 A C statement (sans semicolon) to finish up declaring a variable name
5620 once the compiler has processed its initializer fully and thus has had a
5621 chance to determine the size of an array when controlled by an
5622 initializer. This is used on systems where it's necessary to declare
5623 something about the size of the object.
5625 If you don't define this macro, that is equivalent to defining it to do
5628 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
5629 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
5632 @hook TARGET_ASM_GLOBALIZE_LABEL
5634 @hook TARGET_ASM_GLOBALIZE_DECL_NAME
5636 @hook TARGET_ASM_ASSEMBLE_UNDEFINED_DECL
5638 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
5639 A C statement (sans semicolon) to output to the stdio stream
5640 @var{stream} some commands that will make the label @var{name} weak;
5641 that is, available for reference from other files but only used if
5642 no other definition is available. Use the expression
5643 @code{assemble_name (@var{stream}, @var{name})} to output the name
5644 itself; before and after that, output the additional assembler syntax
5645 for making that name weak, and a newline.
5647 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
5648 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
5652 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
5653 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
5654 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
5655 or variable decl. If @var{value} is not @code{NULL}, this C statement
5656 should output to the stdio stream @var{stream} assembler code which
5657 defines (equates) the weak symbol @var{name} to have the value
5658 @var{value}. If @var{value} is @code{NULL}, it should output commands
5659 to make @var{name} weak.
5662 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
5663 Outputs a directive that enables @var{name} to be used to refer to
5664 symbol @var{value} with weak-symbol semantics. @code{decl} is the
5665 declaration of @code{name}.
5668 @defmac SUPPORTS_WEAK
5669 A preprocessor constant expression which evaluates to true if the target
5670 supports weak symbols.
5672 If you don't define this macro, @file{defaults.h} provides a default
5673 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
5674 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
5677 @defmac TARGET_SUPPORTS_WEAK
5678 A C expression which evaluates to true if the target supports weak symbols.
5680 If you don't define this macro, @file{defaults.h} provides a default
5681 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
5682 this macro if you want to control weak symbol support with a compiler
5683 flag such as @option{-melf}.
5686 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
5687 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5688 public symbol such that extra copies in multiple translation units will
5689 be discarded by the linker. Define this macro if your object file
5690 format provides support for this concept, such as the @samp{COMDAT}
5691 section flags in the Microsoft Windows PE/COFF format, and this support
5692 requires changes to @var{decl}, such as putting it in a separate section.
5695 @defmac SUPPORTS_ONE_ONLY
5696 A C expression which evaluates to true if the target supports one-only
5699 If you don't define this macro, @file{varasm.c} provides a default
5700 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5701 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5702 you want to control one-only symbol support with a compiler flag, or if
5703 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5704 be emitted as one-only.
5707 @hook TARGET_ASM_ASSEMBLE_VISIBILITY
5709 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
5710 A C expression that evaluates to true if the target's linker expects
5711 that weak symbols do not appear in a static archive's table of contents.
5712 The default is @code{0}.
5714 Leaving weak symbols out of an archive's table of contents means that,
5715 if a symbol will only have a definition in one translation unit and
5716 will have undefined references from other translation units, that
5717 symbol should not be weak. Defining this macro to be nonzero will
5718 thus have the effect that certain symbols that would normally be weak
5719 (explicit template instantiations, and vtables for polymorphic classes
5720 with noninline key methods) will instead be nonweak.
5722 The C++ ABI requires this macro to be zero. Define this macro for
5723 targets where full C++ ABI compliance is impossible and where linker
5724 restrictions require weak symbols to be left out of a static archive's
5728 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5729 A C statement (sans semicolon) to output to the stdio stream
5730 @var{stream} any text necessary for declaring the name of an external
5731 symbol named @var{name} which is referenced in this compilation but
5732 not defined. The value of @var{decl} is the tree node for the
5735 This macro need not be defined if it does not need to output anything.
5736 The GNU assembler and most Unix assemblers don't require anything.
5739 @hook TARGET_ASM_EXTERNAL_LIBCALL
5741 @hook TARGET_ASM_MARK_DECL_PRESERVED
5743 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5744 A C statement (sans semicolon) to output to the stdio stream
5745 @var{stream} a reference in assembler syntax to a label named
5746 @var{name}. This should add @samp{_} to the front of the name, if that
5747 is customary on your operating system, as it is in most Berkeley Unix
5748 systems. This macro is used in @code{assemble_name}.
5751 @hook TARGET_MANGLE_ASSEMBLER_NAME
5753 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
5754 A C statement (sans semicolon) to output a reference to
5755 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
5756 will be used to output the name of the symbol. This macro may be used
5757 to modify the way a symbol is referenced depending on information
5758 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
5761 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
5762 A C statement (sans semicolon) to output a reference to @var{buf}, the
5763 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
5764 @code{assemble_name} will be used to output the name of the symbol.
5765 This macro is not used by @code{output_asm_label}, or the @code{%l}
5766 specifier that calls it; the intention is that this macro should be set
5767 when it is necessary to output a label differently when its address is
5771 @hook TARGET_ASM_INTERNAL_LABEL
5773 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
5774 A C statement to output to the stdio stream @var{stream} a debug info
5775 label whose name is made from the string @var{prefix} and the number
5776 @var{num}. This is useful for VLIW targets, where debug info labels
5777 may need to be treated differently than branch target labels. On some
5778 systems, branch target labels must be at the beginning of instruction
5779 bundles, but debug info labels can occur in the middle of instruction
5782 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
5786 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5787 A C statement to store into the string @var{string} a label whose name
5788 is made from the string @var{prefix} and the number @var{num}.
5790 This string, when output subsequently by @code{assemble_name}, should
5791 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
5792 with the same @var{prefix} and @var{num}.
5794 If the string begins with @samp{*}, then @code{assemble_name} will
5795 output the rest of the string unchanged. It is often convenient for
5796 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5797 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5798 to output the string, and may change it. (Of course,
5799 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5800 you should know what it does on your machine.)
5803 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5804 A C expression to assign to @var{outvar} (which is a variable of type
5805 @code{char *}) a newly allocated string made from the string
5806 @var{name} and the number @var{number}, with some suitable punctuation
5807 added. Use @code{alloca} to get space for the string.
5809 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5810 produce an assembler label for an internal static variable whose name is
5811 @var{name}. Therefore, the string must be such as to result in valid
5812 assembler code. The argument @var{number} is different each time this
5813 macro is executed; it prevents conflicts between similarly-named
5814 internal static variables in different scopes.
5816 Ideally this string should not be a valid C identifier, to prevent any
5817 conflict with the user's own symbols. Most assemblers allow periods
5818 or percent signs in assembler symbols; putting at least one of these
5819 between the name and the number will suffice.
5821 If this macro is not defined, a default definition will be provided
5822 which is correct for most systems.
5825 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5826 A C statement to output to the stdio stream @var{stream} assembler code
5827 which defines (equates) the symbol @var{name} to have the value @var{value}.
5830 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5831 correct for most systems.
5834 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
5835 A C statement to output to the stdio stream @var{stream} assembler code
5836 which defines (equates) the symbol whose tree node is @var{decl_of_name}
5837 to have the value of the tree node @var{decl_of_value}. This macro will
5838 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
5839 the tree nodes are available.
5842 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5843 correct for most systems.
5846 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
5847 A C statement that evaluates to true if the assembler code which defines
5848 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
5849 of the tree node @var{decl_of_value} should be emitted near the end of the
5850 current compilation unit. The default is to not defer output of defines.
5851 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
5852 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
5855 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5856 A C statement to output to the stdio stream @var{stream} assembler code
5857 which defines (equates) the weak symbol @var{name} to have the value
5858 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
5859 an undefined weak symbol.
5861 Define this macro if the target only supports weak aliases; define
5862 @code{ASM_OUTPUT_DEF} instead if possible.
5865 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5866 Define this macro to override the default assembler names used for
5867 Objective-C methods.
5869 The default name is a unique method number followed by the name of the
5870 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
5871 the category is also included in the assembler name (e.g.@:
5874 These names are safe on most systems, but make debugging difficult since
5875 the method's selector is not present in the name. Therefore, particular
5876 systems define other ways of computing names.
5878 @var{buf} is an expression of type @code{char *} which gives you a
5879 buffer in which to store the name; its length is as long as
5880 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
5881 50 characters extra.
5883 The argument @var{is_inst} specifies whether the method is an instance
5884 method or a class method; @var{class_name} is the name of the class;
5885 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
5886 in a category); and @var{sel_name} is the name of the selector.
5888 On systems where the assembler can handle quoted names, you can use this
5889 macro to provide more human-readable names.
5892 @node Initialization
5893 @subsection How Initialization Functions Are Handled
5894 @cindex initialization routines
5895 @cindex termination routines
5896 @cindex constructors, output of
5897 @cindex destructors, output of
5899 The compiled code for certain languages includes @dfn{constructors}
5900 (also called @dfn{initialization routines})---functions to initialize
5901 data in the program when the program is started. These functions need
5902 to be called before the program is ``started''---that is to say, before
5903 @code{main} is called.
5905 Compiling some languages generates @dfn{destructors} (also called
5906 @dfn{termination routines}) that should be called when the program
5909 To make the initialization and termination functions work, the compiler
5910 must output something in the assembler code to cause those functions to
5911 be called at the appropriate time. When you port the compiler to a new
5912 system, you need to specify how to do this.
5914 There are two major ways that GCC currently supports the execution of
5915 initialization and termination functions. Each way has two variants.
5916 Much of the structure is common to all four variations.
5918 @findex __CTOR_LIST__
5919 @findex __DTOR_LIST__
5920 The linker must build two lists of these functions---a list of
5921 initialization functions, called @code{__CTOR_LIST__}, and a list of
5922 termination functions, called @code{__DTOR_LIST__}.
5924 Each list always begins with an ignored function pointer (which may hold
5925 0, @minus{}1, or a count of the function pointers after it, depending on
5926 the environment). This is followed by a series of zero or more function
5927 pointers to constructors (or destructors), followed by a function
5928 pointer containing zero.
5930 Depending on the operating system and its executable file format, either
5931 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
5932 time and exit time. Constructors are called in reverse order of the
5933 list; destructors in forward order.
5935 The best way to handle static constructors works only for object file
5936 formats which provide arbitrarily-named sections. A section is set
5937 aside for a list of constructors, and another for a list of destructors.
5938 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
5939 object file that defines an initialization function also puts a word in
5940 the constructor section to point to that function. The linker
5941 accumulates all these words into one contiguous @samp{.ctors} section.
