1 @c Copyright (C) 1988-2023 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 * Rust Language and ABI:: Controlling Rust ABI changes.
57 * Named Address Spaces:: Adding support for named address spaces
58 * Misc:: Everything else.
61 @node Target Structure
62 @section The Global @code{targetm} Variable
64 @cindex target functions
66 @deftypevar {struct gcc_target} targetm
67 The target @file{.c} file must define the global @code{targetm} variable
68 which contains pointers to functions and data relating to the target
69 machine. The variable is declared in @file{target.h};
70 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
71 used to initialize the variable, and macros for the default initializers
72 for elements of the structure. The @file{.c} file should override those
73 macros for which the default definition is inappropriate. For example:
76 #include "target-def.h"
78 /* @r{Initialize the GCC target structure.} */
80 #undef TARGET_COMP_TYPE_ATTRIBUTES
81 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
83 struct gcc_target targetm = TARGET_INITIALIZER;
87 Where a macro should be defined in the @file{.c} file in this manner to
88 form part of the @code{targetm} structure, it is documented below as a
89 ``Target Hook'' with a prototype. Many macros will change in future
90 from being defined in the @file{.h} file to being part of the
91 @code{targetm} structure.
93 Similarly, there is a @code{targetcm} variable for hooks that are
94 specific to front ends for C-family languages, documented as ``C
95 Target Hook''. This is declared in @file{c-family/c-target.h}, the
96 initializer @code{TARGETCM_INITIALIZER} in
97 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
98 themselves, they should set @code{target_has_targetcm=yes} in
99 @file{config.gcc}; otherwise a default definition is used.
101 Similarly, there is a @code{targetm_common} variable for hooks that
102 are shared between the compiler driver and the compilers proper,
103 documented as ``Common Target Hook''. This is declared in
104 @file{common/common-target.h}, the initializer
105 @code{TARGETM_COMMON_INITIALIZER} in
106 @file{common/common-target-def.h}. If targets initialize
107 @code{targetm_common} themselves, they should set
108 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
109 default definition is used.
111 Similarly, there is a @code{targetdm} variable for hooks that are
112 specific to the D language front end, documented as ``D Target Hook''.
113 This is declared in @file{d/d-target.h}, the initializer
114 @code{TARGETDM_INITIALIZER} in @file{d/d-target-def.h}. If targets
115 initialize @code{targetdm} themselves, they should set
116 @code{target_has_targetdm=yes} in @file{config.gcc}; otherwise a default
119 Similarly, there is a @code{targetrustm} variable for hooks that are
120 specific to the Rust language front end, documented as ``Rust Target
121 Hook''. This is declared in @file{rust/rust-target.h}, the initializer
122 @code{TARGETRUSTM_INITIALIZER} in @file{rust/rust-target-def.h}.
123 If targets initialize @code{targetrustm} themselves, they should set
124 @code{target_has_targetrustm=yes} in @file{config.gcc}; otherwise a
125 default definition is used.
128 @section Controlling the Compilation Driver, @file{gcc}
130 @cindex controlling the compilation driver
132 @c prevent bad page break with this line
133 You can control the compilation driver.
135 @defmac DRIVER_SELF_SPECS
136 A list of specs for the driver itself. It should be a suitable
137 initializer for an array of strings, with no surrounding braces.
139 The driver applies these specs to its own command line between loading
140 default @file{specs} files (but not command-line specified ones) and
141 choosing the multilib directory or running any subcommands. It
142 applies them in the order given, so each spec can depend on the
143 options added by earlier ones. It is also possible to remove options
144 using @samp{%<@var{option}} in the usual way.
146 This macro can be useful when a port has several interdependent target
147 options. It provides a way of standardizing the command line so
148 that the other specs are easier to write.
150 Do not define this macro if it does not need to do anything.
153 @defmac OPTION_DEFAULT_SPECS
154 A list of specs used to support configure-time default options (i.e.@:
155 @option{--with} options) in the driver. It should be a suitable initializer
156 for an array of structures, each containing two strings, without the
157 outermost pair of surrounding braces.
159 The first item in the pair is the name of the default. This must match
160 the code in @file{config.gcc} for the target. The second item is a spec
161 to apply if a default with this name was specified. The string
162 @samp{%(VALUE)} in the spec will be replaced by the value of the default
163 everywhere it occurs.
165 The driver will apply these specs to its own command line between loading
166 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
167 the same mechanism as @code{DRIVER_SELF_SPECS}.
169 Do not define this macro if it does not need to do anything.
173 A C string constant that tells the GCC driver program options to
174 pass to CPP@. It can also specify how to translate options you
175 give to GCC into options for GCC to pass to the CPP@.
177 Do not define this macro if it does not need to do anything.
180 @defmac CPLUSPLUS_CPP_SPEC
181 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
182 than C@. If you do not define this macro, then the value of
183 @code{CPP_SPEC} (if any) will be used instead.
187 A C string constant that tells the GCC driver program options to
188 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
190 It can also specify how to translate options you give to GCC into options
191 for GCC to pass to front ends.
193 Do not define this macro if it does not need to do anything.
197 A C string constant that tells the GCC driver program options to
198 pass to @code{cc1plus}. It can also specify how to translate options you
199 give to GCC into options for GCC to pass to the @code{cc1plus}.
201 Do not define this macro if it does not need to do anything.
202 Note that everything defined in CC1_SPEC is already passed to
203 @code{cc1plus} so there is no need to duplicate the contents of
204 CC1_SPEC in CC1PLUS_SPEC@.
208 A C string constant that tells the GCC driver program options to
209 pass to the assembler. It can also specify how to translate options
210 you give to GCC into options for GCC to pass to the assembler.
211 See the file @file{sun3.h} for an example of this.
213 Do not define this macro if it does not need to do anything.
216 @defmac ASM_FINAL_SPEC
217 A C string constant that tells the GCC driver program how to
218 run any programs which cleanup after the normal assembler.
219 Normally, this is not needed. See the file @file{mips.h} for
222 Do not define this macro if it does not need to do anything.
225 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
226 Define this macro, with no value, if the driver should give the assembler
227 an argument consisting of a single dash, @option{-}, to instruct it to
228 read from its standard input (which will be a pipe connected to the
229 output of the compiler proper). This argument is given after any
230 @option{-o} option specifying the name of the output file.
232 If you do not define this macro, the assembler is assumed to read its
233 standard input if given no non-option arguments. If your assembler
234 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
235 see @file{mips.h} for instance.
239 A C string constant that tells the GCC driver program options to
240 pass to the linker. It can also specify how to translate options you
241 give to GCC into options for GCC to pass to the linker.
243 Do not define this macro if it does not need to do anything.
247 Another C string constant used much like @code{LINK_SPEC}. The difference
248 between the two is that @code{LIB_SPEC} is used at the end of the
249 command given to the linker.
251 If this macro is not defined, a default is provided that
252 loads the standard C library from the usual place. See @file{gcc.cc}.
256 Another C string constant that tells the GCC driver program
257 how and when to place a reference to @file{libgcc.a} into the
258 linker command line. This constant is placed both before and after
259 the value of @code{LIB_SPEC}.
261 If this macro is not defined, the GCC driver provides a default that
262 passes the string @option{-lgcc} to the linker.
265 @defmac REAL_LIBGCC_SPEC
266 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
267 @code{LIBGCC_SPEC} is not directly used by the driver program but is
268 instead modified to refer to different versions of @file{libgcc.a}
269 depending on the values of the command line flags @option{-static},
270 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
271 targets where these modifications are inappropriate, define
272 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
273 driver how to place a reference to @file{libgcc} on the link command
274 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
277 @defmac USE_LD_AS_NEEDED
278 A macro that controls the modifications to @code{LIBGCC_SPEC}
279 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
280 generated that uses @option{--as-needed} or equivalent options and the
281 shared @file{libgcc} in place of the
282 static exception handler library, when linking without any of
283 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
287 If defined, this C string constant is added to @code{LINK_SPEC}.
288 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
289 the modifications to @code{LIBGCC_SPEC} mentioned in
290 @code{REAL_LIBGCC_SPEC}.
293 @defmac STARTFILE_SPEC
294 Another C string constant used much like @code{LINK_SPEC}. The
295 difference between the two is that @code{STARTFILE_SPEC} is used at
296 the very beginning of the command given to the linker.
298 If this macro is not defined, a default is provided that loads the
299 standard C startup file from the usual place. See @file{gcc.cc}.
303 Another C string constant used much like @code{LINK_SPEC}. The
304 difference between the two is that @code{ENDFILE_SPEC} is used at
305 the very end of the command given to the linker.
307 Do not define this macro if it does not need to do anything.
310 @defmac THREAD_MODEL_SPEC
311 GCC @code{-v} will print the thread model GCC was configured to use.
312 However, this doesn't work on platforms that are multilibbed on thread
313 models, such as AIX 4.3. On such platforms, define
314 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
315 blanks that names one of the recognized thread models. @code{%*}, the
316 default value of this macro, will expand to the value of
317 @code{thread_file} set in @file{config.gcc}.
320 @defmac SYSROOT_SUFFIX_SPEC
321 Define this macro to add a suffix to the target sysroot when GCC is
322 configured with a sysroot. This will cause GCC to search for usr/lib,
323 et al, within sysroot+suffix.
326 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
327 Define this macro to add a headers_suffix to the target sysroot when
328 GCC is configured with a sysroot. This will cause GCC to pass the
329 updated sysroot+headers_suffix to CPP, causing it to search for
330 usr/include, et al, within sysroot+headers_suffix.
334 Define this macro to provide additional specifications to put in the
335 @file{specs} file that can be used in various specifications like
338 The definition should be an initializer for an array of structures,
339 containing a string constant, that defines the specification name, and a
340 string constant that provides the specification.
342 Do not define this macro if it does not need to do anything.
344 @code{EXTRA_SPECS} is useful when an architecture contains several
345 related targets, which have various @code{@dots{}_SPECS} which are similar
346 to each other, and the maintainer would like one central place to keep
349 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
350 define either @code{_CALL_SYSV} when the System V calling sequence is
351 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
354 The @file{config/rs6000/rs6000.h} target file defines:
357 #define EXTRA_SPECS \
358 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
360 #define CPP_SYS_DEFAULT ""
363 The @file{config/rs6000/sysv.h} target file defines:
367 "%@{posix: -D_POSIX_SOURCE @} \
368 %@{mcall-sysv: -D_CALL_SYSV @} \
369 %@{!mcall-sysv: %(cpp_sysv_default) @} \
370 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
372 #undef CPP_SYSV_DEFAULT
373 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
376 while the @file{config/rs6000/eabiaix.h} target file defines
377 @code{CPP_SYSV_DEFAULT} as:
380 #undef CPP_SYSV_DEFAULT
381 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
385 @defmac LINK_LIBGCC_SPECIAL_1
386 Define this macro if the driver program should find the library
387 @file{libgcc.a}. If you do not define this macro, the driver program will pass
388 the argument @option{-lgcc} to tell the linker to do the search.
391 @defmac LINK_GCC_C_SEQUENCE_SPEC
392 The sequence in which libgcc and libc are specified to the linker.
393 By default this is @code{%G %L %G}.
396 @defmac POST_LINK_SPEC
397 Define this macro to add additional steps to be executed after linker.
398 The default value of this macro is empty string.
401 @defmac LINK_COMMAND_SPEC
402 A C string constant giving the complete command line need to execute the
403 linker. When you do this, you will need to update your port each time a
404 change is made to the link command line within @file{gcc.cc}. Therefore,
405 define this macro only if you need to completely redefine the command
406 line for invoking the linker and there is no other way to accomplish
407 the effect you need. Overriding this macro may be avoidable by overriding
408 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
411 @hook TARGET_ALWAYS_STRIP_DOTDOT
413 @defmac MULTILIB_DEFAULTS
414 Define this macro as a C expression for the initializer of an array of
415 string to tell the driver program which options are defaults for this
416 target and thus do not need to be handled specially when using
417 @code{MULTILIB_OPTIONS}.
419 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
420 the target makefile fragment or if none of the options listed in
421 @code{MULTILIB_OPTIONS} are set by default.
422 @xref{Target Fragment}.
425 @defmac RELATIVE_PREFIX_NOT_LINKDIR
426 Define this macro to tell @command{gcc} that it should only translate
427 a @option{-B} prefix into a @option{-L} linker option if the prefix
428 indicates an absolute file name.
431 @defmac MD_EXEC_PREFIX
432 If defined, this macro is an additional prefix to try after
433 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
434 when the compiler is built as a cross
435 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
436 to the list of directories used to find the assembler in @file{configure.ac}.
439 @defmac STANDARD_STARTFILE_PREFIX
440 Define this macro as a C string constant if you wish to override the
441 standard choice of @code{libdir} as the default prefix to
442 try when searching for startup files such as @file{crt0.o}.
443 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
444 is built as a cross compiler.
447 @defmac STANDARD_STARTFILE_PREFIX_1
448 Define this macro as a C string constant if you wish to override the
449 standard choice of @code{/lib} as a prefix to try after the default prefix
450 when searching for startup files such as @file{crt0.o}.
451 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
452 is built as a cross compiler.
455 @defmac STANDARD_STARTFILE_PREFIX_2
456 Define this macro as a C string constant if you wish to override the
457 standard choice of @code{/lib} as yet another prefix to try after the
458 default prefix when searching for startup files such as @file{crt0.o}.
459 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
460 is built as a cross compiler.
463 @defmac MD_STARTFILE_PREFIX
464 If defined, this macro supplies an additional prefix to try after the
465 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
466 compiler is built as a cross compiler.
469 @defmac MD_STARTFILE_PREFIX_1
470 If defined, this macro supplies yet another prefix to try after the
471 standard prefixes. It is not searched when the compiler is built as a
475 @defmac INIT_ENVIRONMENT
476 Define this macro as a C string constant if you wish to set environment
477 variables for programs called by the driver, such as the assembler and
478 loader. The driver passes the value of this macro to @code{putenv} to
479 initialize the necessary environment variables.
482 @defmac LOCAL_INCLUDE_DIR
483 Define this macro as a C string constant if you wish to override the
484 standard choice of @file{/usr/local/include} as the default prefix to
485 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
486 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
487 @file{config.gcc}, normally @file{/usr/include}) in the search order.
489 Cross compilers do not search either @file{/usr/local/include} or its
493 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
494 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
495 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
496 If you do not define this macro, no component is used.
499 @defmac INCLUDE_DEFAULTS
500 Define this macro if you wish to override the entire default search path
501 for include files. For a native compiler, the default search path
502 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
503 @code{GPLUSPLUS_INCLUDE_DIR}, and
504 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
505 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
506 and specify private search areas for GCC@. The directory
507 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
509 The definition should be an initializer for an array of structures.
510 Each array element should have four elements: the directory name (a
511 string constant), the component name (also a string constant), a flag
512 for C++-only directories,
513 and a flag showing that the includes in the directory don't need to be
514 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
515 the array with a null element.
517 The component name denotes what GNU package the include file is part of,
518 if any, in all uppercase letters. For example, it might be @samp{GCC}
519 or @samp{BINUTILS}. If the package is part of a vendor-supplied
520 operating system, code the component name as @samp{0}.
522 For example, here is the definition used for VAX/VMS:
525 #define INCLUDE_DEFAULTS \
527 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
528 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
529 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
536 Here is the order of prefixes tried for exec files:
540 Any prefixes specified by the user with @option{-B}.
543 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
544 is not set and the compiler has not been installed in the configure-time
545 @var{prefix}, the location in which the compiler has actually been installed.
548 The directories specified by the environment variable @code{COMPILER_PATH}.
551 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
552 in the configured-time @var{prefix}.
555 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
558 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
561 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
565 Here is the order of prefixes tried for startfiles:
569 Any prefixes specified by the user with @option{-B}.
572 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
573 value based on the installed toolchain location.
576 The directories specified by the environment variable @code{LIBRARY_PATH}
577 (or port-specific name; native only, cross compilers do not use this).
580 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
581 in the configured @var{prefix} or this is a native compiler.
584 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
587 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
591 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
592 native compiler, or we have a target system root.
595 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
596 native compiler, or we have a target system root.
599 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
600 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
601 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
604 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
605 compiler, or we have a target system root. The default for this macro is
609 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
610 compiler, or we have a target system root. The default for this macro is
614 @node Run-time Target
615 @section Run-time Target Specification
616 @cindex run-time target specification
617 @cindex predefined macros
618 @cindex target specifications
620 @c prevent bad page break with this line
621 Here are run-time target specifications.
623 @defmac TARGET_CPU_CPP_BUILTINS ()
624 This function-like macro expands to a block of code that defines
625 built-in preprocessor macros and assertions for the target CPU, using
626 the functions @code{builtin_define}, @code{builtin_define_std} and
627 @code{builtin_assert}. When the front end
628 calls this macro it provides a trailing semicolon, and since it has
629 finished command line option processing your code can use those
632 @code{builtin_assert} takes a string in the form you pass to the
633 command-line option @option{-A}, such as @code{cpu=mips}, and creates
634 the assertion. @code{builtin_define} takes a string in the form
635 accepted by option @option{-D} and unconditionally defines the macro.
637 @code{builtin_define_std} takes a string representing the name of an
638 object-like macro. If it doesn't lie in the user's namespace,
639 @code{builtin_define_std} defines it unconditionally. Otherwise, it
640 defines a version with two leading underscores, and another version
641 with two leading and trailing underscores, and defines the original
642 only if an ISO standard was not requested on the command line. For
643 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
644 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
645 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
646 defines only @code{_ABI64}.
648 You can also test for the C dialect being compiled. The variable
649 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
650 or @code{clk_objective_c}. Note that if we are preprocessing
651 assembler, this variable will be @code{clk_c} but the function-like
652 macro @code{preprocessing_asm_p()} will return true, so you might want
653 to check for that first. If you need to check for strict ANSI, the
654 variable @code{flag_iso} can be used. The function-like macro
655 @code{preprocessing_trad_p()} can be used to check for traditional
659 @defmac TARGET_OS_CPP_BUILTINS ()
660 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
661 and is used for the target operating system instead.
664 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
665 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
666 and is used for the target object format. @file{elfos.h} uses this
667 macro to define @code{__ELF__}, so you probably do not need to define
671 @deftypevar {extern int} target_flags
672 This variable is declared in @file{options.h}, which is included before
673 any target-specific headers.
676 @hook TARGET_DEFAULT_TARGET_FLAGS
677 This variable specifies the initial value of @code{target_flags}.
678 Its default setting is 0.
681 @cindex optional hardware or system features
682 @cindex features, optional, in system conventions
684 @hook TARGET_HANDLE_OPTION
685 This hook is called whenever the user specifies one of the
686 target-specific options described by the @file{.opt} definition files
687 (@pxref{Options}). It has the opportunity to do some option-specific
688 processing and should return true if the option is valid. The default
689 definition does nothing but return true.
691 @var{decoded} specifies the option and its arguments. @var{opts} and
692 @var{opts_set} are the @code{gcc_options} structures to be used for
693 storing option state, and @var{loc} is the location at which the
694 option was passed (@code{UNKNOWN_LOCATION} except for options passed
698 @hook TARGET_HANDLE_C_OPTION
699 This target hook is called whenever the user specifies one of the
700 target-specific C language family options described by the @file{.opt}
701 definition files(@pxref{Options}). It has the opportunity to do some
702 option-specific processing and should return true if the option is
703 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
704 default definition does nothing but return false.
706 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
707 options. However, if processing an option requires routines that are
708 only available in the C (and related language) front ends, then you
709 should use @code{TARGET_HANDLE_C_OPTION} instead.
712 @hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT
714 @hook TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE
716 @hook TARGET_OBJC_DECLARE_CLASS_DEFINITION
718 @hook TARGET_STRING_OBJECT_REF_TYPE_P
720 @hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG
722 @hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE
724 @defmac C_COMMON_OVERRIDE_OPTIONS
725 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
726 but is only used in the C
727 language frontends (C, Objective-C, C++, Objective-C++) and so can be
728 used to alter option flag variables which only exist in those
732 @hook TARGET_OPTION_OPTIMIZATION_TABLE
733 Some machines may desire to change what optimizations are performed for
734 various optimization levels. This variable, if defined, describes
735 options to enable at particular sets of optimization levels. These
736 options are processed once
737 just after the optimization level is determined and before the remainder
738 of the command options have been parsed, so may be overridden by other
739 options passed explicitly.
741 This processing is run once at program startup and when the optimization
742 options are changed via @code{#pragma GCC optimize} or by using the
743 @code{optimize} attribute.