5942 Termination functions are handled similarly.
5944 This method will be chosen as the default by @file{target-def.h} if
5945 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
5946 support arbitrary sections, but does support special designated
5947 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
5948 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
5950 When arbitrary sections are available, there are two variants, depending
5951 upon how the code in @file{crtstuff.c} is called. On systems that
5952 support a @dfn{.init} section which is executed at program startup,
5953 parts of @file{crtstuff.c} are compiled into that section. The
5954 program is linked by the @command{gcc} driver like this:
5957 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
5960 The prologue of a function (@code{__init}) appears in the @code{.init}
5961 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
5962 for the function @code{__fini} in the @dfn{.fini} section. Normally these
5963 files are provided by the operating system or by the GNU C library, but
5964 are provided by GCC for a few targets.
5966 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
5967 compiled from @file{crtstuff.c}. They contain, among other things, code
5968 fragments within the @code{.init} and @code{.fini} sections that branch
5969 to routines in the @code{.text} section. The linker will pull all parts
5970 of a section together, which results in a complete @code{__init} function
5971 that invokes the routines we need at startup.
5973 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
5976 If no init section is available, when GCC compiles any function called
5977 @code{main} (or more accurately, any function designated as a program
5978 entry point by the language front end calling @code{expand_main_function}),
5979 it inserts a procedure call to @code{__main} as the first executable code
5980 after the function prologue. The @code{__main} function is defined
5981 in @file{libgcc2.c} and runs the global constructors.
5983 In file formats that don't support arbitrary sections, there are again
5984 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
5985 and an `a.out' format must be used. In this case,
5986 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
5987 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
5988 and with the address of the void function containing the initialization
5989 code as its value. The GNU linker recognizes this as a request to add
5990 the value to a @dfn{set}; the values are accumulated, and are eventually
5991 placed in the executable as a vector in the format described above, with
5992 a leading (ignored) count and a trailing zero element.
5993 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
5994 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
5995 the compilation of @code{main} to call @code{__main} as above, starting
5996 the initialization process.
5998 The last variant uses neither arbitrary sections nor the GNU linker.
5999 This is preferable when you want to do dynamic linking and when using
6000 file formats which the GNU linker does not support, such as `ECOFF'@. In
6001 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
6002 termination functions are recognized simply by their names. This requires
6003 an extra program in the linkage step, called @command{collect2}. This program
6004 pretends to be the linker, for use with GCC; it does its job by running
6005 the ordinary linker, but also arranges to include the vectors of
6006 initialization and termination functions. These functions are called
6007 via @code{__main} as described above. In order to use this method,
6008 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6011 The following section describes the specific macros that control and
6012 customize the handling of initialization and termination functions.
6015 @node Macros for Initialization
6016 @subsection Macros Controlling Initialization Routines
6018 Here are the macros that control how the compiler handles initialization
6019 and termination functions:
6021 @defmac INIT_SECTION_ASM_OP
6022 If defined, a C string constant, including spacing, for the assembler
6023 operation to identify the following data as initialization code. If not
6024 defined, GCC will assume such a section does not exist. When you are
6025 using special sections for initialization and termination functions, this
6026 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6027 run the initialization functions.
6030 @defmac HAS_INIT_SECTION
6031 If defined, @code{main} will not call @code{__main} as described above.
6032 This macro should be defined for systems that control start-up code
6033 on a symbol-by-symbol basis, such as OSF/1, and should not
6034 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
6037 @defmac LD_INIT_SWITCH
6038 If defined, a C string constant for a switch that tells the linker that
6039 the following symbol is an initialization routine.
6042 @defmac LD_FINI_SWITCH
6043 If defined, a C string constant for a switch that tells the linker that
6044 the following symbol is a finalization routine.
6047 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
6048 If defined, a C statement that will write a function that can be
6049 automatically called when a shared library is loaded. The function
6050 should call @var{func}, which takes no arguments. If not defined, and
6051 the object format requires an explicit initialization function, then a
6052 function called @code{_GLOBAL__DI} will be generated.
6054 This function and the following one are used by collect2 when linking a
6055 shared library that needs constructors or destructors, or has DWARF2
6056 exception tables embedded in the code.
6059 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
6060 If defined, a C statement that will write a function that can be
6061 automatically called when a shared library is unloaded. The function
6062 should call @var{func}, which takes no arguments. If not defined, and
6063 the object format requires an explicit finalization function, then a
6064 function called @code{_GLOBAL__DD} will be generated.
6067 @defmac INVOKE__main
6068 If defined, @code{main} will call @code{__main} despite the presence of
6069 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6070 where the init section is not actually run automatically, but is still
6071 useful for collecting the lists of constructors and destructors.
6074 @defmac SUPPORTS_INIT_PRIORITY
6075 If nonzero, the C++ @code{init_priority} attribute is supported and the
6076 compiler should emit instructions to control the order of initialization
6077 of objects. If zero, the compiler will issue an error message upon
6078 encountering an @code{init_priority} attribute.
6081 @hook TARGET_HAVE_CTORS_DTORS
6083 @hook TARGET_ASM_CONSTRUCTOR
6085 @hook TARGET_ASM_DESTRUCTOR
6087 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
6088 generated for the generated object file will have static linkage.
6090 If your system uses @command{collect2} as the means of processing
6091 constructors, then that program normally uses @command{nm} to scan
6092 an object file for constructor functions to be called.
6094 On certain kinds of systems, you can define this macro to make
6095 @command{collect2} work faster (and, in some cases, make it work at all):
6097 @defmac OBJECT_FORMAT_COFF
6098 Define this macro if the system uses COFF (Common Object File Format)
6099 object files, so that @command{collect2} can assume this format and scan
6100 object files directly for dynamic constructor/destructor functions.
6102 This macro is effective only in a native compiler; @command{collect2} as
6103 part of a cross compiler always uses @command{nm} for the target machine.
6106 @defmac REAL_NM_FILE_NAME
6107 Define this macro as a C string constant containing the file name to use
6108 to execute @command{nm}. The default is to search the path normally for
6113 @command{collect2} calls @command{nm} to scan object files for static
6114 constructors and destructors and LTO info. By default, @option{-n} is
6115 passed. Define @code{NM_FLAGS} to a C string constant if other options
6116 are needed to get the same output format as GNU @command{nm -n}
6120 If your system supports shared libraries and has a program to list the
6121 dynamic dependencies of a given library or executable, you can define
6122 these macros to enable support for running initialization and
6123 termination functions in shared libraries:
6126 Define this macro to a C string constant containing the name of the program
6127 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
6130 @defmac PARSE_LDD_OUTPUT (@var{ptr})
6131 Define this macro to be C code that extracts filenames from the output
6132 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
6133 of type @code{char *} that points to the beginning of a line of output
6134 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6135 code must advance @var{ptr} to the beginning of the filename on that
6136 line. Otherwise, it must set @var{ptr} to @code{NULL}.
6139 @defmac SHLIB_SUFFIX
6140 Define this macro to a C string constant containing the default shared
6141 library extension of the target (e.g., @samp{".so"}). @command{collect2}
6142 strips version information after this suffix when generating global
6143 constructor and destructor names. This define is only needed on targets
6144 that use @command{collect2} to process constructors and destructors.
6147 @node Instruction Output
6148 @subsection Output of Assembler Instructions
6150 @c prevent bad page break with this line
6151 This describes assembler instruction output.
6153 @defmac REGISTER_NAMES
6154 A C initializer containing the assembler's names for the machine
6155 registers, each one as a C string constant. This is what translates
6156 register numbers in the compiler into assembler language.
6159 @defmac ADDITIONAL_REGISTER_NAMES
6160 If defined, a C initializer for an array of structures containing a name
6161 and a register number. This macro defines additional names for hard
6162 registers, thus allowing the @code{asm} option in declarations to refer
6163 to registers using alternate names.
6166 @defmac OVERLAPPING_REGISTER_NAMES
6167 If defined, a C initializer for an array of structures containing a
6168 name, a register number and a count of the number of consecutive
6169 machine registers the name overlaps. This macro defines additional
6170 names for hard registers, thus allowing the @code{asm} option in
6171 declarations to refer to registers using alternate names. Unlike
6172 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
6173 register name implies multiple underlying registers.
6175 This macro should be used when it is important that a clobber in an
6176 @code{asm} statement clobbers all the underlying values implied by the
6177 register name. For example, on ARM, clobbering the double-precision
6178 VFP register ``d0'' implies clobbering both single-precision registers
6182 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6183 Define this macro if you are using an unusual assembler that
6184 requires different names for the machine instructions.
6186 The definition is a C statement or statements which output an
6187 assembler instruction opcode to the stdio stream @var{stream}. The
6188 macro-operand @var{ptr} is a variable of type @code{char *} which
6189 points to the opcode name in its ``internal'' form---the form that is
6190 written in the machine description. The definition should output the
6191 opcode name to @var{stream}, performing any translation you desire, and
6192 increment the variable @var{ptr} to point at the end of the opcode
6193 so that it will not be output twice.
6195 In fact, your macro definition may process less than the entire opcode
6196 name, or more than the opcode name; but if you want to process text
6197 that includes @samp{%}-sequences to substitute operands, you must take
6198 care of the substitution yourself. Just be sure to increment
6199 @var{ptr} over whatever text should not be output normally.
6201 @findex recog_data.operand
6202 If you need to look at the operand values, they can be found as the
6203 elements of @code{recog_data.operand}.
6205 If the macro definition does nothing, the instruction is output
6209 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6210 If defined, a C statement to be executed just prior to the output of
6211 assembler code for @var{insn}, to modify the extracted operands so
6212 they will be output differently.
6214 Here the argument @var{opvec} is the vector containing the operands
6215 extracted from @var{insn}, and @var{noperands} is the number of
6216 elements of the vector which contain meaningful data for this insn.
6217 The contents of this vector are what will be used to convert the insn
6218 template into assembler code, so you can change the assembler output
6219 by changing the contents of the vector.
6221 This macro is useful when various assembler syntaxes share a single
6222 file of instruction patterns; by defining this macro differently, you
6223 can cause a large class of instructions to be output differently (such
6224 as with rearranged operands). Naturally, variations in assembler
6225 syntax affecting individual insn patterns ought to be handled by
6226 writing conditional output routines in those patterns.