746 @hook TARGET_OPTION_INIT_STRUCT
748 @hook TARGET_COMPUTE_MULTILIB
751 @defmac SWITCHABLE_TARGET
752 Some targets need to switch between substantially different subtargets
753 during compilation. For example, the MIPS target has one subtarget for
754 the traditional MIPS architecture and another for MIPS16. Source code
755 can switch between these two subarchitectures using the @code{mips16}
756 and @code{nomips16} attributes.
758 Such subtargets can differ in things like the set of available
759 registers, the set of available instructions, the costs of various
760 operations, and so on. GCC caches a lot of this type of information
761 in global variables, and recomputing them for each subtarget takes a
762 significant amount of time. The compiler therefore provides a facility
763 for maintaining several versions of the global variables and quickly
764 switching between them; see @file{target-globals.h} for details.
766 Define this macro to 1 if your target needs this facility. The default
770 @hook TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P
772 @node Per-Function Data
773 @section Defining data structures for per-function information.
774 @cindex per-function data
775 @cindex data structures
777 If the target needs to store information on a per-function basis, GCC
778 provides a macro and a couple of variables to allow this. Note, just
779 using statics to store the information is a bad idea, since GCC supports
780 nested functions, so you can be halfway through encoding one function
781 when another one comes along.
783 GCC defines a data structure called @code{struct function} which
784 contains all of the data specific to an individual function. This
785 structure contains a field called @code{machine} whose type is
786 @code{struct machine_function *}, which can be used by targets to point
787 to their own specific data.
789 If a target needs per-function specific data it should define the type
790 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
791 This macro should be used to initialize the function pointer
792 @code{init_machine_status}. This pointer is explained below.
794 One typical use of per-function, target specific data is to create an
795 RTX to hold the register containing the function's return address. This
796 RTX can then be used to implement the @code{__builtin_return_address}
797 function, for level 0.
799 Note---earlier implementations of GCC used a single data area to hold
800 all of the per-function information. Thus when processing of a nested
801 function began the old per-function data had to be pushed onto a
802 stack, and when the processing was finished, it had to be popped off the
803 stack. GCC used to provide function pointers called
804 @code{save_machine_status} and @code{restore_machine_status} to handle
805 the saving and restoring of the target specific information. Since the
806 single data area approach is no longer used, these pointers are no
809 @defmac INIT_EXPANDERS
810 Macro called to initialize any target specific information. This macro
811 is called once per function, before generation of any RTL has begun.
812 The intention of this macro is to allow the initialization of the
813 function pointer @code{init_machine_status}.
816 @deftypevar {void (*)(struct function *)} init_machine_status
817 If this function pointer is non-@code{NULL} it will be called once per
818 function, before function compilation starts, in order to allow the
819 target to perform any target specific initialization of the
820 @code{struct function} structure. It is intended that this would be
821 used to initialize the @code{machine} of that structure.
823 @code{struct machine_function} structures are expected to be freed by GC@.
824 Generally, any memory that they reference must be allocated by using
825 GC allocation, including the structure itself.
829 @section Storage Layout
830 @cindex storage layout
832 Note that the definitions of the macros in this table which are sizes or
833 alignments measured in bits do not need to be constant. They can be C
834 expressions that refer to static variables, such as the @code{target_flags}.
835 @xref{Run-time Target}.
837 @defmac BITS_BIG_ENDIAN
838 Define this macro to have the value 1 if the most significant bit in a
839 byte has the lowest number; otherwise define it to have the value zero.
840 This means that bit-field instructions count from the most significant
841 bit. If the machine has no bit-field instructions, then this must still
842 be defined, but it doesn't matter which value it is defined to. This
843 macro need not be a constant.
845 This macro does not affect the way structure fields are packed into
846 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
849 @defmac BYTES_BIG_ENDIAN
850 Define this macro to have the value 1 if the most significant byte in a
851 word has the lowest number. This macro need not be a constant.
854 @defmac WORDS_BIG_ENDIAN
855 Define this macro to have the value 1 if, in a multiword object, the
856 most significant word has the lowest number. This applies to both
857 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
858 order of words in memory is not the same as the order in registers. This
859 macro need not be a constant.
862 @defmac REG_WORDS_BIG_ENDIAN
863 On some machines, the order of words in a multiword object differs between
864 registers in memory. In such a situation, define this macro to describe
865 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
866 the order of words in memory.
869 @defmac FLOAT_WORDS_BIG_ENDIAN
870 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
871 @code{TFmode} floating point numbers are stored in memory with the word
872 containing the sign bit at the lowest address; otherwise define it to
873 have the value 0. This macro need not be a constant.
875 You need not define this macro if the ordering is the same as for
879 @defmac BITS_PER_WORD
880 Number of bits in a word. If you do not define this macro, the default
881 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
884 @defmac MAX_BITS_PER_WORD
885 Maximum number of bits in a word. If this is undefined, the default is
886 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
887 largest value that @code{BITS_PER_WORD} can have at run-time.
890 @defmac UNITS_PER_WORD
891 Number of storage units in a word; normally the size of a general-purpose
892 register, a power of two from 1 or 8.
895 @defmac MIN_UNITS_PER_WORD
896 Minimum number of units in a word. If this is undefined, the default is
897 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
898 smallest value that @code{UNITS_PER_WORD} can have at run-time.
902 Width of a pointer, in bits. You must specify a value no wider than the
903 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
904 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
905 a value the default is @code{BITS_PER_WORD}.
908 @defmac POINTERS_EXTEND_UNSIGNED
909 A C expression that determines how pointers should be extended from
910 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
911 greater than zero if pointers should be zero-extended, zero if they
912 should be sign-extended, and negative if some other sort of conversion
913 is needed. In the last case, the extension is done by the target's
914 @code{ptr_extend} instruction.
916 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
917 and @code{word_mode} are all the same width.
920 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
921 A macro to update @var{m} and @var{unsignedp} when an object whose type
922 is @var{type} and which has the specified mode and signedness is to be
923 stored in a register. This macro is only called when @var{type} is a
926 On most RISC machines, which only have operations that operate on a full
927 register, define this macro to set @var{m} to @code{word_mode} if
928 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
929 cases, only integer modes should be widened because wider-precision
930 floating-point operations are usually more expensive than their narrower
933 For most machines, the macro definition does not change @var{unsignedp}.
934 However, some machines, have instructions that preferentially handle
935 either signed or unsigned quantities of certain modes. For example, on
936 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
937 sign-extend the result to 64 bits. On such machines, set
938 @var{unsignedp} according to which kind of extension is more efficient.
940 Do not define this macro if it would never modify @var{m}.
943 @hook TARGET_C_EXCESS_PRECISION
944 Return a value, with the same meaning as the C99 macro
945 @code{FLT_EVAL_METHOD} that describes which excess precision should be
948 @hook TARGET_C_BITINT_TYPE_INFO
950 @hook TARGET_PROMOTE_FUNCTION_MODE
952 @defmac PARM_BOUNDARY
953 Normal alignment required for function parameters on the stack, in
954 bits. All stack parameters receive at least this much alignment
955 regardless of data type. On most machines, this is the same as the
959 @defmac STACK_BOUNDARY
960 Define this macro to the minimum alignment enforced by hardware for the
961 stack pointer on this machine. The definition is a C expression for the
962 desired alignment (measured in bits). This value is used as a default
963 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
964 this should be the same as @code{PARM_BOUNDARY}.
967 @defmac PREFERRED_STACK_BOUNDARY
968 Define this macro if you wish to preserve a certain alignment for the
969 stack pointer, greater than what the hardware enforces. The definition
970 is a C expression for the desired alignment (measured in bits). This
971 macro must evaluate to a value equal to or larger than
972 @code{STACK_BOUNDARY}.
975 @defmac INCOMING_STACK_BOUNDARY
976 Define this macro if the incoming stack boundary may be different
977 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
978 to a value equal to or larger than @code{STACK_BOUNDARY}.
981 @defmac FUNCTION_BOUNDARY
982 Alignment required for a function entry point, in bits.
985 @defmac BIGGEST_ALIGNMENT
986 Biggest alignment that any data type can require on this machine, in
987 bits. Note that this is not the biggest alignment that is supported,
988 just the biggest alignment that, when violated, may cause a fault.
991 @hook TARGET_ABSOLUTE_BIGGEST_ALIGNMENT
993 @defmac MALLOC_ABI_ALIGNMENT
994 Alignment, in bits, a C conformant malloc implementation has to
995 provide. If not defined, the default value is @code{BITS_PER_WORD}.
998 @defmac ATTRIBUTE_ALIGNED_VALUE
999 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1000 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1003 @defmac MINIMUM_ATOMIC_ALIGNMENT
1004 If defined, the smallest alignment, in bits, that can be given to an
1005 object that can be referenced in one operation, without disturbing any
1006 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1007 on machines that don't have byte or half-word store operations.
1010 @defmac BIGGEST_FIELD_ALIGNMENT
1011 Biggest alignment that any structure or union field can require on this
1012 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1013 structure and union fields only, unless the field alignment has been set
1014 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1017 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{type}, @var{computed})
1018 An expression for the alignment of a structure field @var{field} of
1019 type @var{type} if the alignment computed in the usual way (including
1020 applying of @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1021 alignment) is @var{computed}. It overrides alignment only if the
1022 field alignment has not been set by the
1023 @code{__attribute__ ((aligned (@var{n})))} construct. Note that @var{field}
1024 may be @code{NULL_TREE} in case we just query for the minimum alignment
1025 of a field of type @var{type} in structure context.
1028 @defmac MAX_STACK_ALIGNMENT
1029 Biggest stack alignment guaranteed by the backend. Use this macro
1030 to specify the maximum alignment of a variable on stack.
1032 If not defined, the default value is @code{STACK_BOUNDARY}.
1034 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1035 @c But the fix for PR 32893 indicates that we can only guarantee
1036 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1037 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1040 @defmac MAX_OFILE_ALIGNMENT
1041 Biggest alignment supported by the object file format of this machine.
1042 Use this macro to limit the alignment which can be specified using the
1043 @code{__attribute__ ((aligned (@var{n})))} construct for functions and
1044 objects with static storage duration. The alignment of automatic
1045 objects may exceed the object file format maximum up to the maximum
1046 supported by GCC. If not defined, the default value is
1047 @code{BIGGEST_ALIGNMENT}.
1049 On systems that use ELF, the default (in @file{config/elfos.h}) is
1050 the largest supported 32-bit ELF section alignment representable on
1051 a 32-bit host e.g.@: @samp{(((uint64_t) 1 << 28) * 8)}.
1052 On 32-bit ELF the largest supported section alignment in bits is
1053 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1056 @hook TARGET_LOWER_LOCAL_DECL_ALIGNMENT
1058 @hook TARGET_STATIC_RTX_ALIGNMENT
1060 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1061 If defined, a C expression to compute the alignment for a variable in
1062 the static store. @var{type} is the data type, and @var{basic-align} is
1063 the alignment that the object would ordinarily have. The value of this
1064 macro is used instead of that alignment to align the object.
1066 If this macro is not defined, then @var{basic-align} is used.
1069 One use of this macro is to increase alignment of medium-size data to
1070 make it all fit in fewer cache lines. Another is to cause character
1071 arrays to be word-aligned so that @code{strcpy} calls that copy
1072 constants to character arrays can be done inline.
1075 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1076 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1077 some alignment increase, instead of optimization only purposes. E.g.@
1078 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1079 must be aligned to 16 byte boundaries.
1081 If this macro is not defined, then @var{basic-align} is used.
1084 @hook TARGET_CONSTANT_ALIGNMENT
1086 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1087 If defined, a C expression to compute the alignment for a variable in
1088 the local store. @var{type} is the data type, and @var{basic-align} is
1089 the alignment that the object would ordinarily have. The value of this
1090 macro is used instead of that alignment to align the object.
1092 If this macro is not defined, then @var{basic-align} is used.
1094 One use of this macro is to increase alignment of medium-size data to
1095 make it all fit in fewer cache lines.
1097 If the value of this macro has a type, it should be an unsigned type.
1100 @hook TARGET_VECTOR_ALIGNMENT
1102 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1103 If defined, a C expression to compute the alignment for stack slot.
1104 @var{type} is the data type, @var{mode} is the widest mode available,
1105 and @var{basic-align} is the alignment that the slot would ordinarily
1106 have. The value of this macro is used instead of that alignment to
1109 If this macro is not defined, then @var{basic-align} is used when
1110 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1113 This macro is to set alignment of stack slot to the maximum alignment
1114 of all possible modes which the slot may have.
1116 If the value of this macro has a type, it should be an unsigned type.
1119 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1120 If defined, a C expression to compute the alignment for a local
1121 variable @var{decl}.
1123 If this macro is not defined, then
1124 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1127 One use of this macro is to increase alignment of medium-size data to
1128 make it all fit in fewer cache lines.
1130 If the value of this macro has a type, it should be an unsigned type.
1133 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1134 If defined, a C expression to compute the minimum required alignment
1135 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1136 @var{mode}, assuming normal alignment @var{align}.
1138 If this macro is not defined, then @var{align} will be used.
1141 @defmac EMPTY_FIELD_BOUNDARY
1142 Alignment in bits to be given to a structure bit-field that follows an
1143 empty field such as @code{int : 0;}.
1145 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1148 @defmac STRUCTURE_SIZE_BOUNDARY
1149 Number of bits which any structure or union's size must be a multiple of.
1150 Each structure or union's size is rounded up to a multiple of this.
1152 If you do not define this macro, the default is the same as
1153 @code{BITS_PER_UNIT}.
1156 @defmac STRICT_ALIGNMENT
1157 Define this macro to be the value 1 if instructions will fail to work
1158 if given data not on the nominal alignment. If instructions will merely
1159 go slower in that case, define this macro as 0.
1162 @defmac PCC_BITFIELD_TYPE_MATTERS
1163 Define this if you wish to imitate the way many other C compilers handle
1164 alignment of bit-fields and the structures that contain them.
1166 The behavior is that the type written for a named bit-field (@code{int},
1167 @code{short}, or other integer type) imposes an alignment for the entire
1168 structure, as if the structure really did contain an ordinary field of
1169 that type. In addition, the bit-field is placed within the structure so
1170 that it would fit within such a field, not crossing a boundary for it.
1172 Thus, on most machines, a named bit-field whose type is written as
1173 @code{int} would not cross a four-byte boundary, and would force
1174 four-byte alignment for the whole structure. (The alignment used may
1175 not be four bytes; it is controlled by the other alignment parameters.)
1177 An unnamed bit-field will not affect the alignment of the containing
1180 If the macro is defined, its definition should be a C expression;
1181 a nonzero value for the expression enables this behavior.
1183 Note that if this macro is not defined, or its value is zero, some
1184 bit-fields may cross more than one alignment boundary. The compiler can
1185 support such references if there are @samp{insv}, @samp{extv}, and
1186 @samp{extzv} insns that can directly reference memory.
1188 The other known way of making bit-fields work is to define
1189 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1190 Then every structure can be accessed with fullwords.
1192 Unless the machine has bit-field instructions or you define
1193 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1194 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1196 If your aim is to make GCC use the same conventions for laying out
1197 bit-fields as are used by another compiler, here is how to investigate
1198 what the other compiler does. Compile and run this program:
1217 printf ("Size of foo1 is %d\n",
1218 sizeof (struct foo1));
1219 printf ("Size of foo2 is %d\n",
1220 sizeof (struct foo2));
1225 If this prints 2 and 5, then the compiler's behavior is what you would
1226 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1229 @defmac BITFIELD_NBYTES_LIMITED
1230 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1231 to aligning a bit-field within the structure.
1234 @hook TARGET_ALIGN_ANON_BITFIELD
1236 @hook TARGET_NARROW_VOLATILE_BITFIELD
1238 @hook TARGET_MEMBER_TYPE_FORCES_BLK
1240 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1241 Define this macro as an expression for the alignment of a type (given
1242 by @var{type} as a tree node) if the alignment computed in the usual
1243 way is @var{computed} and the alignment explicitly specified was
1246 The default is to use @var{specified} if it is larger; otherwise, use
1247 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1250 @defmac MAX_FIXED_MODE_SIZE
1251 An integer expression for the size in bits of the largest integer
1252 machine mode that should actually be used. All integer machine modes of
1253 this size or smaller can be used for structures and unions with the
1254 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1255 (DImode)} is assumed.
1258 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1259 If defined, an expression of type @code{machine_mode} that
1260 specifies the mode of the save area operand of a
1261 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1262 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1263 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1264 having its mode specified.
1266 You need not define this macro if it always returns @code{Pmode}. You
1267 would most commonly define this macro if the
1268 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1272 @defmac STACK_SIZE_MODE
1273 If defined, an expression of type @code{machine_mode} that
1274 specifies the mode of the size increment operand of an
1275 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1277 You need not define this macro if it always returns @code{word_mode}.
1278 You would most commonly define this macro if the @code{allocate_stack}
1279 pattern needs to support both a 32- and a 64-bit mode.
1282 @hook TARGET_LIBGCC_CMP_RETURN_MODE
1284 @hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1286 @hook TARGET_UNWIND_WORD_MODE
1288 @hook TARGET_MS_BITFIELD_LAYOUT_P
1290 @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1292 @hook TARGET_FIXED_POINT_SUPPORTED_P
1294 @hook TARGET_EXPAND_TO_RTL_HOOK
1296 @hook TARGET_INSTANTIATE_DECLS
1298 @hook TARGET_MANGLE_TYPE
1300 @hook TARGET_EMIT_SUPPORT_TINFOS
1303 @section Layout of Source Language Data Types
1305 These macros define the sizes and other characteristics of the standard
1306 basic data types used in programs being compiled. Unlike the macros in
1307 the previous section, these apply to specific features of C and related
1308 languages, rather than to fundamental aspects of storage layout.
1310 @defmac INT_TYPE_SIZE
1311 A C expression for the size in bits of the type @code{int} on the
1312 target machine. If you don't define this, the default is one word.
1315 @defmac SHORT_TYPE_SIZE
1316 A C expression for the size in bits of the type @code{short} on the
1317 target machine. If you don't define this, the default is half a word.
1318 (If this would be less than one storage unit, it is rounded up to one
1322 @defmac LONG_TYPE_SIZE
1323 A C expression for the size in bits of the type @code{long} on the
1324 target machine. If you don't define this, the default is one word.
1327 @defmac ADA_LONG_TYPE_SIZE
1328 On some machines, the size used for the Ada equivalent of the type
1329 @code{long} by a native Ada compiler differs from that used by C@. In
1330 that situation, define this macro to be a C expression to be used for
1331 the size of that type. If you don't define this, the default is the
1332 value of @code{LONG_TYPE_SIZE}.
1335 @defmac LONG_LONG_TYPE_SIZE
1336 A C expression for the size in bits of the type @code{long long} on the
1337 target machine. If you don't define this, the default is two
1338 words. If you want to support GNU Ada on your machine, the value of this
1339 macro must be at least 64.
1342 @defmac CHAR_TYPE_SIZE
1343 A C expression for the size in bits of the type @code{char} on the
1344 target machine. If you don't define this, the default is
1345 @code{BITS_PER_UNIT}.
1348 @defmac BOOL_TYPE_SIZE
1349 A C expression for the size in bits of the C++ type @code{bool} and
1350 C99 type @code{_Bool} on the target machine. If you don't define
1351 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1354 @defmac FLOAT_TYPE_SIZE
1355 A C expression for the size in bits of the type @code{float} on the
1356 target machine. If you don't define this, the default is one word.
1359 @defmac DOUBLE_TYPE_SIZE
1360 A C expression for the size in bits of the type @code{double} on the
1361 target machine. If you don't define this, the default is two
1365 @defmac LONG_DOUBLE_TYPE_SIZE
1366 A C expression for the size in bits of the type @code{long double} on
1367 the target machine. If you don't define this, the default is two
1371 @defmac SHORT_FRACT_TYPE_SIZE
1372 A C expression for the size in bits of the type @code{short _Fract} on
1373 the target machine. If you don't define this, the default is
1374 @code{BITS_PER_UNIT}.
1377 @defmac FRACT_TYPE_SIZE
1378 A C expression for the size in bits of the type @code{_Fract} on
1379 the target machine. If you don't define this, the default is
1380 @code{BITS_PER_UNIT * 2}.
1383 @defmac LONG_FRACT_TYPE_SIZE
1384 A C expression for the size in bits of the type @code{long _Fract} on
1385 the target machine. If you don't define this, the default is
1386 @code{BITS_PER_UNIT * 4}.