6228 If this macro is not defined, it is equivalent to a null statement.
6231 @hook TARGET_ASM_FINAL_POSTSCAN_INSN
6233 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6234 A C compound statement to output to stdio stream @var{stream} the
6235 assembler syntax for an instruction operand @var{x}. @var{x} is an
6238 @var{code} is a value that can be used to specify one of several ways
6239 of printing the operand. It is used when identical operands must be
6240 printed differently depending on the context. @var{code} comes from
6241 the @samp{%} specification that was used to request printing of the
6242 operand. If the specification was just @samp{%@var{digit}} then
6243 @var{code} is 0; if the specification was @samp{%@var{ltr}
6244 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6247 If @var{x} is a register, this macro should print the register's name.
6248 The names can be found in an array @code{reg_names} whose type is
6249 @code{char *[]}. @code{reg_names} is initialized from
6250 @code{REGISTER_NAMES}.
6252 When the machine description has a specification @samp{%@var{punct}}
6253 (a @samp{%} followed by a punctuation character), this macro is called
6254 with a null pointer for @var{x} and the punctuation character for
6258 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6259 A C expression which evaluates to true if @var{code} is a valid
6260 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6261 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6262 punctuation characters (except for the standard one, @samp{%}) are used
6266 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6267 A C compound statement to output to stdio stream @var{stream} the
6268 assembler syntax for an instruction operand that is a memory reference
6269 whose address is @var{x}. @var{x} is an RTL expression.
6271 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
6272 On some machines, the syntax for a symbolic address depends on the
6273 section that the address refers to. On these machines, define the hook
6274 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
6275 @code{symbol_ref}, and then check for it here. @xref{Assembler
6279 @findex dbr_sequence_length
6280 @defmac DBR_OUTPUT_SEQEND (@var{file})
6281 A C statement, to be executed after all slot-filler instructions have
6282 been output. If necessary, call @code{dbr_sequence_length} to
6283 determine the number of slots filled in a sequence (zero if not
6284 currently outputting a sequence), to decide how many no-ops to output,
6287 Don't define this macro if it has nothing to do, but it is helpful in
6288 reading assembly output if the extent of the delay sequence is made
6289 explicit (e.g.@: with white space).
6292 @findex final_sequence
6293 Note that output routines for instructions with delay slots must be
6294 prepared to deal with not being output as part of a sequence
6295 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
6296 found.) The variable @code{final_sequence} is null when not
6297 processing a sequence, otherwise it contains the @code{sequence} rtx
6301 @defmac REGISTER_PREFIX
6302 @defmacx LOCAL_LABEL_PREFIX
6303 @defmacx USER_LABEL_PREFIX
6304 @defmacx IMMEDIATE_PREFIX
6305 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6306 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6307 @file{final.c}). These are useful when a single @file{md} file must
6308 support multiple assembler formats. In that case, the various @file{tm.h}
6309 files can define these macros differently.
6312 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
6313 If defined this macro should expand to a series of @code{case}
6314 statements which will be parsed inside the @code{switch} statement of
6315 the @code{asm_fprintf} function. This allows targets to define extra
6316 printf formats which may useful when generating their assembler
6317 statements. Note that uppercase letters are reserved for future
6318 generic extensions to asm_fprintf, and so are not available to target
6319 specific code. The output file is given by the parameter @var{file}.
6320 The varargs input pointer is @var{argptr} and the rest of the format
6321 string, starting the character after the one that is being switched
6322 upon, is pointed to by @var{format}.
6325 @defmac ASSEMBLER_DIALECT
6326 If your target supports multiple dialects of assembler language (such as
6327 different opcodes), define this macro as a C expression that gives the
6328 numeric index of the assembler language dialect to use, with zero as the
6331 If this macro is defined, you may use constructs of the form
6333 @samp{@{option0|option1|option2@dots{}@}}
6336 in the output templates of patterns (@pxref{Output Template}) or in the
6337 first argument of @code{asm_fprintf}. This construct outputs
6338 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
6339 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
6340 within these strings retain their usual meaning. If there are fewer
6341 alternatives within the braces than the value of
6342 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
6343 to print curly braces or @samp{|} character in assembler output directly,
6344 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
6346 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6347 @samp{@}} do not have any special meaning when used in templates or
6348 operands to @code{asm_fprintf}.
6350 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6351 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6352 the variations in assembler language syntax with that mechanism. Define
6353 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6354 if the syntax variant are larger and involve such things as different
6355 opcodes or operand order.
6358 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6359 A C expression to output to @var{stream} some assembler code
6360 which will push hard register number @var{regno} onto the stack.
6361 The code need not be optimal, since this macro is used only when
6365 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6366 A C expression to output to @var{stream} some assembler code
6367 which will pop hard register number @var{regno} off of the stack.
6368 The code need not be optimal, since this macro is used only when
6372 @node Dispatch Tables
6373 @subsection Output of Dispatch Tables
6375 @c prevent bad page break with this line
6376 This concerns dispatch tables.
6378 @cindex dispatch table
6379 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6380 A C statement to output to the stdio stream @var{stream} an assembler
6381 pseudo-instruction to generate a difference between two labels.
6382 @var{value} and @var{rel} are the numbers of two internal labels. The
6383 definitions of these labels are output using
6384 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
6385 way here. For example,
6388 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6389 @var{value}, @var{rel})
6392 You must provide this macro on machines where the addresses in a
6393 dispatch table are relative to the table's own address. If defined, GCC
6394 will also use this macro on all machines when producing PIC@.
6395 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
6396 mode and flags can be read.
6399 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6400 This macro should be provided on machines where the addresses
6401 in a dispatch table are absolute.
6403 The definition should be a C statement to output to the stdio stream
6404 @var{stream} an assembler pseudo-instruction to generate a reference to
6405 a label. @var{value} is the number of an internal label whose
6406 definition is output using @code{(*targetm.asm_out.internal_label)}.
6410 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6414 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6415 Define this if the label before a jump-table needs to be output
6416 specially. The first three arguments are the same as for
6417 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
6418 jump-table which follows (a @code{jump_table_data} containing an
6419 @code{addr_vec} or @code{addr_diff_vec}).
6421 This feature is used on system V to output a @code{swbeg} statement
6424 If this macro is not defined, these labels are output with
6425 @code{(*targetm.asm_out.internal_label)}.
6428 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6429 Define this if something special must be output at the end of a
6430 jump-table. The definition should be a C statement to be executed
6431 after the assembler code for the table is written. It should write
6432 the appropriate code to stdio stream @var{stream}. The argument
6433 @var{table} is the jump-table insn, and @var{num} is the label-number
6434 of the preceding label.
6436 If this macro is not defined, nothing special is output at the end of
6440 @hook TARGET_ASM_POST_CFI_STARTPROC
6442 @hook TARGET_ASM_EMIT_UNWIND_LABEL
6444 @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
6446 @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
6448 @hook TARGET_ASM_UNWIND_EMIT
6450 @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
6452 @node Exception Region Output
6453 @subsection Assembler Commands for Exception Regions
6455 @c prevent bad page break with this line
6457 This describes commands marking the start and the end of an exception
6460 @defmac EH_FRAME_SECTION_NAME
6461 If defined, a C string constant for the name of the section containing
6462 exception handling frame unwind information. If not defined, GCC will
6463 provide a default definition if the target supports named sections.
6464 @file{crtstuff.c} uses this macro to switch to the appropriate section.
6466 You should define this symbol if your target supports DWARF 2 frame
6467 unwind information and the default definition does not work.
6470 @defmac EH_FRAME_THROUGH_COLLECT2
6471 If defined, DWARF 2 frame unwind information will identified by
6472 specially named labels. The collect2 process will locate these
6473 labels and generate code to register the frames.
6475 This might be necessary, for instance, if the system linker will not
6476 place the eh_frames in-between the sentinals from @file{crtstuff.c},
6477 or if the system linker does garbage collection and sections cannot
6478 be marked as not to be collected.
6481 @defmac EH_TABLES_CAN_BE_READ_ONLY
6482 Define this macro to 1 if your target is such that no frame unwind
6483 information encoding used with non-PIC code will ever require a
6484 runtime relocation, but the linker may not support merging read-only
6485 and read-write sections into a single read-write section.
6488 @defmac MASK_RETURN_ADDR
6489 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
6490 that it does not contain any extraneous set bits in it.
6493 @defmac DWARF2_UNWIND_INFO
6494 Define this macro to 0 if your target supports DWARF 2 frame unwind
6495 information, but it does not yet work with exception handling.
6496 Otherwise, if your target supports this information (if it defines
6497 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
6498 GCC will provide a default definition of 1.
6501 @hook TARGET_EXCEPT_UNWIND_INFO
6502 This hook defines the mechanism that will be used for exception handling
6503 by the target. If the target has ABI specified unwind tables, the hook
6504 should return @code{UI_TARGET}. If the target is to use the
6505 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
6506 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
6507 information, the hook should return @code{UI_DWARF2}.
6509 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
6510 This may end up simplifying other parts of target-specific code. The
6511 default implementation of this hook never returns @code{UI_NONE}.
6513 Note that the value returned by this hook should be constant. It should
6514 not depend on anything except the command-line switches described by
6515 @var{opts}. In particular, the
6516 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
6517 macros and builtin functions related to exception handling are set up
6518 depending on this setting.
6520 The default implementation of the hook first honors the
6521 @option{--enable-sjlj-exceptions} configure option, then
6522 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
6523 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
6524 must define this hook so that @var{opts} is used correctly.
6527 @hook TARGET_UNWIND_TABLES_DEFAULT
6528 This variable should be set to @code{true} if the target ABI requires unwinding
6529 tables even when exceptions are not used. It must not be modified by
6530 command-line option processing.
6533 @defmac DONT_USE_BUILTIN_SETJMP
6534 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
6535 should use the @code{setjmp}/@code{longjmp} functions from the C library
6536 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
6539 @defmac JMP_BUF_SIZE
6540 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
6541 defined. Define this macro if the default size of @code{jmp_buf} buffer
6542 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
6543 is not large enough, or if it is much too large.
6544 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
6547 @defmac DWARF_CIE_DATA_ALIGNMENT
6548 This macro need only be defined if the target might save registers in the
6549 function prologue at an offset to the stack pointer that is not aligned to
6550 @code{UNITS_PER_WORD}. The definition should be the negative minimum
6551 alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive
6552 minimum alignment otherwise. @xref{DWARF}. Only applicable if
6553 the target supports DWARF 2 frame unwind information.