1389 @defmac LONG_LONG_FRACT_TYPE_SIZE
1390 A C expression for the size in bits of the type @code{long long _Fract} on
1391 the target machine. If you don't define this, the default is
1392 @code{BITS_PER_UNIT * 8}.
1395 @defmac SHORT_ACCUM_TYPE_SIZE
1396 A C expression for the size in bits of the type @code{short _Accum} on
1397 the target machine. If you don't define this, the default is
1398 @code{BITS_PER_UNIT * 2}.
1401 @defmac ACCUM_TYPE_SIZE
1402 A C expression for the size in bits of the type @code{_Accum} on
1403 the target machine. If you don't define this, the default is
1404 @code{BITS_PER_UNIT * 4}.
1407 @defmac LONG_ACCUM_TYPE_SIZE
1408 A C expression for the size in bits of the type @code{long _Accum} on
1409 the target machine. If you don't define this, the default is
1410 @code{BITS_PER_UNIT * 8}.
1413 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1414 A C expression for the size in bits of the type @code{long long _Accum} on
1415 the target machine. If you don't define this, the default is
1416 @code{BITS_PER_UNIT * 16}.
1419 @defmac LIBGCC2_GNU_PREFIX
1420 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1421 hook and should be defined if that hook is overriden to be true. It
1422 causes function names in libgcc to be changed to use a @code{__gnu_}
1423 prefix for their name rather than the default @code{__}. A port which
1424 uses this macro should also arrange to use @file{t-gnu-prefix} in
1425 the libgcc @file{config.host}.
1428 @defmac WIDEST_HARDWARE_FP_SIZE
1429 A C expression for the size in bits of the widest floating-point format
1430 supported by the hardware. If you define this macro, you must specify a
1431 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1432 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1436 @defmac DEFAULT_SIGNED_CHAR
1437 An expression whose value is 1 or 0, according to whether the type
1438 @code{char} should be signed or unsigned by default. The user can
1439 always override this default with the options @option{-fsigned-char}
1440 and @option{-funsigned-char}.
1443 @hook TARGET_DEFAULT_SHORT_ENUMS
1446 A C expression for a string describing the name of the data type to use
1447 for size values. The typedef name @code{size_t} is defined using the
1448 contents of the string.
1450 The string can contain more than one keyword. If so, separate them with
1451 spaces, and write first any length keyword, then @code{unsigned} if
1452 appropriate, and finally @code{int}. The string must exactly match one
1453 of the data type names defined in the function
1454 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.cc}.
1455 You may not omit @code{int} or change the order---that would cause the
1456 compiler to crash on startup.
1458 If you don't define this macro, the default is @code{"long unsigned
1463 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1464 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1465 dealing with size. This macro is a C expression for a string describing
1466 the name of the data type from which the precision of @code{sizetype}
1469 The string has the same restrictions as @code{SIZE_TYPE} string.
1471 If you don't define this macro, the default is @code{SIZE_TYPE}.
1474 @defmac PTRDIFF_TYPE
1475 A C expression for a string describing the name of the data type to use
1476 for the result of subtracting two pointers. The typedef name
1477 @code{ptrdiff_t} is defined using the contents of the string. See
1478 @code{SIZE_TYPE} above for more information.
1480 If you don't define this macro, the default is @code{"long int"}.
1484 A C expression for a string describing the name of the data type to use
1485 for wide characters. The typedef name @code{wchar_t} is defined using
1486 the contents of the string. See @code{SIZE_TYPE} above for more
1489 If you don't define this macro, the default is @code{"int"}.
1492 @defmac WCHAR_TYPE_SIZE
1493 A C expression for the size in bits of the data type for wide
1494 characters. This is used in @code{cpp}, which cannot make use of
1499 A C expression for a string describing the name of the data type to
1500 use for wide characters passed to @code{printf} and returned from
1501 @code{getwc}. The typedef name @code{wint_t} is defined using the
1502 contents of the string. See @code{SIZE_TYPE} above for more
1505 If you don't define this macro, the default is @code{"unsigned int"}.
1509 A C expression for a string describing the name of the data type that
1510 can represent any value of any standard or extended signed integer type.
1511 The typedef name @code{intmax_t} is defined using the contents of the
1512 string. See @code{SIZE_TYPE} above for more information.
1514 If you don't define this macro, the default is the first of
1515 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1516 much precision as @code{long long int}.
1519 @defmac UINTMAX_TYPE
1520 A C expression for a string describing the name of the data type that
1521 can represent any value of any standard or extended unsigned integer
1522 type. The typedef name @code{uintmax_t} is defined using the contents
1523 of the string. See @code{SIZE_TYPE} above for more information.
1525 If you don't define this macro, the default is the first of
1526 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1527 unsigned int"} that has as much precision as @code{long long unsigned
1531 @defmac SIG_ATOMIC_TYPE
1537 @defmacx UINT16_TYPE
1538 @defmacx UINT32_TYPE
1539 @defmacx UINT64_TYPE
1540 @defmacx INT_LEAST8_TYPE
1541 @defmacx INT_LEAST16_TYPE
1542 @defmacx INT_LEAST32_TYPE
1543 @defmacx INT_LEAST64_TYPE
1544 @defmacx UINT_LEAST8_TYPE
1545 @defmacx UINT_LEAST16_TYPE
1546 @defmacx UINT_LEAST32_TYPE
1547 @defmacx UINT_LEAST64_TYPE
1548 @defmacx INT_FAST8_TYPE
1549 @defmacx INT_FAST16_TYPE
1550 @defmacx INT_FAST32_TYPE
1551 @defmacx INT_FAST64_TYPE
1552 @defmacx UINT_FAST8_TYPE
1553 @defmacx UINT_FAST16_TYPE
1554 @defmacx UINT_FAST32_TYPE
1555 @defmacx UINT_FAST64_TYPE
1556 @defmacx INTPTR_TYPE
1557 @defmacx UINTPTR_TYPE
1558 C expressions for the standard types @code{sig_atomic_t},
1559 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1560 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1561 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1562 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1563 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1564 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1565 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1566 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1567 @code{SIZE_TYPE} above for more information.
1569 If any of these macros evaluates to a null pointer, the corresponding
1570 type is not supported; if GCC is configured to provide
1571 @code{<stdint.h>} in such a case, the header provided may not conform
1572 to C99, depending on the type in question. The defaults for all of
1573 these macros are null pointers.
1576 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1577 The C++ compiler represents a pointer-to-member-function with a struct
1584 ptrdiff_t vtable_index;
1591 The C++ compiler must use one bit to indicate whether the function that
1592 will be called through a pointer-to-member-function is virtual.
1593 Normally, we assume that the low-order bit of a function pointer must
1594 always be zero. Then, by ensuring that the vtable_index is odd, we can
1595 distinguish which variant of the union is in use. But, on some
1596 platforms function pointers can be odd, and so this doesn't work. In
1597 that case, we use the low-order bit of the @code{delta} field, and shift
1598 the remainder of the @code{delta} field to the left.
1600 GCC will automatically make the right selection about where to store
1601 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1602 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1603 set such that functions always start at even addresses, but the lowest
1604 bit of pointers to functions indicate whether the function at that
1605 address is in ARM or Thumb mode. If this is the case of your
1606 architecture, you should define this macro to
1607 @code{ptrmemfunc_vbit_in_delta}.
1609 In general, you should not have to define this macro. On architectures
1610 in which function addresses are always even, according to
1611 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1612 @code{ptrmemfunc_vbit_in_pfn}.
1615 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1616 Normally, the C++ compiler uses function pointers in vtables. This
1617 macro allows the target to change to use ``function descriptors''
1618 instead. Function descriptors are found on targets for whom a
1619 function pointer is actually a small data structure. Normally the
1620 data structure consists of the actual code address plus a data
1621 pointer to which the function's data is relative.
1623 If vtables are used, the value of this macro should be the number
1624 of words that the function descriptor occupies.
1627 @defmac TARGET_VTABLE_ENTRY_ALIGN
1628 By default, the vtable entries are void pointers, the so the alignment
1629 is the same as pointer alignment. The value of this macro specifies
1630 the alignment of the vtable entry in bits. It should be defined only
1631 when special alignment is necessary. */
1634 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1635 There are a few non-descriptor entries in the vtable at offsets below
1636 zero. If these entries must be padded (say, to preserve the alignment
1637 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1638 of words in each data entry.
1642 @section Register Usage
1643 @cindex register usage
1645 This section explains how to describe what registers the target machine
1646 has, and how (in general) they can be used.
1648 The description of which registers a specific instruction can use is
1649 done with register classes; see @ref{Register Classes}. For information
1650 on using registers to access a stack frame, see @ref{Frame Registers}.
1651 For passing values in registers, see @ref{Register Arguments}.
1652 For returning values in registers, see @ref{Scalar Return}.
1655 * Register Basics:: Number and kinds of registers.
1656 * Allocation Order:: Order in which registers are allocated.
1657 * Values in Registers:: What kinds of values each reg can hold.
1658 * Leaf Functions:: Renumbering registers for leaf functions.
1659 * Stack Registers:: Handling a register stack such as 80387.
1662 @node Register Basics
1663 @subsection Basic Characteristics of Registers
1665 @c prevent bad page break with this line
1666 Registers have various characteristics.
1668 @defmac FIRST_PSEUDO_REGISTER
1669 Number of hardware registers known to the compiler. They receive
1670 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1671 pseudo register's number really is assigned the number
1672 @code{FIRST_PSEUDO_REGISTER}.
1675 @defmac FIXED_REGISTERS
1676 @cindex fixed register
1677 An initializer that says which registers are used for fixed purposes
1678 all throughout the compiled code and are therefore not available for
1679 general allocation. These would include the stack pointer, the frame
1680 pointer (except on machines where that can be used as a general
1681 register when no frame pointer is needed), the program counter on
1682 machines where that is considered one of the addressable registers,
1683 and any other numbered register with a standard use.
1685 This information is expressed as a sequence of numbers, separated by
1686 commas and surrounded by braces. The @var{n}th number is 1 if
1687 register @var{n} is fixed, 0 otherwise.
1689 The table initialized from this macro, and the table initialized by
1690 the following one, may be overridden at run time either automatically,
1691 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1692 the user with the command options @option{-ffixed-@var{reg}},
1693 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1696 @defmac CALL_USED_REGISTERS
1697 @cindex call-used register
1698 @cindex call-clobbered register
1699 @cindex call-saved register
1700 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1701 clobbered (in general) by function calls as well as for fixed
1702 registers. This macro therefore identifies the registers that are not
1703 available for general allocation of values that must live across
1706 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1707 automatically saves it on function entry and restores it on function
1708 exit, if the register is used within the function.
1710 Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS}
1711 must be defined. Modern ports should define @code{CALL_REALLY_USED_REGISTERS}.
1714 @defmac CALL_REALLY_USED_REGISTERS
1715 @cindex call-used register
1716 @cindex call-clobbered register
1717 @cindex call-saved register
1718 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1719 that the entire set of @code{FIXED_REGISTERS} be included.
1720 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1722 Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS}
1723 must be defined. Modern ports should define @code{CALL_REALLY_USED_REGISTERS}.
1726 @cindex call-used register
1727 @cindex call-clobbered register
1728 @cindex call-saved register
1729 @hook TARGET_FNTYPE_ABI
1731 @hook TARGET_INSN_CALLEE_ABI
1733 @cindex call-used register
1734 @cindex call-clobbered register
1735 @cindex call-saved register
1736 @hook TARGET_HARD_REGNO_CALL_PART_CLOBBERED
1738 @hook TARGET_GET_MULTILIB_ABI_NAME
1741 @findex call_used_regs
1744 @findex reg_class_contents
1745 @hook TARGET_CONDITIONAL_REGISTER_USAGE
1747 @defmac INCOMING_REGNO (@var{out})
1748 Define this macro if the target machine has register windows. This C
1749 expression returns the register number as seen by the called function
1750 corresponding to the register number @var{out} as seen by the calling
1751 function. Return @var{out} if register number @var{out} is not an
1755 @defmac OUTGOING_REGNO (@var{in})
1756 Define this macro if the target machine has register windows. This C
1757 expression returns the register number as seen by the calling function
1758 corresponding to the register number @var{in} as seen by the called
1759 function. Return @var{in} if register number @var{in} is not an inbound
1763 @defmac LOCAL_REGNO (@var{regno})
1764 Define this macro if the target machine has register windows. This C
1765 expression returns true if the register is call-saved but is in the
1766 register window. Unlike most call-saved registers, such registers
1767 need not be explicitly restored on function exit or during non-local
1772 If the program counter has a register number, define this as that
1773 register number. Otherwise, do not define it.
1776 @node Allocation Order
1777 @subsection Order of Allocation of Registers
1778 @cindex order of register allocation
1779 @cindex register allocation order
1781 @c prevent bad page break with this line
1782 Registers are allocated in order.
1784 @defmac REG_ALLOC_ORDER
1785 If defined, an initializer for a vector of integers, containing the
1786 numbers of hard registers in the order in which GCC should prefer
1787 to use them (from most preferred to least).
1789 If this macro is not defined, registers are used lowest numbered first
1790 (all else being equal).
1792 One use of this macro is on machines where the highest numbered
1793 registers must always be saved and the save-multiple-registers
1794 instruction supports only sequences of consecutive registers. On such
1795 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1796 the highest numbered allocable register first.
1799 @defmac ADJUST_REG_ALLOC_ORDER
1800 A C statement (sans semicolon) to choose the order in which to allocate
1801 hard registers for pseudo-registers local to a basic block.
1803 Store the desired register order in the array @code{reg_alloc_order}.
1804 Element 0 should be the register to allocate first; element 1, the next
1805 register; and so on.
1807 The macro body should not assume anything about the contents of
1808 @code{reg_alloc_order} before execution of the macro.
1810 On most machines, it is not necessary to define this macro.
1813 @defmac HONOR_REG_ALLOC_ORDER
1814 Normally, IRA tries to estimate the costs for saving a register in the
1815 prologue and restoring it in the epilogue. This discourages it from
1816 using call-saved registers. If a machine wants to ensure that IRA
1817 allocates registers in the order given by REG_ALLOC_ORDER even if some
1818 call-saved registers appear earlier than call-used ones, then define this
1819 macro as a C expression to nonzero. Default is 0.
1822 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
1823 In some case register allocation order is not enough for the
1824 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
1825 If this macro is defined, it should return a floating point value
1826 based on @var{regno}. The cost of using @var{regno} for a pseudo will
1827 be increased by approximately the pseudo's usage frequency times the
1828 value returned by this macro. Not defining this macro is equivalent
1829 to having it always return @code{0.0}.
1831 On most machines, it is not necessary to define this macro.
1834 @node Values in Registers
1835 @subsection How Values Fit in Registers
1837 This section discusses the macros that describe which kinds of values
1838 (specifically, which machine modes) each register can hold, and how many
1839 consecutive registers are needed for a given mode.
1841 @hook TARGET_HARD_REGNO_NREGS
1843 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
1844 A C expression that is nonzero if a value of mode @var{mode}, stored
1845 in memory, ends with padding that causes it to take up more space than
1846 in registers starting at register number @var{regno} (as determined by
1847 multiplying GCC's notion of the size of the register when containing
1848 this mode by the number of registers returned by
1849 @code{TARGET_HARD_REGNO_NREGS}). By default this is zero.
1851 For example, if a floating-point value is stored in three 32-bit
1852 registers but takes up 128 bits in memory, then this would be
1855 This macros only needs to be defined if there are cases where
1856 @code{subreg_get_info}
1857 would otherwise wrongly determine that a @code{subreg} can be
1858 represented by an offset to the register number, when in fact such a
1859 @code{subreg} would contain some of the padding not stored in
1860 registers and so not be representable.
1863 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
1864 For values of @var{regno} and @var{mode} for which
1865 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
1866 returning the greater number of registers required to hold the value
1867 including any padding. In the example above, the value would be four.
1870 @defmac REGMODE_NATURAL_SIZE (@var{mode})
1871 Define this macro if the natural size of registers that hold values
1872 of mode @var{mode} is not the word size. It is a C expression that
1873 should give the natural size in bytes for the specified mode. It is
1874 used by the register allocator to try to optimize its results. This
1875 happens for example on SPARC 64-bit where the natural size of
1876 floating-point registers is still 32-bit.
1879 @hook TARGET_HARD_REGNO_MODE_OK
1881 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
1882 A C expression that is nonzero if it is OK to rename a hard register
1883 @var{from} to another hard register @var{to}.
1885 One common use of this macro is to prevent renaming of a register to
1886 another register that is not saved by a prologue in an interrupt
1889 The default is always nonzero.
1892 @hook TARGET_MODES_TIEABLE_P
1894 @hook TARGET_HARD_REGNO_SCRATCH_OK
1896 @defmac AVOID_CCMODE_COPIES
1897 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1898 registers. You should only define this macro if support for copying to/from
1899 @code{CCmode} is incomplete.
1902 @node Leaf Functions
1903 @subsection Handling Leaf Functions
1905 @cindex leaf functions
1906 @cindex functions, leaf
1907 On some machines, a leaf function (i.e., one which makes no calls) can run
1908 more efficiently if it does not make its own register window. Often this
1909 means it is required to receive its arguments in the registers where they
1910 are passed by the caller, instead of the registers where they would
1913 The special treatment for leaf functions generally applies only when
1914 other conditions are met; for example, often they may use only those
1915 registers for its own variables and temporaries. We use the term ``leaf
1916 function'' to mean a function that is suitable for this special
1917 handling, so that functions with no calls are not necessarily ``leaf
1920 GCC assigns register numbers before it knows whether the function is
1921 suitable for leaf function treatment. So it needs to renumber the
1922 registers in order to output a leaf function. The following macros
1925 @defmac LEAF_REGISTERS
1926 Name of a char vector, indexed by hard register number, which
1927 contains 1 for a register that is allowable in a candidate for leaf
1930 If leaf function treatment involves renumbering the registers, then the
1931 registers marked here should be the ones before renumbering---those that
1932 GCC would ordinarily allocate. The registers which will actually be
1933 used in the assembler code, after renumbering, should not be marked with 1
1936 Define this macro only if the target machine offers a way to optimize
1937 the treatment of leaf functions.
1940 @defmac LEAF_REG_REMAP (@var{regno})
1941 A C expression whose value is the register number to which @var{regno}
1942 should be renumbered, when a function is treated as a leaf function.
1944 If @var{regno} is a register number which should not appear in a leaf
1945 function before renumbering, then the expression should yield @minus{}1, which
1946 will cause the compiler to abort.
1948 Define this macro only if the target machine offers a way to optimize the
1949 treatment of leaf functions, and registers need to be renumbered to do
1953 @findex current_function_is_leaf
1954 @findex current_function_uses_only_leaf_regs
1955 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
1956 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
1957 specially. They can test the C variable @code{current_function_is_leaf}
1958 which is nonzero for leaf functions. @code{current_function_is_leaf} is
1959 set prior to local register allocation and is valid for the remaining
1960 compiler passes. They can also test the C variable
1961 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
1962 functions which only use leaf registers.
1963 @code{current_function_uses_only_leaf_regs} is valid after all passes
1964 that modify the instructions have been run and is only useful if
1965 @code{LEAF_REGISTERS} is defined.
1966 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1967 @c of the next paragraph?! --mew 2feb93
1969 @node Stack Registers
1970 @subsection Registers That Form a Stack
1972 There are special features to handle computers where some of the
1973 ``registers'' form a stack. Stack registers are normally written by
1974 pushing onto the stack, and are numbered relative to the top of the
1977 Currently, GCC can only handle one group of stack-like registers, and
1978 they must be consecutively numbered. Furthermore, the existing
1979 support for stack-like registers is specific to the 80387 floating
1980 point coprocessor. If you have a new architecture that uses
1981 stack-like registers, you will need to do substantial work on
1982 @file{reg-stack.cc} and write your machine description to cooperate
1983 with it, as well as defining these macros.
1986 Define this if the machine has any stack-like registers.
1989 @defmac STACK_REG_COVER_CLASS
1990 This is a cover class containing the stack registers. Define this if
1991 the machine has any stack-like registers.
1994 @defmac FIRST_STACK_REG
1995 The number of the first stack-like register. This one is the top
1999 @defmac LAST_STACK_REG
2000 The number of the last stack-like register. This one is the bottom of
2004 @node Register Classes
2005 @section Register Classes
2006 @cindex register class definitions
2007 @cindex class definitions, register
2009 On many machines, the numbered registers are not all equivalent.
2010 For example, certain registers may not be allowed for indexed addressing;
2011 certain registers may not be allowed in some instructions. These machine
2012 restrictions are described to the compiler using @dfn{register classes}.