6556 @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
6558 @hook TARGET_DWARF_REGISTER_SPAN
6560 @hook TARGET_DWARF_FRAME_REG_MODE
6562 @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
6564 @hook TARGET_ASM_TTYPE
6566 @hook TARGET_ARM_EABI_UNWINDER
6568 @node Alignment Output
6569 @subsection Assembler Commands for Alignment
6571 @c prevent bad page break with this line
6572 This describes commands for alignment.
6574 @defmac JUMP_ALIGN (@var{label})
6575 The alignment (log base 2) to put in front of @var{label}, which is
6576 a common destination of jumps and has no fallthru incoming edge.
6578 This macro need not be defined if you don't want any special alignment
6579 to be done at such a time. Most machine descriptions do not currently
6582 Unless it's necessary to inspect the @var{label} parameter, it is better
6583 to set the variable @var{align_jumps} in the target's
6584 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6585 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
6588 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
6589 The alignment (log base 2) to put in front of @var{label}, which follows
6592 This macro need not be defined if you don't want any special alignment
6593 to be done at such a time. Most machine descriptions do not currently
6597 @defmac LOOP_ALIGN (@var{label})
6598 The alignment (log base 2) to put in front of @var{label} that heads
6599 a frequently executed basic block (usually the header of a loop).
6601 This macro need not be defined if you don't want any special alignment
6602 to be done at such a time. Most machine descriptions do not currently
6605 Unless it's necessary to inspect the @var{label} parameter, it is better
6606 to set the variable @code{align_loops} in the target's
6607 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6608 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
6611 @defmac LABEL_ALIGN (@var{label})
6612 The alignment (log base 2) to put in front of @var{label}.
6613 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
6614 the maximum of the specified values is used.
6616 Unless it's necessary to inspect the @var{label} parameter, it is better
6617 to set the variable @code{align_labels} in the target's
6618 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6619 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
6622 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6623 A C statement to output to the stdio stream @var{stream} an assembler
6624 instruction to advance the location counter by @var{nbytes} bytes.
6625 Those bytes should be zero when loaded. @var{nbytes} will be a C
6626 expression of type @code{unsigned HOST_WIDE_INT}.
6629 @defmac ASM_NO_SKIP_IN_TEXT
6630 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6631 text section because it fails to put zeros in the bytes that are skipped.
6632 This is true on many Unix systems, where the pseudo--op to skip bytes
6633 produces no-op instructions rather than zeros when used in the text
6637 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6638 A C statement to output to the stdio stream @var{stream} an assembler
6639 command to advance the location counter to a multiple of 2 to the
6640 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6643 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
6644 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
6645 for padding, if necessary.
6648 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6649 A C statement to output to the stdio stream @var{stream} an assembler
6650 command to advance the location counter to a multiple of 2 to the
6651 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6652 satisfy the alignment request. @var{power} and @var{max_skip} will be
6653 a C expression of type @code{int}.
6657 @node Debugging Info
6658 @section Controlling Debugging Information Format
6660 @c prevent bad page break with this line
6661 This describes how to specify debugging information.
6664 * All Debuggers:: Macros that affect all debugging formats uniformly.
6665 * DBX Options:: Macros enabling specific options in DBX format.
6666 * DBX Hooks:: Hook macros for varying DBX format.
6667 * File Names and DBX:: Macros controlling output of file names in DBX format.
6668 * DWARF:: Macros for DWARF format.
6669 * VMS Debug:: Macros for VMS debug format.
6673 @subsection Macros Affecting All Debugging Formats
6675 @c prevent bad page break with this line
6676 These macros affect all debugging formats.
6678 @defmac DBX_REGISTER_NUMBER (@var{regno})
6679 A C expression that returns the DBX register number for the compiler
6680 register number @var{regno}. In the default macro provided, the value
6681 of this expression will be @var{regno} itself. But sometimes there are
6682 some registers that the compiler knows about and DBX does not, or vice
6683 versa. In such cases, some register may need to have one number in the
6684 compiler and another for DBX@.
6686 If two registers have consecutive numbers inside GCC, and they can be
6687 used as a pair to hold a multiword value, then they @emph{must} have
6688 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6689 Otherwise, debuggers will be unable to access such a pair, because they
6690 expect register pairs to be consecutive in their own numbering scheme.
6692 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6693 does not preserve register pairs, then what you must do instead is
6694 redefine the actual register numbering scheme.
6697 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
6698 A C expression that returns the integer offset value for an automatic
6699 variable having address @var{x} (an RTL expression). The default
6700 computation assumes that @var{x} is based on the frame-pointer and
6701 gives the offset from the frame-pointer. This is required for targets
6702 that produce debugging output for DBX and allow the frame-pointer to be
6703 eliminated when the @option{-g} option is used.
6706 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6707 A C expression that returns the integer offset value for an argument
6708 having address @var{x} (an RTL expression). The nominal offset is
6712 @defmac PREFERRED_DEBUGGING_TYPE
6713 A C expression that returns the type of debugging output GCC should
6714 produce when the user specifies just @option{-g}. Define
6715 this if you have arranged for GCC to support more than one format of
6716 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
6717 @code{DWARF2_DEBUG}, @code{XCOFF_DEBUG}, @code{VMS_DEBUG},
6718 and @code{VMS_AND_DWARF2_DEBUG}.
6720 When the user specifies @option{-ggdb}, GCC normally also uses the
6721 value of this macro to select the debugging output format, but with two
6722 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
6723 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
6724 defined, GCC uses @code{DBX_DEBUG}.
6726 The value of this macro only affects the default debugging output; the
6727 user can always get a specific type of output by using @option{-gstabs},
6728 @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
6732 @subsection Specific Options for DBX Output
6734 @c prevent bad page break with this line
6735 These are specific options for DBX output.
6737 @defmac DBX_DEBUGGING_INFO
6738 Define this macro if GCC should produce debugging output for DBX
6739 in response to the @option{-g} option.
6742 @defmac XCOFF_DEBUGGING_INFO
6743 Define this macro if GCC should produce XCOFF format debugging output
6744 in response to the @option{-g} option. This is a variant of DBX format.
6747 @defmac DEFAULT_GDB_EXTENSIONS
6748 Define this macro to control whether GCC should by default generate
6749 GDB's extended version of DBX debugging information (assuming DBX-format
6750 debugging information is enabled at all). If you don't define the
6751 macro, the default is 1: always generate the extended information
6752 if there is any occasion to.
6755 @defmac DEBUG_SYMS_TEXT
6756 Define this macro if all @code{.stabs} commands should be output while
6757 in the text section.
6760 @defmac ASM_STABS_OP
6761 A C string constant, including spacing, naming the assembler pseudo op to
6762 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
6763 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
6764 applies only to DBX debugging information format.
6767 @defmac ASM_STABD_OP
6768 A C string constant, including spacing, naming the assembler pseudo op to
6769 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
6770 value is the current location. If you don't define this macro,
6771 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
6775 @defmac ASM_STABN_OP
6776 A C string constant, including spacing, naming the assembler pseudo op to
6777 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
6778 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
6779 macro applies only to DBX debugging information format.
6782 @defmac DBX_NO_XREFS
6783 Define this macro if DBX on your system does not support the construct
6784 @samp{xs@var{tagname}}. On some systems, this construct is used to
6785 describe a forward reference to a structure named @var{tagname}.
6786 On other systems, this construct is not supported at all.
6789 @defmac DBX_CONTIN_LENGTH
6790 A symbol name in DBX-format debugging information is normally
6791 continued (split into two separate @code{.stabs} directives) when it
6792 exceeds a certain length (by default, 80 characters). On some
6793 operating systems, DBX requires this splitting; on others, splitting
6794 must not be done. You can inhibit splitting by defining this macro
6795 with the value zero. You can override the default splitting-length by
6796 defining this macro as an expression for the length you desire.
6799 @defmac DBX_CONTIN_CHAR
6800 Normally continuation is indicated by adding a @samp{\} character to
6801 the end of a @code{.stabs} string when a continuation follows. To use
6802 a different character instead, define this macro as a character
6803 constant for the character you want to use. Do not define this macro
6804 if backslash is correct for your system.
6807 @defmac DBX_STATIC_STAB_DATA_SECTION
6808 Define this macro if it is necessary to go to the data section before
6809 outputting the @samp{.stabs} pseudo-op for a non-global static
6813 @defmac DBX_TYPE_DECL_STABS_CODE
6814 The value to use in the ``code'' field of the @code{.stabs} directive
6815 for a typedef. The default is @code{N_LSYM}.
6818 @defmac DBX_STATIC_CONST_VAR_CODE
6819 The value to use in the ``code'' field of the @code{.stabs} directive
6820 for a static variable located in the text section. DBX format does not
6821 provide any ``right'' way to do this. The default is @code{N_FUN}.
6824 @defmac DBX_REGPARM_STABS_CODE
6825 The value to use in the ``code'' field of the @code{.stabs} directive
6826 for a parameter passed in registers. DBX format does not provide any
6827 ``right'' way to do this. The default is @code{N_RSYM}.
6830 @defmac DBX_REGPARM_STABS_LETTER
6831 The letter to use in DBX symbol data to identify a symbol as a parameter
6832 passed in registers. DBX format does not customarily provide any way to
6833 do this. The default is @code{'P'}.
6836 @defmac DBX_FUNCTION_FIRST
6837 Define this macro if the DBX information for a function and its
6838 arguments should precede the assembler code for the function. Normally,
6839 in DBX format, the debugging information entirely follows the assembler
6843 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
6844 Define this macro, with value 1, if the value of a symbol describing
6845 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
6846 relative to the start of the enclosing function. Normally, GCC uses
6847 an absolute address.
6850 @defmac DBX_LINES_FUNCTION_RELATIVE
6851 Define this macro, with value 1, if the value of a symbol indicating
6852 the current line number (@code{N_SLINE}) should be relative to the
6853 start of the enclosing function. Normally, GCC uses an absolute address.
6856 @defmac DBX_USE_BINCL
6857 Define this macro if GCC should generate @code{N_BINCL} and
6858 @code{N_EINCL} stabs for included header files, as on Sun systems. This
6859 macro also directs GCC to output a type number as a pair of a file
6860 number and a type number within the file. Normally, GCC does not
6861 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6862 number for a type number.
6866 @subsection Open-Ended Hooks for DBX Format
6868 @c prevent bad page break with this line
6869 These are hooks for DBX format.
6871 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6872 A C statement to output DBX debugging information before code for line
6873 number @var{line} of the current source file to the stdio stream
6874 @var{stream}. @var{counter} is the number of time the macro was
6875 invoked, including the current invocation; it is intended to generate
6876 unique labels in the assembly output.