2014 You define a number of register classes, giving each one a name and saying
2015 which of the registers belong to it. Then you can specify register classes
2016 that are allowed as operands to particular instruction patterns.
2020 In general, each register will belong to several classes. In fact, one
2021 class must be named @code{ALL_REGS} and contain all the registers. Another
2022 class must be named @code{NO_REGS} and contain no registers. Often the
2023 union of two classes will be another class; however, this is not required.
2025 @findex GENERAL_REGS
2026 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2027 terribly special about the name, but the operand constraint letters
2028 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2029 the same as @code{ALL_REGS}, just define it as a macro which expands
2032 Order the classes so that if class @var{x} is contained in class @var{y}
2033 then @var{x} has a lower class number than @var{y}.
2035 The way classes other than @code{GENERAL_REGS} are specified in operand
2036 constraints is through machine-dependent operand constraint letters.
2037 You can define such letters to correspond to various classes, then use
2038 them in operand constraints.
2040 You must define the narrowest register classes for allocatable
2041 registers, so that each class either has no subclasses, or that for
2042 some mode, the move cost between registers within the class is
2043 cheaper than moving a register in the class to or from memory
2046 You should define a class for the union of two classes whenever some
2047 instruction allows both classes. For example, if an instruction allows
2048 either a floating point (coprocessor) register or a general register for a
2049 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2050 which includes both of them. Otherwise you will get suboptimal code,
2051 or even internal compiler errors when reload cannot find a register in the
2052 class computed via @code{reg_class_subunion}.
2054 You must also specify certain redundant information about the register
2055 classes: for each class, which classes contain it and which ones are
2056 contained in it; for each pair of classes, the largest class contained
2059 When a value occupying several consecutive registers is expected in a
2060 certain class, all the registers used must belong to that class.
2061 Therefore, register classes cannot be used to enforce a requirement for
2062 a register pair to start with an even-numbered register. The way to
2063 specify this requirement is with @code{TARGET_HARD_REGNO_MODE_OK}.
2065 Register classes used for input-operands of bitwise-and or shift
2066 instructions have a special requirement: each such class must have, for
2067 each fixed-point machine mode, a subclass whose registers can transfer that
2068 mode to or from memory. For example, on some machines, the operations for
2069 single-byte values (@code{QImode}) are limited to certain registers. When
2070 this is so, each register class that is used in a bitwise-and or shift
2071 instruction must have a subclass consisting of registers from which
2072 single-byte values can be loaded or stored. This is so that
2073 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2075 @deftp {Data type} {enum reg_class}
2076 An enumerated type that must be defined with all the register class names
2077 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2078 must be the last register class, followed by one more enumerated value,
2079 @code{LIM_REG_CLASSES}, which is not a register class but rather
2080 tells how many classes there are.
2082 Each register class has a number, which is the value of casting
2083 the class name to type @code{int}. The number serves as an index
2084 in many of the tables described below.
2087 @defmac N_REG_CLASSES
2088 The number of distinct register classes, defined as follows:
2091 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2095 @defmac REG_CLASS_NAMES
2096 An initializer containing the names of the register classes as C string
2097 constants. These names are used in writing some of the debugging dumps.
2100 @defmac REG_CLASS_CONTENTS
2101 An initializer containing the contents of the register classes, as integers
2102 which are bit masks. The @var{n}th integer specifies the contents of class
2103 @var{n}. The way the integer @var{mask} is interpreted is that
2104 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2106 When the machine has more than 32 registers, an integer does not suffice.
2107 Then the integers are replaced by sub-initializers, braced groupings containing
2108 several integers. Each sub-initializer must be suitable as an initializer
2109 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2110 In this situation, the first integer in each sub-initializer corresponds to
2111 registers 0 through 31, the second integer to registers 32 through 63, and
2115 @defmac REGNO_REG_CLASS (@var{regno})
2116 A C expression whose value is a register class containing hard register
2117 @var{regno}. In general there is more than one such class; choose a class
2118 which is @dfn{minimal}, meaning that no smaller class also contains the
2122 @defmac BASE_REG_CLASS
2123 A macro whose definition is the name of the class to which a valid
2124 base register must belong. A base register is one used in an address
2125 which is the register value plus a displacement.
2128 @defmac MODE_BASE_REG_CLASS (@var{mode})
2129 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2130 the selection of a base register in a mode dependent manner. If
2131 @var{mode} is VOIDmode then it should return the same value as
2132 @code{BASE_REG_CLASS}.
2135 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2136 A C expression whose value is the register class to which a valid
2137 base register must belong in order to be used in a base plus index
2138 register address. You should define this macro if base plus index
2139 addresses have different requirements than other base register uses.
2142 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2143 A C expression whose value is the register class to which a valid
2144 base register for a memory reference in mode @var{mode} to address
2145 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2146 define the context in which the base register occurs. @var{outer_code} is
2147 the code of the immediately enclosing expression (@code{MEM} for the top level
2148 of an address, @code{ADDRESS} for something that occurs in an
2149 @code{address_operand}). @var{index_code} is the code of the corresponding
2150 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2153 @defmac INSN_BASE_REG_CLASS (@var{insn})
2154 A C expression whose value is the register class to which a valid
2155 base register for a specified @var{insn} must belong. This macro is
2156 used when some backend insns may have limited usage of base register
2157 compared with other insns. If you define this macro, the compiler will
2158 use it instead of all other defined macros that relate to
2162 @defmac INDEX_REG_CLASS
2163 A macro whose definition is the name of the class to which a valid
2164 index register must belong. An index register is one used in an
2165 address where its value is either multiplied by a scale factor or
2166 added to another register (as well as added to a displacement).
2169 @defmac INSN_INDEX_REG_CLASS (@var{insn})
2170 A C expression whose value is the register class to which a valid
2171 index register for a specified @var{insn} must belong. This macro is
2172 used when some backend insns may have limited usage of index register
2173 compared with other insns. If you defined this macro, the compiler
2174 will use it instead of @code{INDEX_REG_CLASS}.
2177 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2178 A C expression which is nonzero if register number @var{num} is
2179 suitable for use as a base register in operand addresses.
2182 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2183 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2184 that expression may examine the mode of the memory reference in
2185 @var{mode}. You should define this macro if the mode of the memory
2186 reference affects whether a register may be used as a base register. If
2187 you define this macro, the compiler will use it instead of
2188 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2189 addresses that appear outside a @code{MEM}, i.e., as an
2190 @code{address_operand}.
2193 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2194 A C expression which is nonzero if register number @var{num} is suitable for
2195 use as a base register in base plus index operand addresses, accessing
2196 memory in mode @var{mode}. It may be either a suitable hard register or a
2197 pseudo register that has been allocated such a hard register. You should
2198 define this macro if base plus index addresses have different requirements
2199 than other base register uses.
2201 Use of this macro is deprecated; please use the more general
2202 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2205 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2206 A C expression which is nonzero if register number @var{num} is
2207 suitable for use as a base register in operand addresses, accessing
2208 memory in mode @var{mode} in address space @var{address_space}.
2209 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2210 that that expression may examine the context in which the register
2211 appears in the memory reference. @var{outer_code} is the code of the
2212 immediately enclosing expression (@code{MEM} if at the top level of the
2213 address, @code{ADDRESS} for something that occurs in an
2214 @code{address_operand}). @var{index_code} is the code of the
2215 corresponding index expression if @var{outer_code} is @code{PLUS};
2216 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2217 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2220 @defmac REGNO_OK_FOR_INSN_BASE_P (@var{num}, @var{insn})
2221 A C expression which is nonzero if register number @var{num} is
2222 suitable for use as a base register in operand addresses for a specified
2223 @var{insn}. This macro is used when some backend insn may have limited
2224 usage of base register compared with other insns. If you define this
2225 macro, the compiler will use it instead of all other defined macros
2226 that relate to REGNO_OK_FOR_BASE_P.
2229 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2230 A C expression which is nonzero if register number @var{num} is
2231 suitable for use as an index register in operand addresses. It may be
2232 either a suitable hard register or a pseudo register that has been
2233 allocated such a hard register.
2235 The difference between an index register and a base register is that
2236 the index register may be scaled. If an address involves the sum of
2237 two registers, neither one of them scaled, then either one may be
2238 labeled the ``base'' and the other the ``index''; but whichever
2239 labeling is used must fit the machine's constraints of which registers
2240 may serve in each capacity. The compiler will try both labelings,
2241 looking for one that is valid, and will reload one or both registers
2242 only if neither labeling works.
2245 @hook TARGET_PREFERRED_RENAME_CLASS
2247 @hook TARGET_PREFERRED_RELOAD_CLASS
2249 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2250 A C expression that places additional restrictions on the register class
2251 to use when it is necessary to copy value @var{x} into a register in class
2252 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2253 another, smaller class. On many machines, the following definition is
2257 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2260 Sometimes returning a more restrictive class makes better code. For
2261 example, on the 68000, when @var{x} is an integer constant that is in range
2262 for a @samp{moveq} instruction, the value of this macro is always
2263 @code{DATA_REGS} as long as @var{class} includes the data registers.
2264 Requiring a data register guarantees that a @samp{moveq} will be used.
2266 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2267 @var{class} is if @var{x} is a legitimate constant which cannot be
2268 loaded into some register class. By returning @code{NO_REGS} you can
2269 force @var{x} into a memory location. For example, rs6000 can load
2270 immediate values into general-purpose registers, but does not have an
2271 instruction for loading an immediate value into a floating-point
2272 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2273 @var{x} is a floating-point constant. If the constant cannot be loaded
2274 into any kind of register, code generation will be better if
2275 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2276 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2278 If an insn has pseudos in it after register allocation, reload will go
2279 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2280 to find the best one. Returning @code{NO_REGS}, in this case, makes
2281 reload add a @code{!} in front of the constraint: the x86 back-end uses
2282 this feature to discourage usage of 387 registers when math is done in
2283 the SSE registers (and vice versa).
2286 @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2288 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2289 A C expression that places additional restrictions on the register class
2290 to use when it is necessary to be able to hold a value of mode
2291 @var{mode} in a reload register for which class @var{class} would
2294 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2295 there are certain modes that simply cannot go in certain reload classes.
2297 The value is a register class; perhaps @var{class}, or perhaps another,
2300 Don't define this macro unless the target machine has limitations which
2301 require the macro to do something nontrivial.
2304 @hook TARGET_SECONDARY_RELOAD
2306 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2307 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2308 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2309 These macros are obsolete, new ports should use the target hook
2310 @code{TARGET_SECONDARY_RELOAD} instead.
2312 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2313 target hook. Older ports still define these macros to indicate to the
2314 reload phase that it may
2315 need to allocate at least one register for a reload in addition to the
2316 register to contain the data. Specifically, if copying @var{x} to a
2317 register @var{class} in @var{mode} requires an intermediate register,
2318 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2319 largest register class all of whose registers can be used as
2320 intermediate registers or scratch registers.
2322 If copying a register @var{class} in @var{mode} to @var{x} requires an
2323 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2324 was supposed to be defined to return the largest register
2325 class required. If the
2326 requirements for input and output reloads were the same, the macro
2327 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2330 The values returned by these macros are often @code{GENERAL_REGS}.
2331 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2332 can be directly copied to or from a register of @var{class} in
2333 @var{mode} without requiring a scratch register. Do not define this
2334 macro if it would always return @code{NO_REGS}.
2336 If a scratch register is required (either with or without an
2337 intermediate register), you were supposed to define patterns for
2338 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2339 (@pxref{Standard Names}. These patterns, which were normally
2340 implemented with a @code{define_expand}, should be similar to the
2341 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2344 These patterns need constraints for the reload register and scratch
2346 contain a single register class. If the original reload register (whose
2347 class is @var{class}) can meet the constraint given in the pattern, the
2348 value returned by these macros is used for the class of the scratch
2349 register. Otherwise, two additional reload registers are required.
2350 Their classes are obtained from the constraints in the insn pattern.
2352 @var{x} might be a pseudo-register or a @code{subreg} of a
2353 pseudo-register, which could either be in a hard register or in memory.
2354 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2355 in memory and the hard register number if it is in a register.
2357 These macros should not be used in the case where a particular class of
2358 registers can only be copied to memory and not to another class of
2359 registers. In that case, secondary reload registers are not needed and
2360 would not be helpful. Instead, a stack location must be used to perform
2361 the copy and the @code{mov@var{m}} pattern should use memory as an
2362 intermediate storage. This case often occurs between floating-point and
2366 @hook TARGET_SECONDARY_MEMORY_NEEDED
2368 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2369 Normally when @code{TARGET_SECONDARY_MEMORY_NEEDED} is defined, the compiler
2370 allocates a stack slot for a memory location needed for register copies.
2371 If this macro is defined, the compiler instead uses the memory location
2372 defined by this macro.
2374 Do not define this macro if you do not define
2375 @code{TARGET_SECONDARY_MEMORY_NEEDED}.
2378 @hook TARGET_SECONDARY_MEMORY_NEEDED_MODE
2380 @hook TARGET_SELECT_EARLY_REMAT_MODES
2382 @hook TARGET_CLASS_LIKELY_SPILLED_P
2384 @hook TARGET_CLASS_MAX_NREGS
2386 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2387 A C expression for the maximum number of consecutive registers
2388 of class @var{class} needed to hold a value of mode @var{mode}.
2390 This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}. In fact,
2391 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2392 should be the maximum value of @code{TARGET_HARD_REGNO_NREGS (@var{regno},
2393 @var{mode})} for all @var{regno} values in the class @var{class}.
2395 This macro helps control the handling of multiple-word values
2399 @hook TARGET_CAN_CHANGE_MODE_CLASS
2401 @hook TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS
2405 @hook TARGET_REGISTER_PRIORITY
2407 @hook TARGET_REGISTER_USAGE_LEVELING_P
2409 @hook TARGET_DIFFERENT_ADDR_DISPLACEMENT_P
2411 @hook TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P
2413 @hook TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT
2415 @hook TARGET_SPILL_CLASS
2417 @hook TARGET_ADDITIONAL_ALLOCNO_CLASS_P
2419 @hook TARGET_CSTORE_MODE
2421 @hook TARGET_COMPUTE_PRESSURE_CLASSES
2423 @node Stack and Calling
2424 @section Stack Layout and Calling Conventions
2425 @cindex calling conventions
2427 @c prevent bad page break with this line
2428 This describes the stack layout and calling conventions.
2432 * Exception Handling::
2437 * Register Arguments::
2439 * Aggregate Return::
2444 * Shrink-wrapping separate components::
2445 * Stack Smashing Protection::
2446 * Miscellaneous Register Hooks::
2450 @subsection Basic Stack Layout
2451 @cindex stack frame layout
2452 @cindex frame layout
2454 @c prevent bad page break with this line
2455 Here is the basic stack layout.
2457 @defmac STACK_GROWS_DOWNWARD
2458 Define this macro to be true if pushing a word onto the stack moves the stack
2459 pointer to a smaller address, and false otherwise.
2462 @defmac STACK_PUSH_CODE
2463 This macro defines the operation used when something is pushed
2464 on the stack. In RTL, a push operation will be
2465 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2467 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2468 and @code{POST_INC}. Which of these is correct depends on
2469 the stack direction and on whether the stack pointer points
2470 to the last item on the stack or whether it points to the
2471 space for the next item on the stack.
2473 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2474 true, which is almost always right, and @code{PRE_INC} otherwise,
2475 which is often wrong.
2478 @defmac FRAME_GROWS_DOWNWARD
2479 Define this macro to nonzero value if the addresses of local variable slots
2480 are at negative offsets from the frame pointer.
2483 @defmac ARGS_GROW_DOWNWARD
2484 Define this macro if successive arguments to a function occupy decreasing
2485 addresses on the stack.
2488 @hook TARGET_STARTING_FRAME_OFFSET
2490 @defmac STACK_ALIGNMENT_NEEDED
2491 Define to zero to disable final alignment of the stack during reload.
2492 The nonzero default for this macro is suitable for most ports.
2494 On ports where @code{TARGET_STARTING_FRAME_OFFSET} is nonzero or where there
2495 is a register save block following the local block that doesn't require
2496 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2497 stack alignment and do it in the backend.
2500 @defmac STACK_POINTER_OFFSET
2501 Offset from the stack pointer register to the first location at which
2502 outgoing arguments are placed. If not specified, the default value of
2503 zero is used. This is the proper value for most machines.
2505 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2506 the first location at which outgoing arguments are placed.
2509 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2510 Offset from the argument pointer register to the first argument's
2511 address. On some machines it may depend on the data type of the
2514 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2515 the first argument's address.
2518 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2519 Offset from the stack pointer register to an item dynamically allocated
2520 on the stack, e.g., by @code{alloca}.
2522 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2523 length of the outgoing arguments. The default is correct for most
2524 machines. See @file{function.cc} for details.
2527 @defmac INITIAL_FRAME_ADDRESS_RTX
2528 A C expression whose value is RTL representing the address of the initial
2529 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2530 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
2531 default value will be used. Define this macro in order to make frame pointer
2532 elimination work in the presence of @code{__builtin_frame_address (count)} and
2533 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
2536 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2537 A C expression whose value is RTL representing the address in a stack
2538 frame where the pointer to the caller's frame is stored. Assume that
2539 @var{frameaddr} is an RTL expression for the address of the stack frame
2542 If you don't define this macro, the default is to return the value
2543 of @var{frameaddr}---that is, the stack frame address is also the
2544 address of the stack word that points to the previous frame.
2547 @defmac SETUP_FRAME_ADDRESSES
2548 A C expression that produces the machine-specific code to
2549 setup the stack so that arbitrary frames can be accessed. For example,
2550 on the SPARC, we must flush all of the register windows to the stack
2551 before we can access arbitrary stack frames. You will seldom need to
2552 define this macro. The default is to do nothing.
2555 @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
2557 @defmac FRAME_ADDR_RTX (@var{frameaddr})
2558 A C expression whose value is RTL representing the value of the frame
2559 address for the current frame. @var{frameaddr} is the frame pointer
2560 of the current frame. This is used for __builtin_frame_address.
2561 You need only define this macro if the frame address is not the same
2562 as the frame pointer. Most machines do not need to define it.
2565 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2566 A C expression whose value is RTL representing the value of the return
2567 address for the frame @var{count} steps up from the current frame, after
2568 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2569 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2570 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero.
2572 The value of the expression must always be the correct address when
2573 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
2574 determine the return address of other frames.
2577 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2578 Define this macro to nonzero value if the return address of a particular
2579 stack frame is accessed from the frame pointer of the previous stack
2580 frame. The zero default for this macro is suitable for most ports.
2583 @defmac INCOMING_RETURN_ADDR_RTX
2584 A C expression whose value is RTL representing the location of the
2585 incoming return address at the beginning of any function, before the
2586 prologue. This RTL is either a @code{REG}, indicating that the return
2587 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2590 You only need to define this macro if you want to support call frame
2591 debugging information like that provided by DWARF 2.
2593 If this RTL is a @code{REG}, you should also define
2594 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2597 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2598 A C expression whose value is an integer giving a DWARF 2 column
2599 number that may be used as an alternative return column. The column
2600 must not correspond to any gcc hard register (that is, it must not
2601 be in the range of @code{DWARF_FRAME_REGNUM}).
2603 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2604 general register, but an alternative column needs to be used for signal
2605 frames. Some targets have also used different frame return columns
2609 @defmac DWARF_ZERO_REG
2610 A C expression whose value is an integer giving a DWARF 2 register
2611 number that is considered to always have the value zero. This should
2612 only be defined if the target has an architected zero register, and
2613 someone decided it was a good idea to use that register number to
2614 terminate the stack backtrace. New ports should avoid this.
2617 @defmac DWARF_VERSION_DEFAULT
2618 A C expression whose value is the default dwarf standard version we'll honor
2619 and advertise when generating dwarf debug information, in absence of
2620 an explicit @option{-gdwarf-@var{version}} option on the command line.
2623 @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
2625 @hook TARGET_DWARF_POLY_INDETERMINATE_VALUE
2627 @defmac INCOMING_FRAME_SP_OFFSET
2628 A C expression whose value is an integer giving the offset, in bytes,
2629 from the value of the stack pointer register to the top of the stack
2630 frame at the beginning of any function, before the prologue. The top of
2631 the frame is defined to be the value of the stack pointer in the
2632 previous frame, just before the call instruction.
2634 You only need to define this macro if you want to support call frame
2635 debugging information like that provided by DWARF 2.
2638 @defmac DEFAULT_INCOMING_FRAME_SP_OFFSET
2639 Like @code{INCOMING_FRAME_SP_OFFSET}, but must be the same for all
2640 functions of the same ABI, and when using GAS @code{.cfi_*} directives
2641 must also agree with the default CFI GAS emits. Define this macro
2642 only if @code{INCOMING_FRAME_SP_OFFSET} can have different values
2643 between different functions of the same ABI or when
2644 @code{INCOMING_FRAME_SP_OFFSET} does not agree with GAS default CFI.
2647 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2648 A C expression whose value is an integer giving the offset, in bytes,
2649 from the argument pointer to the canonical frame address (cfa). The
2650 final value should coincide with that calculated by
2651 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2652 during virtual register instantiation.
2654 The default value for this macro is
2655 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
2656 which is correct for most machines; in general, the arguments are found
2657 immediately before the stack frame. Note that this is not the case on
2658 some targets that save registers into the caller's frame, such as SPARC
2659 and rs6000, and so such targets need to define this macro.
2661 You only need to define this macro if the default is incorrect, and you
2662 want to support call frame debugging information like that provided by
2666 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
2667 If defined, a C expression whose value is an integer giving the offset
2668 in bytes from the frame pointer to the canonical frame address (cfa).
2669 The final value should coincide with that calculated by
2670 @code{INCOMING_FRAME_SP_OFFSET}.
2672 Normally the CFA is calculated as an offset from the argument pointer,
2673 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
2674 variable due to the ABI, this may not be possible. If this macro is
2675 defined, it implies that the virtual register instantiation should be
2676 based on the frame pointer instead of the argument pointer. Only one
2677 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
2681 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
2682 If defined, a C expression whose value is an integer giving the offset
2683 in bytes from the canonical frame address (cfa) to the frame base used
2684 in DWARF 2 debug information. The default is zero. A different value
2685 may reduce the size of debug information on some ports.
2688 @node Exception Handling
2689 @subsection Exception Handling Support
2690 @cindex exception handling
2692 @defmac EH_RETURN_DATA_REGNO (@var{N})
2693 A C expression whose value is the @var{N}th register number used for
2694 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2695 @var{N} registers are usable.
2697 The exception handling library routines communicate with the exception
2698 handlers via a set of agreed upon registers. Ideally these registers
2699 should be call-clobbered; it is possible to use call-saved registers,
2700 but may negatively impact code size. The target must support at least
2701 2 data registers, but should define 4 if there are enough free registers.
2703 You must define this macro if you want to support call frame exception
2704 handling like that provided by DWARF 2.
2707 @defmac EH_RETURN_STACKADJ_RTX
2708 A C expression whose value is RTL representing a location in which
2709 to store a stack adjustment to be applied before function return.
2710 This is used to unwind the stack to an exception handler's call frame.
2711 It will be assigned zero on code paths that return normally.
2713 Typically this is a call-clobbered hard register that is otherwise
2714 untouched by the epilogue, but could also be a stack slot.
2716 Do not define this macro if the stack pointer is saved and restored
2717 by the regular prolog and epilog code in the call frame itself; in
2718 this case, the exception handling library routines will update the
2719 stack location to be restored in place. Otherwise, you must define
2720 this macro if you want to support call frame exception handling like
2721 that provided by DWARF 2.
2724 @defmac EH_RETURN_HANDLER_RTX
2725 A C expression whose value is RTL representing a location in which
2726 to store the address of an exception handler to which we should
2727 return. It will not be assigned on code paths that return normally.
2729 Typically this is the location in the call frame at which the normal
2730 return address is stored. For targets that return by popping an
2731 address off the stack, this might be a memory address just below
2732 the @emph{target} call frame rather than inside the current call
2733 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2734 been assigned, so it may be used to calculate the location of the
2737 Some targets have more complex requirements than storing to an
2738 address calculable during initial code generation. In that case
2739 the @code{eh_return} instruction pattern should be used instead.
2741 If you want to support call frame exception handling, you must
2742 define either this macro or the @code{eh_return} instruction pattern.
2745 @defmac RETURN_ADDR_OFFSET
2746 If defined, an integer-valued C expression for which rtl will be generated
2747 to add it to the exception handler address before it is searched in the
2748 exception handling tables, and to subtract it again from the address before
2749 using it to return to the exception handler.
2752 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
2753 This macro chooses the encoding of pointers embedded in the exception
2754 handling sections. If at all possible, this should be defined such
2755 that the exception handling section will not require dynamic relocations,
2756 and so may be read-only.
2758 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2759 @var{global} is true if the symbol may be affected by dynamic relocations.
2760 The macro should return a combination of the @code{DW_EH_PE_*} defines
2761 as found in @file{dwarf2.h}.
2763 If this macro is not defined, pointers will not be encoded but
2764 represented directly.
2767 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2768 This macro allows the target to emit whatever special magic is required
2769 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2770 Generic code takes care of pc-relative and indirect encodings; this must
2771 be defined if the target uses text-relative or data-relative encodings.
2773 This is a C statement that branches to @var{done} if the format was
2774 handled. @var{encoding} is the format chosen, @var{size} is the number
2775 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2779 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
2780 This macro allows the target to add CPU and operating system specific
2781 code to the call-frame unwinder for use when there is no unwind data
2782 available. The most common reason to implement this macro is to unwind
2783 through signal frames.
2785 This macro is called from @code{uw_frame_state_for} in
2786 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
2787 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2788 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
2789 for the address of the code being executed and @code{context->cfa} for
2790 the stack pointer value. If the frame can be decoded, the register
2791 save addresses should be updated in @var{fs} and the macro should
2792 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
2793 the macro should evaluate to @code{_URC_END_OF_STACK}.
2795 For proper signal handling in Java this macro is accompanied by
2796 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
2799 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
2800 This macro allows the target to add operating system specific code to the
2801 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
2802 usually used for signal or interrupt frames.
2804 This macro is called from @code{uw_update_context} in libgcc's
2805 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2806 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
2807 for the abi and context in the @code{.unwabi} directive. If the
2808 @code{.unwabi} directive can be handled, the register save addresses should
2809 be updated in @var{fs}.
2812 @defmac TARGET_USES_WEAK_UNWIND_INFO
2813 A C expression that evaluates to true if the target requires unwind
2814 info to be given comdat linkage. Define it to be @code{1} if comdat
2815 linkage is necessary. The default is @code{0}.
2818 @node Stack Checking
2819 @subsection Specifying How Stack Checking is Done
2821 GCC will check that stack references are within the boundaries of the
2822 stack, if the option @option{-fstack-check} is specified, in one of
2827 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2828 will assume that you have arranged for full stack checking to be done
2829 at appropriate places in the configuration files. GCC will not do
2830 other special processing.
2833 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
2834 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
2835 that you have arranged for static stack checking (checking of the
2836 static stack frame of functions) to be done at appropriate places
2837 in the configuration files. GCC will only emit code to do dynamic
2838 stack checking (checking on dynamic stack allocations) using the third
2842 If neither of the above are true, GCC will generate code to periodically
2843 ``probe'' the stack pointer using the values of the macros defined below.
2846 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
2847 GCC will change its allocation strategy for large objects if the option
2848 @option{-fstack-check} is specified: they will always be allocated
2849 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
2851 @defmac STACK_CHECK_BUILTIN
2852 A nonzero value if stack checking is done by the configuration files in a
2853 machine-dependent manner. You should define this macro if stack checking
2854 is required by the ABI of your machine or if you would like to do stack
2855 checking in some more efficient way than the generic approach. The default
2856 value of this macro is zero.
2859 @defmac STACK_CHECK_STATIC_BUILTIN
2860 A nonzero value if static stack checking is done by the configuration files
2861 in a machine-dependent manner. You should define this macro if you would
2862 like to do static stack checking in some more efficient way than the generic
2863 approach. The default value of this macro is zero.
2866 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
2867 An integer specifying the interval at which GCC must generate stack probe
2868 instructions, defined as 2 raised to this integer. You will normally
2869 define this macro so that the interval be no larger than the size of
2870 the ``guard pages'' at the end of a stack area. The default value
2871 of 12 (4096-byte interval) is suitable for most systems.
2874 @defmac STACK_CHECK_MOVING_SP
2875 An integer which is nonzero if GCC should move the stack pointer page by page
2876 when doing probes. This can be necessary on systems where the stack pointer
2877 contains the bottom address of the memory area accessible to the executing
2878 thread at any point in time. In this situation an alternate signal stack
2879 is required in order to be able to recover from a stack overflow. The
2880 default value of this macro is zero.
2883 @defmac STACK_CHECK_PROTECT
2884 The number of bytes of stack needed to recover from a stack overflow, for
2885 languages where such a recovery is supported. The default value of 4KB/8KB
2886 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
2887 8KB/12KB with other exception handling mechanisms should be adequate for most
2888 architectures and operating systems.
2891 The following macros are relevant only if neither STACK_CHECK_BUILTIN
2892 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
2893 in the opposite case.
2895 @defmac STACK_CHECK_MAX_FRAME_SIZE
2896 The maximum size of a stack frame, in bytes. GCC will generate probe
2897 instructions in non-leaf functions to ensure at least this many bytes of
2898 stack are available. If a stack frame is larger than this size, stack
2899 checking will not be reliable and GCC will issue a warning. The
2900 default is chosen so that GCC only generates one instruction on most
2901 systems. You should normally not change the default value of this macro.
2904 @defmac STACK_CHECK_FIXED_FRAME_SIZE
2905 GCC uses this value to generate the above warning message. It
2906 represents the amount of fixed frame used by a function, not including
2907 space for any callee-saved registers, temporaries and user variables.
2908 You need only specify an upper bound for this amount and will normally
2909 use the default of four words.
2912 @defmac STACK_CHECK_MAX_VAR_SIZE
2913 The maximum size, in bytes, of an object that GCC will place in the
2914 fixed area of the stack frame when the user specifies
2915 @option{-fstack-check}.
2916 GCC computed the default from the values of the above macros and you will
2917 normally not need to override that default.
2920 @hook TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE
2923 @node Frame Registers
2924 @subsection Registers That Address the Stack Frame
2926 @c prevent bad page break with this line
2927 This discusses registers that address the stack frame.
2929 @defmac STACK_POINTER_REGNUM
2930 The register number of the stack pointer register, which must also be a
2931 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2932 the hardware determines which register this is.
2935 @defmac FRAME_POINTER_REGNUM
2936 The register number of the frame pointer register, which is used to
2937 access automatic variables in the stack frame. On some machines, the
2938 hardware determines which register this is. On other machines, you can
2939 choose any register you wish for this purpose.
2942 @defmac HARD_FRAME_POINTER_REGNUM
2943 On some machines the offset between the frame pointer and starting
2944 offset of the automatic variables is not known until after register
2945 allocation has been done (for example, because the saved registers are
2946 between these two locations). On those machines, define
2947 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2948 be used internally until the offset is known, and define
2949 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2950 used for the frame pointer.
2952 You should define this macro only in the very rare circumstances when it
2953 is not possible to calculate the offset between the frame pointer and
2954 the automatic variables until after register allocation has been
2955 completed. When this macro is defined, you must also indicate in your
2956 definition of @code{ELIMINABLE_REGS} how to eliminate
2957 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2958 or @code{STACK_POINTER_REGNUM}.
2960 Do not define this macro if it would be the same as
2961 @code{FRAME_POINTER_REGNUM}.
2964 @defmac ARG_POINTER_REGNUM
2965 The register number of the arg pointer register, which is used to access
2966 the function's argument list. On some machines, this is the same as the
2967 frame pointer register. On some machines, the hardware determines which
2968 register this is. On other machines, you can choose any register you
2969 wish for this purpose. If this is not the same register as the frame
2970 pointer register, then you must mark it as a fixed register according to
2971 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2972 (@pxref{Elimination}).
2975 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
2976 Define this to a preprocessor constant that is nonzero if
2977 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
2978 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
2979 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
2980 definition is not suitable for use in preprocessor conditionals.
2983 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
2984 Define this to a preprocessor constant that is nonzero if
2985 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
2986 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
2987 ARG_POINTER_REGNUM)}; you only need to define this macro if that
2988 definition is not suitable for use in preprocessor conditionals.
2991 @defmac RETURN_ADDRESS_POINTER_REGNUM
2992 The register number of the return address pointer register, which is used to
2993 access the current function's return address from the stack. On some
2994 machines, the return address is not at a fixed offset from the frame
2995 pointer or stack pointer or argument pointer. This register can be defined
2996 to point to the return address on the stack, and then be converted by
2997 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2999 Do not define this macro unless there is no other way to get the return
3000 address from the stack.
3003 @defmac STATIC_CHAIN_REGNUM
3004 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3005 Register numbers used for passing a function's static chain pointer. If
3006 register windows are used, the register number as seen by the called
3007 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3008 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3009 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3012 The static chain register need not be a fixed register.
3014 If the static chain is passed in memory, these macros should not be
3015 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3018 @hook TARGET_STATIC_CHAIN
3020 @defmac DWARF_FRAME_REGISTERS
3021 This macro specifies the maximum number of hard registers that can be
3022 saved in a call frame. This is used to size data structures used in
3023 DWARF2 exception handling.
3025 Prior to GCC 3.0, this macro was needed in order to establish a stable
3026 exception handling ABI in the face of adding new hard registers for ISA
3027 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3028 in the number of hard registers. Nevertheless, this macro can still be
3029 used to reduce the runtime memory requirements of the exception handling
3030 routines, which can be substantial if the ISA contains a lot of
3031 registers that are not call-saved.
3033 If this macro is not defined, it defaults to
3034 @code{FIRST_PSEUDO_REGISTER}.
3037 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3039 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3040 for backward compatibility in pre GCC 3.0 compiled code.
3042 If this macro is not defined, it defaults to
3043 @code{DWARF_FRAME_REGISTERS}.
3046 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3048 Define this macro if the target's representation for dwarf registers
3049 is different than the internal representation for unwind column.
3050 Given a dwarf register, this macro should return the internal unwind
3051 column number to use instead.
3054 @defmac DWARF_FRAME_REGNUM (@var{regno})
3056 Define this macro if the target's representation for dwarf registers
3057 used in .eh_frame or .debug_frame is different from that used in other
3058 debug info sections. Given a GCC hard register number, this macro
3059 should return the .eh_frame register number. The default is
3060 @code{DEBUGGER_REGNO (@var{regno})}.
3064 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3066 Define this macro to map register numbers held in the call frame info
3067 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3068 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3069 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3070 return @code{@var{regno}}.
3074 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3076 Define this macro if the target stores register values as
3077 @code{_Unwind_Word} type in unwind context. It should be defined if
3078 target register size is larger than the size of @code{void *}. The
3079 default is to store register values as @code{void *} type.
3083 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3085 Define this macro to be 1 if the target always uses extended unwind
3086 context with version, args_size and by_value fields. If it is undefined,
3087 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3088 defined and 0 otherwise.
3092 @defmac DWARF_LAZY_REGISTER_VALUE (@var{regno}, @var{value})
3093 Define this macro if the target has pseudo DWARF registers whose
3094 values need to be computed lazily on demand by the unwinder (such as when
3095 referenced in a CFA expression). The macro returns true if @var{regno}
3096 is such a register and stores its value in @samp{*@var{value}} if so.
3100 @subsection Eliminating Frame Pointer and Arg Pointer
3102 @c prevent bad page break with this line
3103 This is about eliminating the frame pointer and arg pointer.
3105 @hook TARGET_FRAME_POINTER_REQUIRED
3107 @defmac ELIMINABLE_REGS
3108 This macro specifies a table of register pairs used to eliminate
3109 unneeded registers that point into the stack frame.
3111 The definition of this macro is a list of structure initializations, each
3112 of which specifies an original and replacement register.
3114 On some machines, the position of the argument pointer is not known until
3115 the compilation is completed. In such a case, a separate hard register
3116 must be used for the argument pointer. This register can be eliminated by
3117 replacing it with either the frame pointer or the argument pointer,
3118 depending on whether or not the frame pointer has been eliminated.
3120 In this case, you might specify:
3122 #define ELIMINABLE_REGS \
3123 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3124 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3125 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3128 Note that the elimination of the argument pointer with the stack pointer is
3129 specified first since that is the preferred elimination.
3132 @hook TARGET_CAN_ELIMINATE
3134 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3135 This macro returns the initial difference between the specified pair
3136 of registers. The value would be computed from information
3137 such as the result of @code{get_frame_size ()} and the tables of
3138 registers @code{df_regs_ever_live_p} and @code{call_used_regs}.
3141 @hook TARGET_COMPUTE_FRAME_LAYOUT
3143 @node Stack Arguments
3144 @subsection Passing Function Arguments on the Stack
3145 @cindex arguments on stack
3146 @cindex stack arguments
3148 The macros in this section control how arguments are passed
3149 on the stack. See the following section for other macros that
3150 control passing certain arguments in registers.
3152 @hook TARGET_PROMOTE_PROTOTYPES
3154 @hook TARGET_PUSH_ARGUMENT
3156 @defmac PUSH_ARGS_REVERSED
3157 A C expression. If nonzero, function arguments will be evaluated from
3158 last to first, rather than from first to last. If this macro is not
3159 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3160 and args grow in opposite directions, and 0 otherwise.
3163 @defmac PUSH_ROUNDING (@var{npushed})
3164 A C expression that is the number of bytes actually pushed onto the
3165 stack when an instruction attempts to push @var{npushed} bytes.
3167 On some machines, the definition
3170 #define PUSH_ROUNDING(BYTES) (BYTES)
3174 will suffice. But on other machines, instructions that appear
3175 to push one byte actually push two bytes in an attempt to maintain
3176 alignment. Then the definition should be
3179 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3182 If the value of this macro has a type, it should be an unsigned type.
3185 @findex outgoing_args_size
3186 @findex crtl->outgoing_args_size
3187 @defmac ACCUMULATE_OUTGOING_ARGS
3188 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3189 will be computed and placed into
3190 @code{crtl->outgoing_args_size}. No space will be pushed
3191 onto the stack for each call; instead, the function prologue should
3192 increase the stack frame size by this amount.
3194 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3198 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3199 Define this macro if functions should assume that stack space has been
3200 allocated for arguments even when their values are passed in
3203 The value of this macro is the size, in bytes, of the area reserved for
3204 arguments passed in registers for the function represented by @var{fndecl},
3205 which can be zero if GCC is calling a library function.
3206 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3209 This space can be allocated by the caller, or be a part of the
3210 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3213 @c above is overfull. not sure what to do. --mew 5feb93 did
3214 @c something, not sure if it looks good. --mew 10feb93
3216 @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3217 Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3218 Define this macro if space guaranteed when compiling a function body
3219 is different to space required when making a call, a situation that
3220 can arise with K&R style function definitions.
3223 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3224 Define this to a nonzero value if it is the responsibility of the
3225 caller to allocate the area reserved for arguments passed in registers
3226 when calling a function of @var{fntype}. @var{fntype} may be NULL
3227 if the function called is a library function.
3229 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3230 whether the space for these arguments counts in the value of
3231 @code{crtl->outgoing_args_size}.
3234 @defmac STACK_PARMS_IN_REG_PARM_AREA
3235 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3236 stack parameters don't skip the area specified by it.
3237 @c i changed this, makes more sens and it should have taken care of the
3238 @c overfull.. not as specific, tho. --mew 5feb93
3240 Normally, when a parameter is not passed in registers, it is placed on the
3241 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3242 suppresses this behavior and causes the parameter to be passed on the
3243 stack in its natural location.
3246 @hook TARGET_RETURN_POPS_ARGS
3248 @defmac CALL_POPS_ARGS (@var{cum})
3249 A C expression that should indicate the number of bytes a call sequence
3250 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3251 when compiling a function call.
3253 @var{cum} is the variable in which all arguments to the called function
3254 have been accumulated.
3256 On certain architectures, such as the SH5, a call trampoline is used
3257 that pops certain registers off the stack, depending on the arguments
3258 that have been passed to the function. Since this is a property of the
3259 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3263 @node Register Arguments
3264 @subsection Passing Arguments in Registers
3265 @cindex arguments in registers
3266 @cindex registers arguments
3268 This section describes the macros which let you control how various
3269 types of arguments are passed in registers or how they are arranged in
3272 @hook TARGET_FUNCTION_ARG
3274 @hook TARGET_MUST_PASS_IN_STACK
3276 @hook TARGET_FUNCTION_INCOMING_ARG
3278 @hook TARGET_USE_PSEUDO_PIC_REG
3280 @hook TARGET_INIT_PIC_REG
3282 @hook TARGET_ARG_PARTIAL_BYTES
3284 @hook TARGET_PASS_BY_REFERENCE
3286 @hook TARGET_CALLEE_COPIES
3288 @defmac CUMULATIVE_ARGS
3289 A C type for declaring a variable that is used as the first argument
3290 of @code{TARGET_FUNCTION_ARG} and other related values. For some
3291 target machines, the type @code{int} suffices and can hold the number
3292 of bytes of argument so far.