6878 This macro should not be defined if the default output is correct, or
6879 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
6882 @defmac NO_DBX_FUNCTION_END
6883 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
6884 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
6885 On those machines, define this macro to turn this feature off without
6886 disturbing the rest of the gdb extensions.
6889 @defmac NO_DBX_BNSYM_ENSYM
6890 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
6891 extension construct. On those machines, define this macro to turn this
6892 feature off without disturbing the rest of the gdb extensions.
6895 @node File Names and DBX
6896 @subsection File Names in DBX Format
6898 @c prevent bad page break with this line
6899 This describes file names in DBX format.
6901 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
6902 A C statement to output DBX debugging information to the stdio stream
6903 @var{stream}, which indicates that file @var{name} is the main source
6904 file---the file specified as the input file for compilation.
6905 This macro is called only once, at the beginning of compilation.
6907 This macro need not be defined if the standard form of output
6908 for DBX debugging information is appropriate.
6910 It may be necessary to refer to a label equal to the beginning of the
6911 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
6912 to do so. If you do this, you must also set the variable
6913 @var{used_ltext_label_name} to @code{true}.
6916 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
6917 Define this macro, with value 1, if GCC should not emit an indication
6918 of the current directory for compilation and current source language at
6919 the beginning of the file.
6922 @defmac NO_DBX_GCC_MARKER
6923 Define this macro, with value 1, if GCC should not emit an indication
6924 that this object file was compiled by GCC@. The default is to emit
6925 an @code{N_OPT} stab at the beginning of every source file, with
6926 @samp{gcc2_compiled.} for the string and value 0.
6929 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
6930 A C statement to output DBX debugging information at the end of
6931 compilation of the main source file @var{name}. Output should be
6932 written to the stdio stream @var{stream}.
6934 If you don't define this macro, nothing special is output at the end
6935 of compilation, which is correct for most machines.
6938 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
6939 Define this macro @emph{instead of} defining
6940 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
6941 the end of compilation is an @code{N_SO} stab with an empty string,
6942 whose value is the highest absolute text address in the file.
6947 @subsection Macros for DWARF Output
6949 @c prevent bad page break with this line
6950 Here are macros for DWARF output.
6952 @defmac DWARF2_DEBUGGING_INFO
6953 Define this macro if GCC should produce dwarf version 2 format
6954 debugging output in response to the @option{-g} option.
6956 @hook TARGET_DWARF_CALLING_CONVENTION
6958 To support optional call frame debugging information, you must also
6959 define @code{INCOMING_RETURN_ADDR_RTX} and either set
6960 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
6961 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
6962 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
6965 @defmac DWARF2_FRAME_INFO
6966 Define this macro to a nonzero value if GCC should always output
6967 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
6968 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
6969 exceptions are enabled, GCC will output this information not matter
6970 how you define @code{DWARF2_FRAME_INFO}.
6973 @hook TARGET_DEBUG_UNWIND_INFO
6975 @defmac DWARF2_ASM_LINE_DEBUG_INFO
6976 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
6977 line debug info sections. This will result in much more compact line number
6978 tables, and hence is desirable if it works.
6981 @defmac DWARF2_ASM_VIEW_DEBUG_INFO
6982 Define this macro to be a nonzero value if the assembler supports view
6983 assignment and verification in @code{.loc}. If it does not, but the
6984 user enables location views, the compiler may have to fallback to
6985 internal line number tables.
6988 @hook TARGET_RESET_LOCATION_VIEW
6990 @hook TARGET_WANT_DEBUG_PUB_SECTIONS
6992 @hook TARGET_DELAY_SCHED2
6994 @hook TARGET_DELAY_VARTRACK
6996 @hook TARGET_NO_REGISTER_ALLOCATION
6998 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
6999 A C statement to issue assembly directives that create a difference
7000 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
7003 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
7004 A C statement to issue assembly directives that create a difference
7005 between the two given labels in system defined units, e.g.@: instruction
7006 slots on IA64 VMS, using an integer of the given size.
7009 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{offset}, @var{section})
7010 A C statement to issue assembly directives that create a
7011 section-relative reference to the given @var{label} plus @var{offset}, using
7012 an integer of the given @var{size}. The label is known to be defined in the
7013 given @var{section}.
7016 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
7017 A C statement to issue assembly directives that create a self-relative
7018 reference to the given @var{label}, using an integer of the given @var{size}.
7021 @defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label})
7022 A C statement to issue assembly directives that create a reference to the
7023 given @var{label} relative to the dbase, using an integer of the given @var{size}.
7026 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
7027 A C statement to issue assembly directives that create a reference to
7028 the DWARF table identifier @var{label} from the current section. This
7029 is used on some systems to avoid garbage collecting a DWARF table which
7030 is referenced by a function.
7033 @hook TARGET_ASM_OUTPUT_DWARF_DTPREL
7037 @subsection Macros for VMS Debug Format
7039 @c prevent bad page break with this line
7040 Here are macros for VMS debug format.
7042 @defmac VMS_DEBUGGING_INFO
7043 Define this macro if GCC should produce debugging output for VMS
7044 in response to the @option{-g} option. The default behavior for VMS
7045 is to generate minimal debug info for a traceback in the absence of
7046 @option{-g} unless explicitly overridden with @option{-g0}. This
7047 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
7048 @code{TARGET_OPTION_OVERRIDE}.
7051 @node Floating Point
7052 @section Cross Compilation and Floating Point
7053 @cindex cross compilation and floating point
7054 @cindex floating point and cross compilation
7056 While all modern machines use twos-complement representation for integers,
7057 there are a variety of representations for floating point numbers. This
7058 means that in a cross-compiler the representation of floating point numbers
7059 in the compiled program may be different from that used in the machine
7060 doing the compilation.
7062 Because different representation systems may offer different amounts of
7063 range and precision, all floating point constants must be represented in
7064 the target machine's format. Therefore, the cross compiler cannot
7065 safely use the host machine's floating point arithmetic; it must emulate
7066 the target's arithmetic. To ensure consistency, GCC always uses
7067 emulation to work with floating point values, even when the host and
7068 target floating point formats are identical.
7070 The following macros are provided by @file{real.h} for the compiler to
7071 use. All parts of the compiler which generate or optimize
7072 floating-point calculations must use these macros. They may evaluate
7073 their operands more than once, so operands must not have side effects.
7075 @defmac REAL_VALUE_TYPE
7076 The C data type to be used to hold a floating point value in the target
7077 machine's format. Typically this is a @code{struct} containing an
7078 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
7082 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
7083 Truncates @var{x} to a signed integer, rounding toward zero.
7086 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
7087 Truncates @var{x} to an unsigned integer, rounding toward zero. If
7088 @var{x} is negative, returns zero.
7091 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode})
7092 Converts @var{string} into a floating point number in the target machine's
7093 representation for mode @var{mode}. This routine can handle both
7094 decimal and hexadecimal floating point constants, using the syntax
7095 defined by the C language for both.
7098 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
7099 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
7102 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
7103 Determines whether @var{x} represents infinity (positive or negative).
7106 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
7107 Determines whether @var{x} represents a ``NaN'' (not-a-number).
7110 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
7111 Returns the negative of the floating point value @var{x}.
7114 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
7115 Returns the absolute value of @var{x}.
7118 @node Mode Switching
7119 @section Mode Switching Instructions
7120 @cindex mode switching
7121 The following macros control mode switching optimizations:
7123 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
7124 Define this macro if the port needs extra instructions inserted for mode
7125 switching in an optimizing compilation.
7127 For an example, the SH4 can perform both single and double precision
7128 floating point operations, but to perform a single precision operation,
7129 the FPSCR PR bit has to be cleared, while for a double precision
7130 operation, this bit has to be set. Changing the PR bit requires a general
7131 purpose register as a scratch register, hence these FPSCR sets have to
7132 be inserted before reload, i.e.@: you cannot put this into instruction emitting
7133 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
7135 You can have multiple entities that are mode-switched, and select at run time
7136 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7137 return nonzero for any @var{entity} that needs mode-switching.
7138 If you define this macro, you also have to define
7139 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
7140 @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
7141 @code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
7145 @defmac NUM_MODES_FOR_MODE_SWITCHING
7146 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7147 initializer for an array of integers. Each initializer element
7148 N refers to an entity that needs mode switching, and specifies the number
7149 of different modes that might need to be set for this entity.
7150 The position of the initializer in the initializer---starting counting at
7151 zero---determines the integer that is used to refer to the mode-switched
7153 In macros that take mode arguments / yield a mode result, modes are
7154 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
7155 switch is needed / supplied.
7158 @hook TARGET_MODE_EMIT
7160 @hook TARGET_MODE_NEEDED
7162 @hook TARGET_MODE_AFTER
7164 @hook TARGET_MODE_ENTRY
7166 @hook TARGET_MODE_EXIT
7168 @hook TARGET_MODE_PRIORITY
7170 @node Target Attributes
7171 @section Defining target-specific uses of @code{__attribute__}
7172 @cindex target attributes
7173 @cindex machine attributes
7174 @cindex attributes, target-specific
7176 Target-specific attributes may be defined for functions, data and types.
7177 These are described using the following target hooks; they also need to
7178 be documented in @file{extend.texi}.
7180 @hook TARGET_ATTRIBUTE_TABLE
7182 @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
7184 @hook TARGET_COMP_TYPE_ATTRIBUTES
7186 @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
7188 @hook TARGET_MERGE_TYPE_ATTRIBUTES
7190 @hook TARGET_MERGE_DECL_ATTRIBUTES
7192 @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
7194 @defmac TARGET_DECLSPEC
7195 Define this macro to a nonzero value if you want to treat
7196 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
7197 default, this behavior is enabled only for targets that define
7198 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
7199 of @code{__declspec} is via a built-in macro, but you should not rely
7200 on this implementation detail.
7203 @hook TARGET_INSERT_ATTRIBUTES
7205 @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
7207 @hook TARGET_OPTION_VALID_ATTRIBUTE_P
7209 @hook TARGET_OPTION_SAVE
7211 @hook TARGET_OPTION_RESTORE
7213 @hook TARGET_OPTION_POST_STREAM_IN
7215 @hook TARGET_OPTION_PRINT
7217 @hook TARGET_OPTION_PRAGMA_PARSE
7219 @hook TARGET_OPTION_OVERRIDE
7221 @hook TARGET_OPTION_FUNCTION_VERSIONS
7223 @hook TARGET_CAN_INLINE_P
7225 @hook TARGET_RELAYOUT_FUNCTION
7228 @section Emulating TLS
7229 @cindex Emulated TLS
7231 For targets whose psABI does not provide Thread Local Storage via
7232 specific relocations and instruction sequences, an emulation layer is
7233 used. A set of target hooks allows this emulation layer to be
7234 configured for the requirements of a particular target. For instance
7235 the psABI may in fact specify TLS support in terms of an emulation
7238 The emulation layer works by creating a control object for every TLS
7239 object. To access the TLS object, a lookup function is provided
7240 which, when given the address of the control object, will return the
7241 address of the current thread's instance of the TLS object.