3294 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3295 arguments that have been passed on the stack. The compiler has other
3296 variables to keep track of that. For target machines on which all
3297 arguments are passed on the stack, there is no need to store anything in
3298 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3299 should not be empty, so use @code{int}.
3302 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
3303 If defined, this macro is called before generating any code for a
3304 function, but after the @var{cfun} descriptor for the function has been
3305 created. The back end may use this macro to update @var{cfun} to
3306 reflect an ABI other than that which would normally be used by default.
3307 If the compiler is generating code for a compiler-generated function,
3308 @var{fndecl} may be @code{NULL}.
3311 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3312 A C statement (sans semicolon) for initializing the variable
3313 @var{cum} for the state at the beginning of the argument list. The
3314 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3315 is the tree node for the data type of the function which will receive
3316 the args, or 0 if the args are to a compiler support library function.
3317 For direct calls that are not libcalls, @var{fndecl} contain the
3318 declaration node of the function. @var{fndecl} is also set when
3319 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3320 being compiled. @var{n_named_args} is set to the number of named
3321 arguments, including a structure return address if it is passed as a
3322 parameter, when making a call. When processing incoming arguments,
3323 @var{n_named_args} is set to @minus{}1.
3325 When processing a call to a compiler support library function,
3326 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3327 contains the name of the function, as a string. @var{libname} is 0 when
3328 an ordinary C function call is being processed. Thus, each time this
3329 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3330 never both of them at once.
3333 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3334 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3335 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3336 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3337 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3338 0)} is used instead.
3341 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3342 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3343 finding the arguments for the function being compiled. If this macro is
3344 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3346 The value passed for @var{libname} is always 0, since library routines
3347 with special calling conventions are never compiled with GCC@. The
3348 argument @var{libname} exists for symmetry with
3349 @code{INIT_CUMULATIVE_ARGS}.
3350 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3351 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3354 @hook TARGET_FUNCTION_ARG_ADVANCE
3356 @hook TARGET_FUNCTION_ARG_OFFSET
3358 @hook TARGET_FUNCTION_ARG_PADDING
3360 @defmac PAD_VARARGS_DOWN
3361 If defined, a C expression which determines whether the default
3362 implementation of va_arg will attempt to pad down before reading the
3363 next argument, if that argument is smaller than its aligned space as
3364 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3365 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3368 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3369 Specify padding for the last element of a block move between registers and
3370 memory. @var{first} is nonzero if this is the only element. Defining this
3371 macro allows better control of register function parameters on big-endian
3372 machines, without using @code{PARALLEL} rtl. In particular,
3373 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3374 registers, as there is no longer a "wrong" part of a register; For example,
3375 a three byte aggregate may be passed in the high part of a register if so
3379 @hook TARGET_FUNCTION_ARG_BOUNDARY
3381 @hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY
3383 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3384 A C expression that is nonzero if @var{regno} is the number of a hard
3385 register in which function arguments are sometimes passed. This does
3386 @emph{not} include implicit arguments such as the static chain and
3387 the structure-value address. On many machines, no registers can be
3388 used for this purpose since all function arguments are pushed on the
3392 @hook TARGET_SPLIT_COMPLEX_ARG
3394 @hook TARGET_BUILD_BUILTIN_VA_LIST
3396 @hook TARGET_ENUM_VA_LIST_P
3398 @hook TARGET_FN_ABI_VA_LIST
3400 @hook TARGET_CANONICAL_VA_LIST_TYPE
3402 @hook TARGET_GIMPLIFY_VA_ARG_EXPR
3404 @hook TARGET_VALID_POINTER_MODE
3406 @hook TARGET_REF_MAY_ALIAS_ERRNO
3408 @hook TARGET_TRANSLATE_MODE_ATTRIBUTE
3410 @hook TARGET_SCALAR_MODE_SUPPORTED_P
3412 @hook TARGET_VECTOR_MODE_SUPPORTED_P
3414 @hook TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P
3416 @hook TARGET_COMPATIBLE_VECTOR_TYPES_P
3418 @hook TARGET_ARRAY_MODE
3420 @hook TARGET_ARRAY_MODE_SUPPORTED_P
3422 @hook TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P
3424 @hook TARGET_FLOATN_MODE
3426 @hook TARGET_FLOATN_BUILTIN_P
3428 @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
3431 @subsection How Scalar Function Values Are Returned
3432 @cindex return values in registers
3433 @cindex values, returned by functions
3434 @cindex scalars, returned as values
3436 This section discusses the macros that control returning scalars as
3437 values---values that can fit in registers.
3439 @hook TARGET_FUNCTION_VALUE
3441 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3442 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
3443 a new target instead.
3446 @defmac LIBCALL_VALUE (@var{mode})
3447 A C expression to create an RTX representing the place where a library
3448 function returns a value of mode @var{mode}.
3450 Note that ``library function'' in this context means a compiler
3451 support routine, used to perform arithmetic, whose name is known
3452 specially by the compiler and was not mentioned in the C code being
3456 @hook TARGET_LIBCALL_VALUE
3458 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3459 A C expression that is nonzero if @var{regno} is the number of a hard
3460 register in which the values of called function may come back.
3462 A register whose use for returning values is limited to serving as the
3463 second of a pair (for a value of type @code{double}, say) need not be
3464 recognized by this macro. So for most machines, this definition
3468 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3471 If the machine has register windows, so that the caller and the called
3472 function use different registers for the return value, this macro
3473 should recognize only the caller's register numbers.
3475 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
3476 for a new target instead.
3479 @hook TARGET_FUNCTION_VALUE_REGNO_P
3481 @defmac APPLY_RESULT_SIZE
3482 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3483 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3484 saving and restoring an arbitrary return value.
3487 @hook TARGET_OMIT_STRUCT_RETURN_REG
3489 @hook TARGET_RETURN_IN_MSB
3491 @node Aggregate Return
3492 @subsection How Large Values Are Returned
3493 @cindex aggregates as return values
3494 @cindex large return values
3495 @cindex returning aggregate values
3496 @cindex structure value address
3498 When a function value's mode is @code{BLKmode} (and in some other
3499 cases), the value is not returned according to
3500 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
3501 caller passes the address of a block of memory in which the value
3502 should be stored. This address is called the @dfn{structure value
3505 This section describes how to control returning structure values in
3508 @hook TARGET_RETURN_IN_MEMORY
3510 @defmac DEFAULT_PCC_STRUCT_RETURN
3511 Define this macro to be 1 if all structure and union return values must be
3512 in memory. Since this results in slower code, this should be defined
3513 only if needed for compatibility with other compilers or with an ABI@.
3514 If you define this macro to be 0, then the conventions used for structure
3515 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
3518 If not defined, this defaults to the value 1.
3521 @hook TARGET_STRUCT_VALUE_RTX
3523 @defmac PCC_STATIC_STRUCT_RETURN
3524 Define this macro if the usual system convention on the target machine
3525 for returning structures and unions is for the called function to return
3526 the address of a static variable containing the value.
3528 Do not define this if the usual system convention is for the caller to
3529 pass an address to the subroutine.
3531 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3532 nothing when you use @option{-freg-struct-return} mode.
3535 @hook TARGET_GET_RAW_RESULT_MODE
3537 @hook TARGET_GET_RAW_ARG_MODE
3539 @hook TARGET_EMPTY_RECORD_P
3541 @hook TARGET_WARN_PARAMETER_PASSING_ABI
3544 @subsection Caller-Saves Register Allocation
3546 If you enable it, GCC can save registers around function calls. This
3547 makes it possible to use call-clobbered registers to hold variables that
3548 must live across calls.
3550 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3551 A C expression specifying which mode is required for saving @var{nregs}
3552 of a pseudo-register in call-clobbered hard register @var{regno}. If
3553 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3554 returned. For most machines this macro need not be defined since GCC
3555 will select the smallest suitable mode.
3558 @node Function Entry
3559 @subsection Function Entry and Exit
3560 @cindex function entry and exit
3564 This section describes the macros that output function entry
3565 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3567 @hook TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY
3569 @hook TARGET_ASM_FUNCTION_PROLOGUE
3571 @hook TARGET_ASM_FUNCTION_END_PROLOGUE
3573 @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
3575 @hook TARGET_ASM_FUNCTION_EPILOGUE
3578 @findex pretend_args_size
3579 @findex crtl->args.pretend_args_size
3581 A region of @code{crtl->args.pretend_args_size} bytes of
3582 uninitialized space just underneath the first argument arriving on the
3583 stack. (This may not be at the very start of the allocated stack region
3584 if the calling sequence has pushed anything else since pushing the stack
3585 arguments. But usually, on such machines, nothing else has been pushed
3586 yet, because the function prologue itself does all the pushing.) This
3587 region is used on machines where an argument may be passed partly in
3588 registers and partly in memory, and, in some cases to support the
3589 features in @code{<stdarg.h>}.
3592 An area of memory used to save certain registers used by the function.
3593 The size of this area, which may also include space for such things as
3594 the return address and pointers to previous stack frames, is
3595 machine-specific and usually depends on which registers have been used
3596 in the function. Machines with register windows often do not require
3600 A region of at least @var{size} bytes, possibly rounded up to an allocation
3601 boundary, to contain the local variables of the function. On some machines,
3602 this region and the save area may occur in the opposite order, with the
3603 save area closer to the top of the stack.
3605 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3607 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3608 @code{crtl->outgoing_args_size} bytes to be used for outgoing
3609 argument lists of the function. @xref{Stack Arguments}.
3612 @defmac EXIT_IGNORE_STACK
3613 Define this macro as a C expression that is nonzero if the return
3614 instruction or the function epilogue ignores the value of the stack
3615 pointer; in other words, if it is safe to delete an instruction to
3616 adjust the stack pointer before a return from the function. The
3619 Note that this macro's value is relevant only for functions for which
3620 frame pointers are maintained. It is never safe to delete a final
3621 stack adjustment in a function that has no frame pointer, and the
3622 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3625 @defmac EPILOGUE_USES (@var{regno})
3626 Define this macro as a C expression that is nonzero for registers that are
3627 used by the epilogue or the @samp{return} pattern. The stack and frame
3628 pointer registers are already assumed to be used as needed.
3631 @defmac EH_USES (@var{regno})
3632 Define this macro as a C expression that is nonzero for registers that are
3633 used by the exception handling mechanism, and so should be considered live
3634 on entry to an exception edge.
3637 @hook TARGET_ASM_OUTPUT_MI_THUNK
3639 @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
3642 @subsection Generating Code for Profiling
3643 @cindex profiling, code generation
3645 These macros will help you generate code for profiling.
3647 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
3648 A C statement or compound statement to output to @var{file} some
3649 assembler code to call the profiling subroutine @code{mcount}.
3652 The details of how @code{mcount} expects to be called are determined by
3653 your operating system environment, not by GCC@. To figure them out,
3654 compile a small program for profiling using the system's installed C
3655 compiler and look at the assembler code that results.
3657 Older implementations of @code{mcount} expect the address of a counter
3658 variable to be loaded into some register. The name of this variable is
3659 @samp{LP} followed by the number @var{labelno}, so you would generate
3660 the name using @samp{LP%d} in a @code{fprintf}.
3663 @defmac PROFILE_HOOK
3664 A C statement or compound statement to output to @var{file} some assembly
3665 code to call the profiling subroutine @code{mcount} even the target does
3666 not support profiling.
3669 @defmac NO_PROFILE_COUNTERS
3670 Define this macro to be an expression with a nonzero value if the
3671 @code{mcount} subroutine on your system does not need a counter variable
3672 allocated for each function. This is true for almost all modern
3673 implementations. If you define this macro, you must not use the
3674 @var{labelno} argument to @code{FUNCTION_PROFILER}.
3677 @defmac PROFILE_BEFORE_PROLOGUE
3678 Define this macro if the code for function profiling should come before
3679 the function prologue. Normally, the profiling code comes after.
3682 @hook TARGET_KEEP_LEAF_WHEN_PROFILED
3685 @subsection Permitting tail calls
3688 @hook TARGET_FUNCTION_OK_FOR_SIBCALL
3690 @hook TARGET_EXTRA_LIVE_ON_ENTRY
3692 @hook TARGET_SET_UP_BY_PROLOGUE
3694 @hook TARGET_WARN_FUNC_RETURN
3696 @node Shrink-wrapping separate components
3697 @subsection Shrink-wrapping separate components
3698 @cindex shrink-wrapping separate components
3700 The prologue may perform a variety of target dependent tasks such as
3701 saving callee-saved registers, saving the return address, aligning the
3702 stack, creating a stack frame, initializing the PIC register, setting
3703 up the static chain, etc.
3705 On some targets some of these tasks may be independent of others and
3706 thus may be shrink-wrapped separately. These independent tasks are
3707 referred to as components and are handled generically by the target
3708 independent parts of GCC.
3710 Using the following hooks those prologue or epilogue components can be
3711 shrink-wrapped separately, so that the initialization (and possibly
3712 teardown) those components do is not done as frequently on execution
3713 paths where this would unnecessary.
3715 What exactly those components are is up to the target code; the generic
3716 code treats them abstractly, as a bit in an @code{sbitmap}. These
3717 @code{sbitmap}s are allocated by the @code{shrink_wrap.get_separate_components}
3718 and @code{shrink_wrap.components_for_bb} hooks, and deallocated by the
3721 @hook TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS
3723 @hook TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB
3725 @hook TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS
3727 @hook TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS
3729 @hook TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS
3731 @hook TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS
3733 @node Stack Smashing Protection
3734 @subsection Stack smashing protection
3735 @cindex stack smashing protection
3737 @hook TARGET_STACK_PROTECT_GUARD
3739 @hook TARGET_STACK_PROTECT_FAIL
3741 @hook TARGET_STACK_PROTECT_RUNTIME_ENABLED_P
3743 @hook TARGET_SUPPORTS_SPLIT_STACK
3745 @hook TARGET_GET_VALID_OPTION_VALUES
3747 @node Miscellaneous Register Hooks
3748 @subsection Miscellaneous register hooks
3749 @cindex miscellaneous register hooks
3751 @hook TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
3754 @section Implementing the Varargs Macros
3755 @cindex varargs implementation
3757 GCC comes with an implementation of @code{<varargs.h>} and
3758 @code{<stdarg.h>} that work without change on machines that pass arguments
3759 on the stack. Other machines require their own implementations of
3760 varargs, and the two machine independent header files must have
3761 conditionals to include it.
3763 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
3764 the calling convention for @code{va_start}. The traditional
3765 implementation takes just one argument, which is the variable in which
3766 to store the argument pointer. The ISO implementation of
3767 @code{va_start} takes an additional second argument. The user is
3768 supposed to write the last named argument of the function here.
3770 However, @code{va_start} should not use this argument. The way to find
3771 the end of the named arguments is with the built-in functions described
3774 @defmac __builtin_saveregs ()
3775 Use this built-in function to save the argument registers in memory so
3776 that the varargs mechanism can access them. Both ISO and traditional
3777 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3778 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
3780 On some machines, @code{__builtin_saveregs} is open-coded under the
3781 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
3782 other machines, it calls a routine written in assembler language,
3783 found in @file{libgcc2.c}.
3785 Code generated for the call to @code{__builtin_saveregs} appears at the
3786 beginning of the function, as opposed to where the call to
3787 @code{__builtin_saveregs} is written, regardless of what the code is.
3788 This is because the registers must be saved before the function starts
3789 to use them for its own purposes.
3790 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3794 @defmac __builtin_next_arg (@var{lastarg})
3795 This builtin returns the address of the first anonymous stack
3796 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3797 returns the address of the location above the first anonymous stack
3798 argument. Use it in @code{va_start} to initialize the pointer for
3799 fetching arguments from the stack. Also use it in @code{va_start} to
3800 verify that the second parameter @var{lastarg} is the last named argument
3801 of the current function.
3804 @defmac __builtin_classify_type (@var{object})
3805 Since each machine has its own conventions for which data types are
3806 passed in which kind of register, your implementation of @code{va_arg}
3807 has to embody these conventions. The easiest way to categorize the
3808 specified data type is to use @code{__builtin_classify_type} together
3809 with @code{sizeof} and @code{__alignof__}.
3811 @code{__builtin_classify_type} ignores the value of @var{object},
3812 considering only its data type. It returns an integer describing what
3813 kind of type that is---integer, floating, pointer, structure, and so on.
3815 The file @file{typeclass.h} defines an enumeration that you can use to
3816 interpret the values of @code{__builtin_classify_type}.
3819 These machine description macros help implement varargs:
3821 @hook TARGET_EXPAND_BUILTIN_SAVEREGS
3823 @hook TARGET_SETUP_INCOMING_VARARGS
3825 @hook TARGET_STRICT_ARGUMENT_NAMING
3827 @hook TARGET_CALL_ARGS
3829 @hook TARGET_END_CALL_ARGS
3831 @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
3834 @section Support for Nested Functions
3835 @cindex support for nested functions
3836 @cindex trampolines for nested functions
3837 @cindex descriptors for nested functions
3838 @cindex nested functions, support for
3840 Taking the address of a nested function requires special compiler
3841 handling to ensure that the static chain register is loaded when
3842 the function is invoked via an indirect call.
3844 GCC has traditionally supported nested functions by creating an
3845 executable @dfn{trampoline} at run time when the address of a nested
3846 function is taken. This is a small piece of code which normally
3847 resides on the stack, in the stack frame of the containing function.
3848 The trampoline loads the static chain register and then jumps to the
3849 real address of the nested function.
3851 The use of trampolines requires an executable stack, which is a
3852 security risk. To avoid this problem, GCC also supports another
3853 strategy: using descriptors for nested functions. Under this model,
3854 taking the address of a nested function results in a pointer to a
3855 non-executable function descriptor object. Initializing the static chain
3856 from the descriptor is handled at indirect call sites.
3858 On some targets, including HPPA and IA-64, function descriptors may be
3859 mandated by the ABI or be otherwise handled in a target-specific way
3860 by the back end in its code generation strategy for indirect calls.
3861 GCC also provides its own generic descriptor implementation to support the
3862 @option{-fno-trampolines} option. In this case runtime detection of
3863 function descriptors at indirect call sites relies on descriptor
3864 pointers being tagged with a bit that is never set in bare function
3865 addresses. Since GCC's generic function descriptors are
3866 not ABI-compliant, this option is typically used only on a
3867 per-language basis (notably by Ada) or when it can otherwise be
3868 applied to the whole program.
3870 For languages other than Ada, the @code{-ftrampolines} and
3871 @code{-fno-trampolines} options currently have no effect, and
3872 trampolines are always generated on platforms that need them
3873 for nested functions.
3875 Define the following hook if your backend either implements ABI-specified
3876 descriptor support, or can use GCC's generic descriptor implementation
3877 for nested functions.
3879 @hook TARGET_CUSTOM_FUNCTION_DESCRIPTORS
3881 The following macros tell GCC how to generate code to allocate and
3882 initialize an executable trampoline. You can also use this interface
3883 if your back end needs to create ABI-specified non-executable descriptors; in
3884 this case the "trampoline" created is the descriptor containing data only.
3886 The instructions in an executable trampoline must do two things: load
3887 a constant address into the static chain register, and jump to the real
3888 address of the nested function. On CISC machines such as the m68k,
3889 this requires two instructions, a move immediate and a jump. Then the
3890 two addresses exist in the trampoline as word-long immediate operands.
3891 On RISC machines, it is often necessary to load each address into a
3892 register in two parts. Then pieces of each address form separate
3895 The code generated to initialize the trampoline must store the variable
3896 parts---the static chain value and the function address---into the
3897 immediate operands of the instructions. On a CISC machine, this is
3898 simply a matter of copying each address to a memory reference at the
3899 proper offset from the start of the trampoline. On a RISC machine, it
3900 may be necessary to take out pieces of the address and store them
3903 @hook TARGET_ASM_TRAMPOLINE_TEMPLATE
3905 @defmac TRAMPOLINE_SECTION
3906 Return the section into which the trampoline template is to be placed
3907 (@pxref{Sections}). The default value is @code{readonly_data_section}.
3910 @defmac TRAMPOLINE_SIZE
3911 A C expression for the size in bytes of the trampoline, as an integer.
3914 @defmac TRAMPOLINE_ALIGNMENT
3915 Alignment required for trampolines, in bits.
3917 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
3918 is used for aligning trampolines.
3921 @hook TARGET_TRAMPOLINE_INIT
3923 @hook TARGET_EMIT_CALL_BUILTIN___CLEAR_CACHE
3925 @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
3927 Implementing trampolines is difficult on many machines because they have
3928 separate instruction and data caches. Writing into a stack location
3929 fails to clear the memory in the instruction cache, so when the program
3930 jumps to that location, it executes the old contents.
3932 Here are two possible solutions. One is to clear the relevant parts of
3933 the instruction cache whenever a trampoline is set up. The other is to
3934 make all trampolines identical, by having them jump to a standard
3935 subroutine. The former technique makes trampoline execution faster; the
3936 latter makes initialization faster.
3938 To clear the instruction cache when a trampoline is initialized, define
3939 the following macro.
3941 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
3942 If defined, expands to a C expression clearing the @emph{instruction
3943 cache} in the specified interval. The definition of this macro would
3944 typically be a series of @code{asm} statements. Both @var{beg} and
3945 @var{end} are pointer expressions.
3948 To use a standard subroutine, define the following macro. In addition,
3949 you must make sure that the instructions in a trampoline fill an entire
3950 cache line with identical instructions, or else ensure that the
3951 beginning of the trampoline code is always aligned at the same point in
3952 its cache line. Look in @file{m68k.h} as a guide.
3954 @defmac TRANSFER_FROM_TRAMPOLINE
3955 Define this macro if trampolines need a special subroutine to do their
3956 work. The macro should expand to a series of @code{asm} statements
3957 which will be compiled with GCC@. They go in a library function named
3958 @code{__transfer_from_trampoline}.
3960 If you need to avoid executing the ordinary prologue code of a compiled
3961 C function when you jump to the subroutine, you can do so by placing a
3962 special label of your own in the assembler code. Use one @code{asm}
3963 statement to generate an assembler label, and another to make the label
3964 global. Then trampolines can use that label to jump directly to your
3965 special assembler code.
3969 @section Implicit Calls to Library Routines
3970 @cindex library subroutine names
3971 @cindex @file{libgcc.a}
3973 @c prevent bad page break with this line
3974 Here is an explanation of implicit calls to library routines.
3976 @defmac DECLARE_LIBRARY_RENAMES
3977 This macro, if defined, should expand to a piece of C code that will get
3978 expanded when compiling functions for libgcc.a. It can be used to
3979 provide alternate names for GCC's internal library functions if there
3980 are ABI-mandated names that the compiler should provide.
3983 @findex set_optab_libfunc
3984 @findex init_one_libfunc
3985 @hook TARGET_INIT_LIBFUNCS
3987 @hook TARGET_LIBFUNC_GNU_PREFIX
3989 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
3990 This macro should return @code{true} if the library routine that
3991 implements the floating point comparison operator @var{comparison} in
3992 mode @var{mode} will return a boolean, and @var{false} if it will
3995 GCC's own floating point libraries return tristates from the
3996 comparison operators, so the default returns false always. Most ports
3997 don't need to define this macro.
4000 @defmac TARGET_LIB_INT_CMP_BIASED
4001 This macro should evaluate to @code{true} if the integer comparison
4002 functions (like @code{__cmpdi2}) return 0 to indicate that the first
4003 operand is smaller than the second, 1 to indicate that they are equal,
4004 and 2 to indicate that the first operand is greater than the second.
4005 If this macro evaluates to @code{false} the comparison functions return
4006 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
4007 in @file{libgcc.a}, you do not need to define this macro.
4010 @defmac TARGET_HAS_NO_HW_DIVIDE
4011 This macro should be defined if the target has no hardware divide
4012 instructions. If this macro is defined, GCC will use an algorithm which
4013 make use of simple logical and arithmetic operations for 64-bit
4014 division. If the macro is not defined, GCC will use an algorithm which
4015 make use of a 64-bit by 32-bit divide primitive.
4018 @cindex @code{EDOM}, implicit usage
4021 The value of @code{EDOM} on the target machine, as a C integer constant
4022 expression. If you don't define this macro, GCC does not attempt to
4023 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4024 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4027 If you do not define @code{TARGET_EDOM}, then compiled code reports
4028 domain errors by calling the library function and letting it report the
4029 error. If mathematical functions on your system use @code{matherr} when
4030 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4031 that @code{matherr} is used normally.
4034 @cindex @code{errno}, implicit usage
4035 @defmac GEN_ERRNO_RTX
4036 Define this macro as a C expression to create an rtl expression that
4037 refers to the global ``variable'' @code{errno}. (On certain systems,
4038 @code{errno} may not actually be a variable.) If you don't define this
4039 macro, a reasonable default is used.
4042 @hook TARGET_LIBC_HAS_FUNCTION
4044 @hook TARGET_LIBC_HAS_FAST_FUNCTION
4046 @hook TARGET_LIBM_FUNCTION_MAX_ERROR
4048 @defmac NEXT_OBJC_RUNTIME
4049 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
4050 by default. This calling convention involves passing the object, the selector
4051 and the method arguments all at once to the method-lookup library function.
4052 This is the usual setting when targeting Darwin / macOS systems, which have
4053 the NeXT runtime installed.
4055 If the macro is set to 0, the "GNU" Objective-C message sending convention
4056 will be used by default. This convention passes just the object and the
4057 selector to the method-lookup function, which returns a pointer to the method.
4059 In either case, it remains possible to select code-generation for the alternate
4060 scheme, by means of compiler command line switches.
4063 @node Addressing Modes
4064 @section Addressing Modes
4065 @cindex addressing modes
4067 @c prevent bad page break with this line
4068 This is about addressing modes.
4070 @defmac HAVE_PRE_INCREMENT
4071 @defmacx HAVE_PRE_DECREMENT
4072 @defmacx HAVE_POST_INCREMENT
4073 @defmacx HAVE_POST_DECREMENT
4074 A C expression that is nonzero if the machine supports pre-increment,
4075 pre-decrement, post-increment, or post-decrement addressing respectively.
4078 @defmac HAVE_PRE_MODIFY_DISP
4079 @defmacx HAVE_POST_MODIFY_DISP
4080 A C expression that is nonzero if the machine supports pre- or
4081 post-address side-effect generation involving constants other than
4082 the size of the memory operand.
4085 @defmac HAVE_PRE_MODIFY_REG
4086 @defmacx HAVE_POST_MODIFY_REG
4087 A C expression that is nonzero if the machine supports pre- or
4088 post-address side-effect generation involving a register displacement.
4091 @defmac CONSTANT_ADDRESS_P (@var{x})
4092 A C expression that is 1 if the RTX @var{x} is a constant which
4093 is a valid address. On most machines the default definition of
4094 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
4095 is acceptable, but a few machines are more restrictive as to which
4096 constant addresses are supported.
4099 @defmac CONSTANT_P (@var{x})
4100 @code{CONSTANT_P}, which is defined by target-independent code,
4101 accepts integer-values expressions whose values are not explicitly
4102 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4103 expressions and @code{const} arithmetic expressions, in addition to
4104 @code{const_int} and @code{const_double} expressions.
4107 @defmac MAX_REGS_PER_ADDRESS
4108 A number, the maximum number of registers that can appear in a valid
4109 memory address. Note that it is up to you to specify a value equal to
4110 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
4114 @hook TARGET_LEGITIMATE_ADDRESS_P
4116 @defmac TARGET_MEM_CONSTRAINT
4117 A single character to be used instead of the default @code{'m'}
4118 character for general memory addresses. This defines the constraint
4119 letter which matches the memory addresses accepted by
4120 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
4121 support new address formats in your back end without changing the
4122 semantics of the @code{'m'} constraint. This is necessary in order to
4123 preserve functionality of inline assembly constructs using the
4124 @code{'m'} constraint.
4127 @defmac FIND_BASE_TERM (@var{x})
4128 A C expression to determine the base term of address @var{x},
4129 or to provide a simplified version of @var{x} from which @file{alias.cc}
4130 can easily find the base term. This macro is used in only two places:
4131 @code{find_base_value} and @code{find_base_term} in @file{alias.cc}.
4133 It is always safe for this macro to not be defined. It exists so
4134 that alias analysis can understand machine-dependent addresses.
4136 The typical use of this macro is to handle addresses containing
4137 a label_ref or symbol_ref within an UNSPEC@.
4140 @hook TARGET_LEGITIMIZE_ADDRESS
4142 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4143 A C compound statement that attempts to replace @var{x}, which is an address
4144 that needs reloading, with a valid memory address for an operand of mode
4145 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4146 It is not necessary to define this macro, but it might be useful for
4147 performance reasons.
4149 For example, on the i386, it is sometimes possible to use a single
4150 reload register instead of two by reloading a sum of two pseudo
4151 registers into a register. On the other hand, for number of RISC
4152 processors offsets are limited so that often an intermediate address
4153 needs to be generated in order to address a stack slot. By defining
4154 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4155 generated for adjacent some stack slots can be made identical, and thus
4158 @emph{Note}: This macro should be used with caution. It is necessary
4159 to know something of how reload works in order to effectively use this,
4160 and it is quite easy to produce macros that build in too much knowledge
4161 of reload internals.
4163 @emph{Note}: This macro must be able to reload an address created by a
4164 previous invocation of this macro. If it fails to handle such addresses
4165 then the compiler may generate incorrect code or abort.
4168 The macro definition should use @code{push_reload} to indicate parts that
4169 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4170 suitable to be passed unaltered to @code{push_reload}.
4172 The code generated by this macro must not alter the substructure of
4173 @var{x}. If it transforms @var{x} into a more legitimate form, it
4174 should assign @var{x} (which will always be a C variable) a new value.
4175 This also applies to parts that you change indirectly by calling
4178 @findex strict_memory_address_p
4179 The macro definition may use @code{strict_memory_address_p} to test if
4180 the address has become legitimate.
4183 If you want to change only a part of @var{x}, one standard way of doing
4184 this is to use @code{copy_rtx}. Note, however, that it unshares only a
4185 single level of rtl. Thus, if the part to be changed is not at the
4186 top level, you'll need to replace first the top level.
4187 It is not necessary for this macro to come up with a legitimate
4188 address; but often a machine-dependent strategy can generate better code.
4191 @hook TARGET_MODE_DEPENDENT_ADDRESS_P
4193 @hook TARGET_LEGITIMATE_CONSTANT_P
4195 @hook TARGET_PRECOMPUTE_TLS_P
4197 @hook TARGET_DELEGITIMIZE_ADDRESS
4199 @hook TARGET_CONST_NOT_OK_FOR_DEBUG_P
4201 @hook TARGET_CANNOT_FORCE_CONST_MEM
4203 @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
4205 @hook TARGET_USE_BLOCKS_FOR_DECL_P
4207 @hook TARGET_BUILTIN_RECIPROCAL
4209 @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
4211 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
4213 @hook TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT
4215 @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
4217 @hook TARGET_VECTORIZE_VEC_PERM_CONST
4219 @hook TARGET_VECTORIZE_PREFERRED_DIV_AS_SHIFTS_OVER_MULT
4221 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
4223 @hook TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION
4225 @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
4227 @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
4229 @hook TARGET_VECTORIZE_SPLIT_REDUCTION
4231 @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_MODES
4233 @hook TARGET_VECTORIZE_RELATED_MODE
4235 @hook TARGET_VECTORIZE_GET_MASK_MODE
4237 @hook TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE
4239 @hook TARGET_VECTORIZE_CREATE_COSTS
4241 @hook TARGET_VECTORIZE_BUILTIN_GATHER
4243 @hook TARGET_VECTORIZE_BUILTIN_SCATTER
4245 @hook TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN
4247 @hook TARGET_SIMD_CLONE_ADJUST
4249 @hook TARGET_SIMD_CLONE_USABLE
4251 @hook TARGET_SIMT_VF
4253 @hook TARGET_OMP_DEVICE_KIND_ARCH_ISA
4255 @hook TARGET_GOACC_VALIDATE_DIMS
4257 @hook TARGET_GOACC_DIM_LIMIT
4259 @hook TARGET_GOACC_FORK_JOIN
4261 @hook TARGET_GOACC_REDUCTION
4263 @hook TARGET_PREFERRED_ELSE_VALUE
4265 @hook TARGET_GOACC_ADJUST_PRIVATE_DECL
4267 @hook TARGET_GOACC_EXPAND_VAR_DECL
4269 @hook TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD
4271 @hook TARGET_GOACC_SHARED_MEM_LAYOUT
4273 @node Anchored Addresses
4274 @section Anchored Addresses
4275 @cindex anchored addresses
4276 @cindex @option{-fsection-anchors}
4278 GCC usually addresses every static object as a separate entity.
4279 For example, if we have:
4283 int foo (void) @{ return a + b + c; @}
4286 the code for @code{foo} will usually calculate three separate symbolic
4287 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
4288 it would be better to calculate just one symbolic address and access
4289 the three variables relative to it. The equivalent pseudocode would
4295 register int *xr = &x;
4296 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
4300 (which isn't valid C). We refer to shared addresses like @code{x} as
4301 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
4303 The hooks below describe the target properties that GCC needs to know
4304 in order to make effective use of section anchors. It won't use
4305 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
4306 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
4308 @hook TARGET_MIN_ANCHOR_OFFSET
4310 @hook TARGET_MAX_ANCHOR_OFFSET
4312 @hook TARGET_ASM_OUTPUT_ANCHOR
4314 @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
4316 @node Condition Code
4317 @section Condition Code Status
4318 @cindex condition code status
4320 Condition codes in GCC are represented as registers,
4321 which provides better schedulability for
4322 architectures that do have a condition code register, but on which
4323 most instructions do not affect it. The latter category includes
4326 Implicit clobbering would pose a strong restriction on the placement of
4327 the definition and use of the condition code. In the past the definition
4328 and use were always adjacent. However, recent changes to support trapping
4329 arithmetic may result in the definition and user being in different blocks.
4330 Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally,
4331 the definition may be the source of exception handling edges.
4333 These restrictions can prevent important
4334 optimizations on some machines. For example, on the IBM RS/6000, there
4335 is a delay for taken branches unless the condition code register is set
4336 three instructions earlier than the conditional branch. The instruction
4337 scheduler cannot perform this optimization if it is not permitted to
4338 separate the definition and use of the condition code register.
4340 If there is a specific
4341 condition code register in the machine, use a hard register. If the
4342 condition code or comparison result can be placed in any general register,
4343 or if there are multiple condition registers, use a pseudo register.
4344 Registers used to store the condition code value will usually have a mode
4345 that is in class @code{MODE_CC}.
4347 Alternatively, you can use @code{BImode} if the comparison operator is
4348 specified already in the compare instruction. In this case, you are not
4349 interested in most macros in this section.
4352 * MODE_CC Condition Codes:: Modern representation of condition codes.
4355 @node MODE_CC Condition Codes
4356 @subsection Representation of condition codes using registers
4360 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4361 On many machines, the condition code may be produced by other instructions
4362 than compares, for example the branch can use directly the condition
4363 code set by a subtract instruction. However, on some machines
4364 when the condition code is set this way some bits (such as the overflow
4365 bit) are not set in the same way as a test instruction, so that a different
4366 branch instruction must be used for some conditional branches. When
4367 this happens, use the machine mode of the condition code register to
4368 record different formats of the condition code register. Modes can
4369 also be used to record which compare instruction (e.g.@: a signed or an
4370 unsigned comparison) produced the condition codes.
4372 If other modes than @code{CCmode} are required, add them to
4373 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
4374 a mode given an operand of a compare. This is needed because the modes
4375 have to be chosen not only during RTL generation but also, for example,
4376 by instruction combination. The result of @code{SELECT_CC_MODE} should
4377 be consistent with the mode used in the patterns; for example to support
4378 the case of the add on the SPARC discussed above, we have the pattern
4384 (plus:SI (match_operand:SI 0 "register_operand" "%r")
4385 (match_operand:SI 1 "arith_operand" "rI"))
4392 together with a @code{SELECT_CC_MODE} that returns @code{CCNZmode}
4393 for comparisons whose argument is a @code{plus}:
4396 #define SELECT_CC_MODE(OP,X,Y) \
4397 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4398 ? ((OP == LT || OP == LE || OP == GT || OP == GE) \
4399 ? CCFPEmode : CCFPmode) \
4400 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4401 || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \
4402 ? CCNZmode : CCmode))
4405 Another reason to use modes is to retain information on which operands
4406 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
4409 You should define this macro if and only if you define extra CC modes
4410 in @file{@var{machine}-modes.def}.
4413 @hook TARGET_CANONICALIZE_COMPARISON
4415 @defmac REVERSIBLE_CC_MODE (@var{mode})
4416 A C expression whose value is one if it is always safe to reverse a
4417 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4418 can ever return @var{mode} for a floating-point inequality comparison,
4419 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4421 You need not define this macro if it would always returns zero or if the
4422 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4423 For example, here is the definition used on the SPARC, where floating-point
4424 inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}:
4427 #define REVERSIBLE_CC_MODE(MODE) \
4428 ((MODE) != CCFPEmode && (MODE) != CCFPmode)
4432 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
4433 A C expression whose value is reversed condition code of the @var{code} for
4434 comparison done in CC_MODE @var{mode}. The macro is used only in case
4435 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
4436 machine has some non-standard way how to reverse certain conditionals. For
4437 instance in case all floating point conditions are non-trapping, compiler may
4438 freely convert unordered compares to ordered ones. Then definition may look
4442 #define REVERSE_CONDITION(CODE, MODE) \
4443 ((MODE) != CCFPmode ? reverse_condition (CODE) \
4444 : reverse_condition_maybe_unordered (CODE))
4448 @hook TARGET_FIXED_CONDITION_CODE_REGS
4450 @hook TARGET_CC_MODES_COMPATIBLE
4452 @hook TARGET_FLAGS_REGNUM
4455 @section Describing Relative Costs of Operations
4456 @cindex costs of instructions
4457 @cindex relative costs
4458 @cindex speed of instructions
4460 These macros let you describe the relative speed of various operations
4461 on the target machine.
4463 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
4464 A C expression for the cost of moving data of mode @var{mode} from a
4465 register in class @var{from} to one in class @var{to}. The classes are
4466 expressed using the enumeration values such as @code{GENERAL_REGS}. A
4467 value of 2 is the default; other values are interpreted relative to
4470 It is not required that the cost always equal 2 when @var{from} is the
4471 same as @var{to}; on some machines it is expensive to move between
4472 registers if they are not general registers.
4474 If reload sees an insn consisting of a single @code{set} between two
4475 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4476 classes returns a value of 2, reload does not check to ensure that the
4477 constraints of the insn are met. Setting a cost of other than 2 will
4478 allow reload to verify that the constraints are met. You should do this
4479 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4481 These macros are obsolete, new ports should use the target hook
4482 @code{TARGET_REGISTER_MOVE_COST} instead.
4485 @hook TARGET_REGISTER_MOVE_COST
4487 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4488 A C expression for the cost of moving data of mode @var{mode} between a
4489 register of class @var{class} and memory; @var{in} is zero if the value
4490 is to be written to memory, nonzero if it is to be read in. This cost
4491 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4492 registers and memory is more expensive than between two registers, you
4493 should define this macro to express the relative cost.
4495 If you do not define this macro, GCC uses a default cost of 4 plus
4496 the cost of copying via a secondary reload register, if one is
4497 needed. If your machine requires a secondary reload register to copy
4498 between memory and a register of @var{class} but the reload mechanism is
4499 more complex than copying via an intermediate, define this macro to
4500 reflect the actual cost of the move.
4502 GCC defines the function @code{memory_move_secondary_cost} if
4503 secondary reloads are needed. It computes the costs due to copying via
4504 a secondary register. If your machine copies from memory using a
4505 secondary register in the conventional way but the default base value of
4506 4 is not correct for your machine, define this macro to add some other
4507 value to the result of that function. The arguments to that function
4508 are the same as to this macro.
4510 These macros are obsolete, new ports should use the target hook
4511 @code{TARGET_MEMORY_MOVE_COST} instead.
4514 @hook TARGET_MEMORY_MOVE_COST
4516 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
4517 A C expression for the cost of a branch instruction. A value of 1 is
4518 the default; other values are interpreted relative to that. Parameter
4519 @var{speed_p} is true when the branch in question should be optimized
4520 for speed. When it is false, @code{BRANCH_COST} should return a value
4521 optimal for code size rather than performance. @var{predictable_p} is
4522 true for well-predicted branches. On many architectures the
4523 @code{BRANCH_COST} can be reduced then.