7243 @hook TARGET_EMUTLS_GET_ADDRESS
7245 @hook TARGET_EMUTLS_REGISTER_COMMON
7247 @hook TARGET_EMUTLS_VAR_SECTION
7249 @hook TARGET_EMUTLS_TMPL_SECTION
7251 @hook TARGET_EMUTLS_VAR_PREFIX
7253 @hook TARGET_EMUTLS_TMPL_PREFIX
7255 @hook TARGET_EMUTLS_VAR_FIELDS
7257 @hook TARGET_EMUTLS_VAR_INIT
7259 @hook TARGET_EMUTLS_VAR_ALIGN_FIXED
7261 @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
7263 @node MIPS Coprocessors
7264 @section Defining coprocessor specifics for MIPS targets.
7265 @cindex MIPS coprocessor-definition macros
7267 The MIPS specification allows MIPS implementations to have as many as 4
7268 coprocessors, each with as many as 32 private registers. GCC supports
7269 accessing these registers and transferring values between the registers
7270 and memory using asm-ized variables. For example:
7273 register unsigned int cp0count asm ("c0r1");
7279 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
7280 names may be added as described below, or the default names may be
7281 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
7283 Coprocessor registers are assumed to be epilogue-used; sets to them will
7284 be preserved even if it does not appear that the register is used again
7285 later in the function.
7287 Another note: according to the MIPS spec, coprocessor 1 (if present) is
7288 the FPU@. One accesses COP1 registers through standard mips
7289 floating-point support; they are not included in this mechanism.
7292 @section Parameters for Precompiled Header Validity Checking
7293 @cindex parameters, precompiled headers
7295 @hook TARGET_GET_PCH_VALIDITY
7297 @hook TARGET_PCH_VALID_P
7299 @hook TARGET_CHECK_PCH_TARGET_FLAGS
7301 @hook TARGET_PREPARE_PCH_SAVE
7304 @section C++ ABI parameters
7305 @cindex parameters, c++ abi
7307 @hook TARGET_CXX_GUARD_TYPE
7309 @hook TARGET_CXX_GUARD_MASK_BIT
7311 @hook TARGET_CXX_GET_COOKIE_SIZE
7313 @hook TARGET_CXX_COOKIE_HAS_SIZE
7315 @hook TARGET_CXX_IMPORT_EXPORT_CLASS
7317 @hook TARGET_CXX_CDTOR_RETURNS_THIS
7319 @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
7321 @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
7323 @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
7325 @hook TARGET_CXX_LIBRARY_RTTI_COMDAT
7327 @hook TARGET_CXX_USE_AEABI_ATEXIT
7329 @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
7331 @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
7333 @hook TARGET_CXX_DECL_MANGLING_CONTEXT
7335 @node D Language and ABI
7336 @section D ABI parameters
7337 @cindex parameters, d abi
7339 @hook TARGET_D_CPU_VERSIONS
7341 @hook TARGET_D_OS_VERSIONS
7343 @hook TARGET_D_CRITSEC_SIZE
7345 @node Named Address Spaces
7346 @section Adding support for named address spaces
7347 @cindex named address spaces
7349 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
7350 standards committee, @cite{Programming Languages - C - Extensions to
7351 support embedded processors}, specifies a syntax for embedded
7352 processors to specify alternate address spaces. You can configure a
7353 GCC port to support section 5.1 of the draft report to add support for
7354 address spaces other than the default address space. These address
7355 spaces are new keywords that are similar to the @code{volatile} and
7356 @code{const} type attributes.
7358 Pointers to named address spaces can have a different size than
7359 pointers to the generic address space.
7361 For example, the SPU port uses the @code{__ea} address space to refer
7362 to memory in the host processor, rather than memory local to the SPU
7363 processor. Access to memory in the @code{__ea} address space involves
7364 issuing DMA operations to move data between the host processor and the
7365 local processor memory address space. Pointers in the @code{__ea}
7366 address space are either 32 bits or 64 bits based on the
7367 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
7370 Internally, address spaces are represented as a small integer in the
7371 range 0 to 15 with address space 0 being reserved for the generic
7374 To register a named address space qualifier keyword with the C front end,
7375 the target may call the @code{c_register_addr_space} routine. For example,
7376 the SPU port uses the following to declare @code{__ea} as the keyword for
7377 named address space #1:
7379 #define ADDR_SPACE_EA 1
7380 c_register_addr_space ("__ea", ADDR_SPACE_EA);
7383 @hook TARGET_ADDR_SPACE_POINTER_MODE
7385 @hook TARGET_ADDR_SPACE_ADDRESS_MODE
7387 @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
7389 @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
7391 @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
7393 @hook TARGET_ADDR_SPACE_SUBSET_P
7395 @hook TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID
7397 @hook TARGET_ADDR_SPACE_CONVERT
7399 @hook TARGET_ADDR_SPACE_DEBUG
7401 @hook TARGET_ADDR_SPACE_DIAGNOSE_USAGE
7404 @section Miscellaneous Parameters
7405 @cindex parameters, miscellaneous
7407 @c prevent bad page break with this line
7408 Here are several miscellaneous parameters.
7410 @defmac HAS_LONG_COND_BRANCH
7411 Define this boolean macro to indicate whether or not your architecture
7412 has conditional branches that can span all of memory. It is used in
7413 conjunction with an optimization that partitions hot and cold basic
7414 blocks into separate sections of the executable. If this macro is
7415 set to false, gcc will convert any conditional branches that attempt
7416 to cross between sections into unconditional branches or indirect jumps.
7419 @defmac HAS_LONG_UNCOND_BRANCH
7420 Define this boolean macro to indicate whether or not your architecture
7421 has unconditional branches that can span all of memory. It is used in
7422 conjunction with an optimization that partitions hot and cold basic
7423 blocks into separate sections of the executable. If this macro is
7424 set to false, gcc will convert any unconditional branches that attempt
7425 to cross between sections into indirect jumps.
7428 @defmac CASE_VECTOR_MODE
7429 An alias for a machine mode name. This is the machine mode that
7430 elements of a jump-table should have.
7433 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7434 Optional: return the preferred mode for an @code{addr_diff_vec}
7435 when the minimum and maximum offset are known. If you define this,
7436 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7437 To make this work, you also have to define @code{INSN_ALIGN} and
7438 make the alignment for @code{addr_diff_vec} explicit.
7439 The @var{body} argument is provided so that the offset_unsigned and scale
7440 flags can be updated.
7443 @defmac CASE_VECTOR_PC_RELATIVE
7444 Define this macro to be a C expression to indicate when jump-tables
7445 should contain relative addresses. You need not define this macro if
7446 jump-tables never contain relative addresses, or jump-tables should
7447 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
7451 @hook TARGET_CASE_VALUES_THRESHOLD
7453 @defmac WORD_REGISTER_OPERATIONS
7454 Define this macro to 1 if operations between registers with integral mode
7455 smaller than a word are always performed on the entire register. To be
7456 more explicit, if you start with a pair of @code{word_mode} registers with
7457 known values and you do a subword, for example @code{QImode}, addition on
7458 the low part of the registers, then the compiler may consider that the
7459 result has a known value in @code{word_mode} too if the macro is defined
7460 to 1. Most RISC machines have this property and most CISC machines do not.
7463 @hook TARGET_MIN_ARITHMETIC_PRECISION
7465 @defmac LOAD_EXTEND_OP (@var{mem_mode})
7466 Define this macro to be a C expression indicating when insns that read
7467 memory in @var{mem_mode}, an integral mode narrower than a word, set the
7468 bits outside of @var{mem_mode} to be either the sign-extension or the
7469 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7470 of @var{mem_mode} for which the
7471 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7472 @code{UNKNOWN} for other modes.
7474 This macro is not called with @var{mem_mode} non-integral or with a width
7475 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7476 value in this case. Do not define this macro if it would always return
7477 @code{UNKNOWN}. On machines where this macro is defined, you will normally
7478 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7480 You may return a non-@code{UNKNOWN} value even if for some hard registers
7481 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
7482 of these hard registers @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false
7483 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
7484 integral mode larger than this but not larger than @code{word_mode}.
7486 You must return @code{UNKNOWN} if for some hard registers that allow this
7487 mode, @code{TARGET_CAN_CHANGE_MODE_CLASS} says that they cannot change to
7488 @code{word_mode}, but that they can change to another integral mode that
7489 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
7492 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
7493 Define this macro to 1 if loading short immediate values into registers sign
7497 @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
7500 The maximum number of bytes that a single instruction can move quickly
7501 between memory and registers or between two memory locations.
7504 @defmac MAX_MOVE_MAX
7505 The maximum number of bytes that a single instruction can move quickly
7506 between memory and registers or between two memory locations. If this
7507 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7508 constant value that is the largest value that @code{MOVE_MAX} can have
7512 @defmac SHIFT_COUNT_TRUNCATED
7513 A C expression that is nonzero if on this machine the number of bits
7514 actually used for the count of a shift operation is equal to the number
7515 of bits needed to represent the size of the object being shifted. When
7516 this macro is nonzero, the compiler will assume that it is safe to omit
7517 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7518 truncates the count of a shift operation. On machines that have
7519 instructions that act on bit-fields at variable positions, which may
7520 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7521 also enables deletion of truncations of the values that serve as
7522 arguments to bit-field instructions.
7524 If both types of instructions truncate the count (for shifts) and
7525 position (for bit-field operations), or if no variable-position bit-field
7526 instructions exist, you should define this macro.
7528 However, on some machines, such as the 80386 and the 680x0, truncation
7529 only applies to shift operations and not the (real or pretended)
7530 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7531 such machines. Instead, add patterns to the @file{md} file that include
7532 the implied truncation of the shift instructions.
7534 You need not define this macro if it would always have the value of zero.
7537 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
7538 @hook TARGET_SHIFT_TRUNCATION_MASK
7540 @hook TARGET_TRULY_NOOP_TRUNCATION
7542 @hook TARGET_MODE_REP_EXTENDED
7544 @hook TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P
7546 @defmac STORE_FLAG_VALUE
7547 A C expression describing the value returned by a comparison operator
7548 with an integral mode and stored by a store-flag instruction
7549 (@samp{cstore@var{mode}4}) when the condition is true. This description must
7550 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
7551 comparison operators whose results have a @code{MODE_INT} mode.
7553 A value of 1 or @minus{}1 means that the instruction implementing the
7554 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
7555 and 0 when the comparison is false. Otherwise, the value indicates
7556 which bits of the result are guaranteed to be 1 when the comparison is
7557 true. This value is interpreted in the mode of the comparison
7558 operation, which is given by the mode of the first operand in the
7559 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
7560 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7563 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
7564 generate code that depends only on the specified bits. It can also
7565 replace comparison operators with equivalent operations if they cause
7566 the required bits to be set, even if the remaining bits are undefined.
7567 For example, on a machine whose comparison operators return an
7568 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7569 @samp{0x80000000}, saying that just the sign bit is relevant, the
7573 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7580 (ashift:SI @var{x} (const_int @var{n}))
7584 where @var{n} is the appropriate shift count to move the bit being
7585 tested into the sign bit.
7587 There is no way to describe a machine that always sets the low-order bit
7588 for a true value, but does not guarantee the value of any other bits,
7589 but we do not know of any machine that has such an instruction. If you
7590 are trying to port GCC to such a machine, include an instruction to
7591 perform a logical-and of the result with 1 in the pattern for the
7592 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
7594 Often, a machine will have multiple instructions that obtain a value
7595 from a comparison (or the condition codes). Here are rules to guide the
7596 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7601 Use the shortest sequence that yields a valid definition for
7602 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7603 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7604 comparison operators to do so because there may be opportunities to
7605 combine the normalization with other operations.
7608 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
7609 slightly preferred on machines with expensive jumps and 1 preferred on
7613 As a second choice, choose a value of @samp{0x80000001} if instructions
7614 exist that set both the sign and low-order bits but do not define the
7618 Otherwise, use a value of @samp{0x80000000}.
7621 Many machines can produce both the value chosen for
7622 @code{STORE_FLAG_VALUE} and its negation in the same number of
7623 instructions. On those machines, you should also define a pattern for
7624 those cases, e.g., one matching
7627 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7630 Some machines can also perform @code{and} or @code{plus} operations on
7631 condition code values with less instructions than the corresponding
7632 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
7633 machines, define the appropriate patterns. Use the names @code{incscc}
7634 and @code{decscc}, respectively, for the patterns which perform
7635 @code{plus} or @code{minus} operations on condition code values. See
7636 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
7637 find such instruction sequences on other machines.
7639 If this macro is not defined, the default value, 1, is used. You need
7640 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7641 instructions, or if the value generated by these instructions is 1.
7644 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
7645 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
7646 returned when comparison operators with floating-point results are true.
7647 Define this macro on machines that have comparison operations that return
7648 floating-point values. If there are no such operations, do not define
7652 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
7653 A C expression that gives a rtx representing the nonzero true element
7654 for vector comparisons. The returned rtx should be valid for the inner
7655 mode of @var{mode} which is guaranteed to be a vector mode. Define
7656 this macro on machines that have vector comparison operations that
7657 return a vector result. If there are no such operations, do not define
7658 this macro. Typically, this macro is defined as @code{const1_rtx} or
7659 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
7660 the compiler optimizing such vector comparison operations for the
7664 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7665 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7666 A C expression that indicates whether the architecture defines a value
7667 for @code{clz} or @code{ctz} with a zero operand.
7668 A result of @code{0} indicates the value is undefined.
7669 If the value is defined for only the RTL expression, the macro should
7670 evaluate to @code{1}; if the value applies also to the corresponding optab
7671 entry (which is normally the case if it expands directly into
7672 the corresponding RTL), then the macro should evaluate to @code{2}.
7673 In the cases where the value is defined, @var{value} should be set to
7676 If this macro is not defined, the value of @code{clz} or
7677 @code{ctz} at zero is assumed to be undefined.
7679 This macro must be defined if the target's expansion for @code{ffs}
7680 relies on a particular value to get correct results. Otherwise it
7681 is not necessary, though it may be used to optimize some corner cases, and
7682 to provide a default expansion for the @code{ffs} optab.
7684 Note that regardless of this macro the ``definedness'' of @code{clz}
7685 and @code{ctz} at zero do @emph{not} extend to the builtin functions
7686 visible to the user. Thus one may be free to adjust the value at will
7687 to match the target expansion of these operations without fear of
7692 An alias for the machine mode for pointers. On most machines, define
7693 this to be the integer mode corresponding to the width of a hardware
7694 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7695 On some machines you must define this to be one of the partial integer
7696 modes, such as @code{PSImode}.
7698 The width of @code{Pmode} must be at least as large as the value of
7699 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7700 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7704 @defmac FUNCTION_MODE
7705 An alias for the machine mode used for memory references to functions
7706 being called, in @code{call} RTL expressions. On most CISC machines,
7707 where an instruction can begin at any byte address, this should be
7708 @code{QImode}. On most RISC machines, where all instructions have fixed
7709 size and alignment, this should be a mode with the same size and alignment
7710 as the machine instruction words - typically @code{SImode} or @code{HImode}.
7713 @defmac STDC_0_IN_SYSTEM_HEADERS
7714 In normal operation, the preprocessor expands @code{__STDC__} to the
7715 constant 1, to signify that GCC conforms to ISO Standard C@. On some
7716 hosts, like Solaris, the system compiler uses a different convention,
7717 where @code{__STDC__} is normally 0, but is 1 if the user specifies
7718 strict conformance to the C Standard.
7720 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
7721 convention when processing system header files, but when processing user
7722 files @code{__STDC__} will always expand to 1.
7725 @hook TARGET_C_PREINCLUDE
7727 @hook TARGET_CXX_IMPLICIT_EXTERN_C
7729 @defmac SYSTEM_IMPLICIT_EXTERN_C
7730 Define this macro if the system header files do not support C++@.
7731 This macro handles system header files by pretending that system
7732 header files are enclosed in @samp{extern "C" @{@dots{}@}}.
7737 @defmac REGISTER_TARGET_PRAGMAS ()
7738 Define this macro if you want to implement any target-specific pragmas.
7739 If defined, it is a C expression which makes a series of calls to
7740 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
7741 for each pragma. The macro may also do any
7742 setup required for the pragmas.
7744 The primary reason to define this macro is to provide compatibility with
7745 other compilers for the same target. In general, we discourage
7746 definition of target-specific pragmas for GCC@.
7748 If the pragma can be implemented by attributes then you should consider
7749 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
7751 Preprocessor macros that appear on pragma lines are not expanded. All
7752 @samp{#pragma} directives that do not match any registered pragma are
7753 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
7756 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7757 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7759 Each call to @code{c_register_pragma} or
7760 @code{c_register_pragma_with_expansion} establishes one pragma. The
7761 @var{callback} routine will be called when the preprocessor encounters a
7765 #pragma [@var{space}] @var{name} @dots{}
7768 @var{space} is the case-sensitive namespace of the pragma, or
7769 @code{NULL} to put the pragma in the global namespace. The callback
7770 routine receives @var{pfile} as its first argument, which can be passed
7771 on to cpplib's functions if necessary. You can lex tokens after the
7772 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
7773 callback will be silently ignored. The end of the line is indicated by
7774 a token of type @code{CPP_EOF}. Macro expansion occurs on the
7775 arguments of pragmas registered with
7776 @code{c_register_pragma_with_expansion} but not on the arguments of
7777 pragmas registered with @code{c_register_pragma}.
7779 Note that the use of @code{pragma_lex} is specific to the C and C++
7780 compilers. It will not work in the Java or Fortran compilers, or any
7781 other language compilers for that matter. Thus if @code{pragma_lex} is going
7782 to be called from target-specific code, it must only be done so when
7783 building the C and C++ compilers. This can be done by defining the
7784 variables @code{c_target_objs} and @code{cxx_target_objs} in the
7785 target entry in the @file{config.gcc} file. These variables should name
7786 the target-specific, language-specific object file which contains the
7787 code that uses @code{pragma_lex}. Note it will also be necessary to add a
7788 rule to the makefile fragment pointed to by @code{tmake_file} that shows
7789 how to build this object file.
7792 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
7793 Define this macro if macros should be expanded in the
7794 arguments of @samp{#pragma pack}.
7797 @defmac TARGET_DEFAULT_PACK_STRUCT
7798 If your target requires a structure packing default other than 0 (meaning
7799 the machine default), define this macro to the necessary value (in bytes).
7800 This must be a value that would also be valid to use with
7801 @samp{#pragma pack()} (that is, a small power of two).
7804 @defmac DOLLARS_IN_IDENTIFIERS
7805 Define this macro to control use of the character @samp{$} in
7806 identifier names for the C family of languages. 0 means @samp{$} is
7807 not allowed by default; 1 means it is allowed. 1 is the default;
7808 there is no need to define this macro in that case.
7811 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
7812 Define this macro as a C expression that is nonzero if it is safe for the
7813 delay slot scheduler to place instructions in the delay slot of @var{insn},
7814 even if they appear to use a resource set or clobbered in @var{insn}.
7815 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
7816 every @code{call_insn} has this behavior. On machines where some @code{insn}
7817 or @code{jump_insn} is really a function call and hence has this behavior,
7818 you should define this macro.
7820 You need not define this macro if it would always return zero.
7823 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
7824 Define this macro as a C expression that is nonzero if it is safe for the
7825 delay slot scheduler to place instructions in the delay slot of @var{insn},
7826 even if they appear to set or clobber a resource referenced in @var{insn}.
7827 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
7828 some @code{insn} or @code{jump_insn} is really a function call and its operands
7829 are registers whose use is actually in the subroutine it calls, you should
7830 define this macro. Doing so allows the delay slot scheduler to move
7831 instructions which copy arguments into the argument registers into the delay
7834 You need not define this macro if it would always return zero.
7837 @defmac MULTIPLE_SYMBOL_SPACES
7838 Define this macro as a C expression that is nonzero if, in some cases,
7839 global symbols from one translation unit may not be bound to undefined
7840 symbols in another translation unit without user intervention. For
7841 instance, under Microsoft Windows symbols must be explicitly imported
7842 from shared libraries (DLLs).
7844 You need not define this macro if it would always evaluate to zero.
7847 @hook TARGET_MD_ASM_ADJUST
7849 @defmac MATH_LIBRARY
7850 Define this macro as a C string constant for the linker argument to link
7851 in the system math library, minus the initial @samp{"-l"}, or
7852 @samp{""} if the target does not have a
7853 separate math library.
7855 You need only define this macro if the default of @samp{"m"} is wrong.
7858 @defmac LIBRARY_PATH_ENV
7859 Define this macro as a C string constant for the environment variable that
7860 specifies where the linker should look for libraries.
7862 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
7866 @defmac TARGET_POSIX_IO
7867 Define this macro if the target supports the following POSIX@ file
7868 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
7869 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
7870 to use file locking when exiting a program, which avoids race conditions
7871 if the program has forked. It will also create directories at run-time
7872 for cross-profiling.
7875 @defmac MAX_CONDITIONAL_EXECUTE
7877 A C expression for the maximum number of instructions to execute via
7878 conditional execution instructions instead of a branch. A value of
7879 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
7880 1 if it does use cc0.
7883 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
7884 Used if the target needs to perform machine-dependent modifications on the
7885 conditionals used for turning basic blocks into conditionally executed code.
7886 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
7887 contains information about the currently processed blocks. @var{true_expr}
7888 and @var{false_expr} are the tests that are used for converting the
7889 then-block and the else-block, respectively. Set either @var{true_expr} or
7890 @var{false_expr} to a null pointer if the tests cannot be converted.
7893 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
7894 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
7895 if-statements into conditions combined by @code{and} and @code{or} operations.
7896 @var{bb} contains the basic block that contains the test that is currently
7897 being processed and about to be turned into a condition.
7900 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
7901 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
7902 be converted to conditional execution format. @var{ce_info} points to
7903 a data structure, @code{struct ce_if_block}, which contains information
7904 about the currently processed blocks.
7907 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
7908 A C expression to perform any final machine dependent modifications in
7909 converting code to conditional execution. The involved basic blocks
7910 can be found in the @code{struct ce_if_block} structure that is pointed
7911 to by @var{ce_info}.
7914 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
7915 A C expression to cancel any machine dependent modifications in
7916 converting code to conditional execution. The involved basic blocks
7917 can be found in the @code{struct ce_if_block} structure that is pointed
7918 to by @var{ce_info}.
7921 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
7922 A C expression to initialize any machine specific data for if-conversion
7923 of the if-block in the @code{struct ce_if_block} structure that is pointed
7924 to by @var{ce_info}.
7927 @hook TARGET_MACHINE_DEPENDENT_REORG
7929 @hook TARGET_INIT_BUILTINS
7931 @hook TARGET_BUILTIN_DECL
7933 @hook TARGET_EXPAND_BUILTIN
7935 @hook TARGET_RESOLVE_OVERLOADED_BUILTIN
7937 @hook TARGET_FOLD_BUILTIN
7939 @hook TARGET_GIMPLE_FOLD_BUILTIN
7941 @hook TARGET_COMPARE_VERSION_PRIORITY
7943 @hook TARGET_GET_FUNCTION_VERSIONS_DISPATCHER
7945 @hook TARGET_GENERATE_VERSION_DISPATCHER_BODY
7947 @hook TARGET_PREDICT_DOLOOP_P
7949 @hook TARGET_CAN_USE_DOLOOP_P
7951 @hook TARGET_INVALID_WITHIN_DOLOOP
7953 @hook TARGET_LEGITIMATE_COMBINED_INSN
7955 @hook TARGET_CAN_FOLLOW_JUMP
7957 @hook TARGET_COMMUTATIVE_P
7959 @hook TARGET_ALLOCATE_INITIAL_VALUE
7961 @hook TARGET_UNSPEC_MAY_TRAP_P
7963 @hook TARGET_SET_CURRENT_FUNCTION
7965 @defmac TARGET_OBJECT_SUFFIX
7966 Define this macro to be a C string representing the suffix for object
7967 files on your target machine. If you do not define this macro, GCC will
7968 use @samp{.o} as the suffix for object files.
7971 @defmac TARGET_EXECUTABLE_SUFFIX
7972 Define this macro to be a C string representing the suffix to be
7973 automatically added to executable files on your target machine. If you
7974 do not define this macro, GCC will use the null string as the suffix for
7978 @defmac COLLECT_EXPORT_LIST
7979 If defined, @code{collect2} will scan the individual object files
7980 specified on its command line and create an export list for the linker.
7981 Define this macro for systems like AIX, where the linker discards
7982 object files that are not referenced from @code{main} and uses export
7986 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
7987 Define this macro to a C expression representing a variant of the
7988 method call @var{mdecl}, if Java Native Interface (JNI) methods
7989 must be invoked differently from other methods on your target.
7990 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
7991 the @code{stdcall} calling convention and this macro is then
7992 defined as this expression:
7995 build_type_attribute_variant (@var{mdecl},
7997 (get_identifier ("stdcall"),
8002 @hook TARGET_CANNOT_MODIFY_JUMPS_P
8004 @hook TARGET_BRANCH_TARGET_REGISTER_CLASS
8006 @hook TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED
8008 @hook TARGET_HAVE_CONDITIONAL_EXECUTION
8010 @hook TARGET_GEN_CCMP_FIRST
8012 @hook TARGET_GEN_CCMP_NEXT
8014 @hook TARGET_LOOP_UNROLL_ADJUST
8016 @defmac POWI_MAX_MULTS
8017 If defined, this macro is interpreted as a signed integer C expression
8018 that specifies the maximum number of floating point multiplications
8019 that should be emitted when expanding exponentiation by an integer
8020 constant inline. When this value is defined, exponentiation requiring
8021 more than this number of multiplications is implemented by calling the
8022 system library's @code{pow}, @code{powf} or @code{powl} routines.
8023 The default value places no upper bound on the multiplication count.
8026 @deftypefn Macro void TARGET_EXTRA_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. The parameter @var{stdinc} indicates if normal include files
8029 are present. The parameter @var{sysroot} is the system root directory.
8030 The parameter @var{iprefix} is the prefix for the gcc directory.
8033 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8034 This target hook should register any extra include files for the
8035 target before any standard headers. The parameter @var{stdinc}
8036 indicates if normal include files are present. The parameter
8037 @var{sysroot} is the system root directory. The parameter
8038 @var{iprefix} is the prefix for the gcc directory.
8041 @deftypefn Macro void TARGET_OPTF (char *@var{path})
8042 This target hook should register special include paths for the target.
8043 The parameter @var{path} is the include to register. On Darwin
8044 systems, this is used for Framework includes, which have semantics
8045 that are different from @option{-I}.
8048 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
8049 This target macro returns @code{true} if it is safe to use a local alias
8050 for a virtual function @var{fndecl} when constructing thunks,
8051 @code{false} otherwise. By default, the macro returns @code{true} for all
8052 functions, if a target supports aliases (i.e.@: defines
8053 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
8056 @defmac TARGET_FORMAT_TYPES
8057 If defined, this macro is the name of a global variable containing
8058 target-specific format checking information for the @option{-Wformat}
8059 option. The default is to have no target-specific format checks.
8062 @defmac TARGET_N_FORMAT_TYPES
8063 If defined, this macro is the number of entries in
8064 @code{TARGET_FORMAT_TYPES}.
8067 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
8068 If defined, this macro is the name of a global variable containing
8069 target-specific format overrides for the @option{-Wformat} option. The
8070 default is to have no target-specific format overrides. If defined,
8071 @code{TARGET_FORMAT_TYPES} must be defined, too.
8074 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
8075 If defined, this macro specifies the number of entries in
8076 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
8079 @defmac TARGET_OVERRIDES_FORMAT_INIT
8080 If defined, this macro specifies the optional initialization
8081 routine for target specific customizations of the system printf
8082 and scanf formatter settings.
8085 @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
8087 @hook TARGET_INVALID_CONVERSION
8089 @hook TARGET_INVALID_UNARY_OP
8091 @hook TARGET_INVALID_BINARY_OP
8093 @hook TARGET_PROMOTED_TYPE
8095 @hook TARGET_CONVERT_TO_TYPE
8098 This macro determines the size of the objective C jump buffer for the
8099 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
8102 @defmac LIBGCC2_UNWIND_ATTRIBUTE
8103 Define this macro if any target-specific attributes need to be attached
8104 to the functions in @file{libgcc} that provide low-level support for
8105 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
8106 and the associated definitions of those functions.
8109 @hook TARGET_UPDATE_STACK_BOUNDARY
8111 @hook TARGET_GET_DRAP_RTX
8113 @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
8115 @hook TARGET_CONST_ANCHOR
8117 @hook TARGET_ASAN_SHADOW_OFFSET
8119 @hook TARGET_MEMMODEL_CHECK
8121 @hook TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
8123 @hook TARGET_HAS_IFUNC_P
8125 @hook TARGET_ATOMIC_ALIGN_FOR_MODE
8127 @hook TARGET_ATOMIC_ASSIGN_EXPAND_FENV
8129 @hook TARGET_RECORD_OFFLOAD_SYMBOL
8131 @hook TARGET_OFFLOAD_OPTIONS
8133 @defmac TARGET_SUPPORTS_WIDE_INT
8135 On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
8136 objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
8137 to indicate that large integers are stored in
8138 @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows
8139 very large integer constants to be represented. @code{CONST_DOUBLE}
8140 is limited to twice the size of the host's @code{HOST_WIDE_INT}
8143 Converting a port mostly requires looking for the places where
8144 @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
8145 code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i
8146 const_double"} at the port level gets you to 95% of the changes that
8147 need to be made. There are a few places that require a deeper look.
8151 There is no equivalent to @code{hval} and @code{lval} for
8152 @code{CONST_WIDE_INT}s. This would be difficult to express in the md
8153 language since there are a variable number of elements.
8155 Most ports only check that @code{hval} is either 0 or -1 to see if the
8156 value is small. As mentioned above, this will no longer be necessary
8157 since small constants are always @code{CONST_INT}. Of course there
8158 are still a few exceptions, the alpha's constraint used by the zap
8159 instruction certainly requires careful examination by C code.
8160 However, all the current code does is pass the hval and lval to C
8161 code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
8162 not really a large change.
8165 Because there is no standard template that ports use to materialize
8166 constants, there is likely to be some futzing that is unique to each
8170 The rtx costs may have to be adjusted to properly account for larger
8171 constants that are represented as @code{CONST_WIDE_INT}.
8174 All and all it does not take long to convert ports that the
8175 maintainer is familiar with.
8179 @hook TARGET_HAVE_SPECULATION_SAFE_VALUE
8181 @hook TARGET_SPECULATION_SAFE_VALUE
8183 @hook TARGET_RUN_TARGET_SELFTESTS