4526 Here are additional macros which do not specify precise relative costs,
4527 but only that certain actions are more expensive than GCC would
4530 @defmac SLOW_BYTE_ACCESS
4531 Define this macro as a C expression which is nonzero if accessing less
4532 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
4533 faster than accessing a word of memory, i.e., if such access
4534 require more than one instruction or if there is no difference in cost
4535 between byte and (aligned) word loads.
4537 When this macro is not defined, the compiler will access a field by
4538 finding the smallest containing object; when it is defined, a fullword
4539 load will be used if alignment permits. Unless bytes accesses are
4540 faster than word accesses, using word accesses is preferable since it
4541 may eliminate subsequent memory access if subsequent accesses occur to
4542 other fields in the same word of the structure, but to different bytes.
4545 @hook TARGET_SLOW_UNALIGNED_ACCESS
4547 @defmac MOVE_RATIO (@var{speed})
4548 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4549 which a sequence of insns should be generated instead of a
4550 string move insn or a library call. Increasing the value will always
4551 make code faster, but eventually incurs high cost in increased code size.
4553 Note that on machines where the corresponding move insn is a
4554 @code{define_expand} that emits a sequence of insns, this macro counts
4555 the number of such sequences.
4557 The parameter @var{speed} is true if the code is currently being
4558 optimized for speed rather than size.
4560 If you don't define this, a reasonable default is used.
4563 @hook TARGET_USE_BY_PIECES_INFRASTRUCTURE_P
4565 @hook TARGET_OVERLAP_OP_BY_PIECES_P
4567 @hook TARGET_COMPARE_BY_PIECES_BRANCH_RATIO
4569 @defmac MOVE_MAX_PIECES
4570 A C expression used by @code{move_by_pieces} to determine the largest unit
4571 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4574 @defmac STORE_MAX_PIECES
4575 A C expression used by @code{store_by_pieces} to determine the largest unit
4576 a store used to memory is. Defaults to @code{MOVE_MAX_PIECES}, or two times
4577 the size of @code{HOST_WIDE_INT}, whichever is smaller.
4580 @defmac COMPARE_MAX_PIECES
4581 A C expression used by @code{compare_by_pieces} to determine the largest unit
4582 a load or store used to compare memory is. Defaults to
4583 @code{MOVE_MAX_PIECES}.
4586 @defmac CLEAR_RATIO (@var{speed})
4587 The threshold of number of scalar move insns, @emph{below} which a sequence
4588 of insns should be generated to clear memory instead of a string clear insn
4589 or a library call. Increasing the value will always make code faster, but
4590 eventually incurs high cost in increased code size.
4592 The parameter @var{speed} is true if the code is currently being
4593 optimized for speed rather than size.
4595 If you don't define this, a reasonable default is used.
4598 @defmac SET_RATIO (@var{speed})
4599 The threshold of number of scalar move insns, @emph{below} which a sequence
4600 of insns should be generated to set memory to a constant value, instead of
4601 a block set insn or a library call.
4602 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, it defaults to the value of @code{MOVE_RATIO}.
4611 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
4612 A C expression used to determine whether a load postincrement is a good
4613 thing to use for a given mode. Defaults to the value of
4614 @code{HAVE_POST_INCREMENT}.
4617 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
4618 A C expression used to determine whether a load postdecrement is a good
4619 thing to use for a given mode. Defaults to the value of
4620 @code{HAVE_POST_DECREMENT}.
4623 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
4624 A C expression used to determine whether a load preincrement is a good
4625 thing to use for a given mode. Defaults to the value of
4626 @code{HAVE_PRE_INCREMENT}.
4629 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
4630 A C expression used to determine whether a load predecrement is a good
4631 thing to use for a given mode. Defaults to the value of
4632 @code{HAVE_PRE_DECREMENT}.
4635 @defmac USE_STORE_POST_INCREMENT (@var{mode})
4636 A C expression used to determine whether a store postincrement is a good
4637 thing to use for a given mode. Defaults to the value of
4638 @code{HAVE_POST_INCREMENT}.
4641 @defmac USE_STORE_POST_DECREMENT (@var{mode})
4642 A C expression used to determine whether a store postdecrement is a good
4643 thing to use for a given mode. Defaults to the value of
4644 @code{HAVE_POST_DECREMENT}.
4647 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
4648 This macro is used to determine whether a store preincrement is a good
4649 thing to use for a given mode. Defaults to the value of
4650 @code{HAVE_PRE_INCREMENT}.
4653 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
4654 This macro is used to determine whether a store predecrement is a good
4655 thing to use for a given mode. Defaults to the value of
4656 @code{HAVE_PRE_DECREMENT}.
4659 @defmac NO_FUNCTION_CSE
4660 Define this macro to be true if it is as good or better to call a constant
4661 function address than to call an address kept in a register.
4664 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
4665 Define this macro if a non-short-circuit operation produced by
4666 @samp{fold_range_test ()} is optimal. This macro defaults to true if
4667 @code{BRANCH_COST} is greater than or equal to the value 2.
4670 @hook TARGET_OPTAB_SUPPORTED_P
4672 @hook TARGET_RTX_COSTS
4674 @hook TARGET_ADDRESS_COST
4676 @hook TARGET_INSN_COST
4678 @hook TARGET_MAX_NOCE_IFCVT_SEQ_COST
4680 @hook TARGET_NOCE_CONVERSION_PROFITABLE_P
4682 @hook TARGET_NEW_ADDRESS_PROFITABLE_P
4684 @hook TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P
4686 @hook TARGET_ESTIMATED_POLY_VALUE
4689 @section Adjusting the Instruction Scheduler
4691 The instruction scheduler may need a fair amount of machine-specific
4692 adjustment in order to produce good code. GCC provides several target
4693 hooks for this purpose. It is usually enough to define just a few of
4694 them: try the first ones in this list first.
4696 @hook TARGET_SCHED_ISSUE_RATE
4698 @hook TARGET_SCHED_VARIABLE_ISSUE
4700 @hook TARGET_SCHED_ADJUST_COST
4702 @hook TARGET_SCHED_ADJUST_PRIORITY
4704 @hook TARGET_SCHED_REORDER
4706 @hook TARGET_SCHED_REORDER2
4708 @hook TARGET_SCHED_MACRO_FUSION_P
4710 @hook TARGET_SCHED_MACRO_FUSION_PAIR_P
4712 @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
4714 @hook TARGET_SCHED_INIT
4716 @hook TARGET_SCHED_FINISH
4718 @hook TARGET_SCHED_INIT_GLOBAL
4720 @hook TARGET_SCHED_FINISH_GLOBAL
4722 @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
4724 @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
4726 @hook TARGET_SCHED_DFA_POST_CYCLE_INSN
4728 @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
4730 @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
4732 @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
4734 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
4736 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
4738 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
4740 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
4742 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
4744 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
4746 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
4748 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
4750 @hook TARGET_SCHED_DFA_NEW_CYCLE
4752 @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
4754 @hook TARGET_SCHED_H_I_D_EXTENDED
4756 @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
4758 @hook TARGET_SCHED_INIT_SCHED_CONTEXT
4760 @hook TARGET_SCHED_SET_SCHED_CONTEXT
4762 @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
4764 @hook TARGET_SCHED_FREE_SCHED_CONTEXT
4766 @hook TARGET_SCHED_SPECULATE_INSN
4768 @hook TARGET_SCHED_NEEDS_BLOCK_P
4770 @hook TARGET_SCHED_GEN_SPEC_CHECK
4772 @hook TARGET_SCHED_SET_SCHED_FLAGS
4774 @hook TARGET_SCHED_CAN_SPECULATE_INSN
4776 @hook TARGET_SCHED_SMS_RES_MII
4778 @hook TARGET_SCHED_DISPATCH
4780 @hook TARGET_SCHED_DISPATCH_DO
4782 @hook TARGET_SCHED_EXPOSED_PIPELINE
4784 @hook TARGET_SCHED_REASSOCIATION_WIDTH
4786 @hook TARGET_SCHED_FUSION_PRIORITY
4788 @hook TARGET_EXPAND_DIVMOD_LIBFUNC
4791 @section Dividing the Output into Sections (Texts, Data, @dots{})
4792 @c the above section title is WAY too long. maybe cut the part between
4793 @c the (...)? --mew 10feb93
4795 An object file is divided into sections containing different types of
4796 data. In the most common case, there are three sections: the @dfn{text
4797 section}, which holds instructions and read-only data; the @dfn{data
4798 section}, which holds initialized writable data; and the @dfn{bss
4799 section}, which holds uninitialized data. Some systems have other kinds
4802 @file{varasm.cc} provides several well-known sections, such as
4803 @code{text_section}, @code{data_section} and @code{bss_section}.
4804 The normal way of controlling a @code{@var{foo}_section} variable
4805 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
4806 as described below. The macros are only read once, when @file{varasm.cc}
4807 initializes itself, so their values must be run-time constants.
4808 They may however depend on command-line flags.
4810 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
4811 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
4812 to be string literals.
4814 Some assemblers require a different string to be written every time a
4815 section is selected. If your assembler falls into this category, you
4816 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
4817 @code{get_unnamed_section} to set up the sections.
4819 You must always create a @code{text_section}, either by defining
4820 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
4821 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
4822 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
4823 create a distinct @code{readonly_data_section}, the default is to
4824 reuse @code{text_section}.
4826 All the other @file{varasm.cc} sections are optional, and are null
4827 if the target does not provide them.
4829 @defmac TEXT_SECTION_ASM_OP
4830 A C expression whose value is a string, including spacing, containing the
4831 assembler operation that should precede instructions and read-only data.
4832 Normally @code{"\t.text"} is right.
4835 @defmac HOT_TEXT_SECTION_NAME
4836 If defined, a C string constant for the name of the section containing most
4837 frequently executed functions of the program. If not defined, GCC will provide
4838 a default definition if the target supports named sections.
4841 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
4842 If defined, a C string constant for the name of the section containing unlikely
4843 executed functions in the program.
4846 @defmac DATA_SECTION_ASM_OP
4847 A C expression whose value is a string, including spacing, containing the
4848 assembler operation to identify the following data as writable initialized
4849 data. Normally @code{"\t.data"} is right.
4852 @defmac SDATA_SECTION_ASM_OP
4853 If defined, a C expression whose value is a string, including spacing,
4854 containing the assembler operation to identify the following data as
4855 initialized, writable small data.
4858 @defmac READONLY_DATA_SECTION_ASM_OP
4859 A C expression whose value is a string, including spacing, containing the
4860 assembler operation to identify the following data as read-only initialized
4864 @defmac BSS_SECTION_ASM_OP
4865 If defined, a C expression whose value is a string, including spacing,
4866 containing the assembler operation to identify the following data as
4867 uninitialized global data. If not defined, and
4868 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
4869 uninitialized global data will be output in the data section if
4870 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
4874 @defmac SBSS_SECTION_ASM_OP
4875 If defined, a C expression whose value is a string, including spacing,
4876 containing the assembler operation to identify the following data as
4877 uninitialized, writable small data.
4880 @defmac TLS_COMMON_ASM_OP
4881 If defined, a C expression whose value is a string containing the
4882 assembler operation to identify the following data as thread-local
4883 common data. The default is @code{".tls_common"}.
4886 @defmac TLS_SECTION_ASM_FLAG
4887 If defined, a C expression whose value is a character constant
4888 containing the flag used to mark a section as a TLS section. The
4889 default is @code{'T'}.
4892 @defmac INIT_SECTION_ASM_OP
4893 If defined, a C expression whose value is a string, including spacing,
4894 containing the assembler operation to identify the following data as
4895 initialization code. If not defined, GCC will assume such a section does
4896 not exist. This section has no corresponding @code{init_section}
4897 variable; it is used entirely in runtime code.
4900 @defmac FINI_SECTION_ASM_OP
4901 If defined, a C expression whose value is a string, including spacing,
4902 containing the assembler operation to identify the following data as
4903 finalization code. If not defined, GCC will assume such a section does
4904 not exist. This section has no corresponding @code{fini_section}
4905 variable; it is used entirely in runtime code.
4908 @defmac INIT_ARRAY_SECTION_ASM_OP
4909 If defined, a C expression whose value is a string, including spacing,
4910 containing the assembler operation to identify the following data as
4911 part of the @code{.init_array} (or equivalent) section. If not
4912 defined, GCC will assume such a section does not exist. Do not define
4913 both this macro and @code{INIT_SECTION_ASM_OP}.
4916 @defmac FINI_ARRAY_SECTION_ASM_OP
4917 If defined, a C expression whose value is a string, including spacing,
4918 containing the assembler operation to identify the following data as
4919 part of the @code{.fini_array} (or equivalent) section. If not
4920 defined, GCC will assume such a section does not exist. Do not define
4921 both this macro and @code{FINI_SECTION_ASM_OP}.
4924 @defmac MACH_DEP_SECTION_ASM_FLAG
4925 If defined, a C expression whose value is a character constant
4926 containing the flag used to mark a machine-dependent section. This
4927 corresponds to the @code{SECTION_MACH_DEP} section flag.
4930 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
4931 If defined, an ASM statement that switches to a different section
4932 via @var{section_op}, calls @var{function}, and switches back to
4933 the text section. This is used in @file{crtstuff.c} if
4934 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
4935 to initialization and finalization functions from the init and fini
4936 sections. By default, this macro uses a simple function call. Some
4937 ports need hand-crafted assembly code to avoid dependencies on
4938 registers initialized in the function prologue or to ensure that
4939 constant pools don't end up too far way in the text section.
4942 @defmac TARGET_LIBGCC_SDATA_SECTION
4943 If defined, a string which names the section into which small
4944 variables defined in crtstuff and libgcc should go. This is useful
4945 when the target has options for optimizing access to small data, and
4946 you want the crtstuff and libgcc routines to be conservative in what
4947 they expect of your application yet liberal in what your application
4948 expects. For example, for targets with a @code{.sdata} section (like
4949 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
4950 require small data support from your application, but use this macro
4951 to put small data into @code{.sdata} so that your application can
4952 access these variables whether it uses small data or not.
4955 @defmac FORCE_CODE_SECTION_ALIGN
4956 If defined, an ASM statement that aligns a code section to some
4957 arbitrary boundary. This is used to force all fragments of the
4958 @code{.init} and @code{.fini} sections to have to same alignment
4959 and thus prevent the linker from having to add any padding.
4962 @defmac JUMP_TABLES_IN_TEXT_SECTION
4963 Define this macro to be an expression with a nonzero value if jump
4964 tables (for @code{tablejump} insns) should be output in the text
4965 section, along with the assembler instructions. Otherwise, the
4966 readonly data section is used.
4968 This macro is irrelevant if there is no separate readonly data section.
4971 @hook TARGET_ASM_INIT_SECTIONS
4973 @hook TARGET_ASM_RELOC_RW_MASK
4975 @hook TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC
4977 @hook TARGET_ASM_SELECT_SECTION
4979 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
4980 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
4981 for @code{FUNCTION_DECL}s as well as for variables and constants.
4983 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
4984 function has been determined to be likely to be called, and nonzero if
4985 it is unlikely to be called.
4988 @hook TARGET_ASM_UNIQUE_SECTION
4990 @hook TARGET_ASM_FUNCTION_RODATA_SECTION
4992 @hook TARGET_ASM_MERGEABLE_RODATA_PREFIX
4994 @hook TARGET_ASM_TM_CLONE_TABLE_SECTION
4996 @hook TARGET_ASM_SELECT_RTX_SECTION
4998 @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
5000 @hook TARGET_ENCODE_SECTION_INFO
5002 @hook TARGET_STRIP_NAME_ENCODING
5004 @hook TARGET_IN_SMALL_DATA_P
5006 @hook TARGET_HAVE_SRODATA_SECTION
5008 @hook TARGET_PROFILE_BEFORE_PROLOGUE
5010 @hook TARGET_BINDS_LOCAL_P
5012 @hook TARGET_HAVE_TLS
5016 @section Position Independent Code
5017 @cindex position independent code
5020 This section describes macros that help implement generation of position
5021 independent code. Simply defining these macros is not enough to
5022 generate valid PIC; you must also add support to the hook
5023 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
5024 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
5025 must modify the definition of @samp{movsi} to do something appropriate
5026 when the source operand contains a symbolic address. You may also
5027 need to alter the handling of switch statements so that they use
5029 @c i rearranged the order of the macros above to try to force one of
5030 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5032 @defmac PIC_OFFSET_TABLE_REGNUM
5033 The register number of the register used to address a table of static
5034 data addresses in memory. In some cases this register is defined by a
5035 processor's ``application binary interface'' (ABI)@. When this macro
5036 is defined, RTL is generated for this register once, as with the stack
5037 pointer and frame pointer registers. If this macro is not defined, it
5038 is up to the machine-dependent files to allocate such a register (if
5039 necessary). Note that this register must be fixed when in use (e.g.@:
5040 when @code{flag_pic} is true).
5043 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5044 A C expression that is nonzero if the register defined by
5045 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
5046 the default is zero. Do not define
5047 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5050 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
5051 A C expression that is nonzero if @var{x} is a legitimate immediate
5052 operand on the target machine when generating position independent code.
5053 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5054 check this. You can also assume @var{flag_pic} is true, so you need not
5055 check it either. You need not define this macro if all constants
5056 (including @code{SYMBOL_REF}) can be immediate operands when generating
5057 position independent code.
5060 @node Assembler Format
5061 @section Defining the Output Assembler Language
5063 This section describes macros whose principal purpose is to describe how
5064 to write instructions in assembler language---rather than what the
5068 * File Framework:: Structural information for the assembler file.
5069 * Data Output:: Output of constants (numbers, strings, addresses).
5070 * Uninitialized Data:: Output of uninitialized variables.
5071 * Label Output:: Output and generation of labels.
5072 * Initialization:: General principles of initialization
5073 and termination routines.
5074 * Macros for Initialization::
5075 Specific macros that control the handling of
5076 initialization and termination routines.
5077 * Instruction Output:: Output of actual instructions.
5078 * Dispatch Tables:: Output of jump tables.
5079 * Exception Region Output:: Output of exception region code.
5080 * Alignment Output:: Pseudo ops for alignment and skipping data.
5083 @node File Framework
5084 @subsection The Overall Framework of an Assembler File
5085 @cindex assembler format
5086 @cindex output of assembler code
5088 @c prevent bad page break with this line
5089 This describes the overall framework of an assembly file.
5091 @findex default_file_start
5092 @hook TARGET_ASM_FILE_START
5094 @hook TARGET_ASM_FILE_START_APP_OFF
5096 @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
5098 @hook TARGET_ASM_FILE_END
5100 @deftypefun void file_end_indicate_exec_stack ()
5101 Some systems use a common convention, the @samp{.note.GNU-stack}
5102 special section, to indicate whether or not an object file relies on
5103 the stack being executable. If your system uses this convention, you
5104 should define @code{TARGET_ASM_FILE_END} to this function. If you
5105 need to do other things in that hook, have your hook function call
5109 @hook TARGET_ASM_LTO_START
5111 @hook TARGET_ASM_LTO_END
5113 @hook TARGET_ASM_CODE_END
5115 @defmac ASM_COMMENT_START
5116 A C string constant describing how to begin a comment in the target
5117 assembler language. The compiler assumes that the comment will end at
5118 the end of the line.
5122 A C string constant for text to be output before each @code{asm}
5123 statement or group of consecutive ones. Normally this is
5124 @code{"#APP"}, which is a comment that has no effect on most
5125 assemblers but tells the GNU assembler that it must check the lines
5126 that follow for all valid assembler constructs.
5130 A C string constant for text to be output after each @code{asm}
5131 statement or group of consecutive ones. Normally this is
5132 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5133 time-saving assumptions that are valid for ordinary compiler output.
5136 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5137 A C statement to output COFF information or DWARF debugging information
5138 which indicates that filename @var{name} is the current source file to
5139 the stdio stream @var{stream}.
5141 This macro need not be defined if the standard form of output
5142 for the file format in use is appropriate.
5145 @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
5147 @hook TARGET_ASM_OUTPUT_IDENT
5149 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5150 A C statement to output the string @var{string} to the stdio stream
5151 @var{stream}. If you do not call the function @code{output_quoted_string}
5152 in your config files, GCC will only call it to output filenames to
5153 the assembler source. So you can use it to canonicalize the format
5154 of the filename using this macro.
5157 @hook TARGET_ASM_NAMED_SECTION
5159 @hook TARGET_ASM_ELF_FLAGS_NUMERIC
5161 @hook TARGET_ASM_FUNCTION_SECTION
5163 @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
5165 @hook TARGET_HAVE_NAMED_SECTIONS
5166 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
5167 It must not be modified by command-line option processing.
5170 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
5171 @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
5173 @hook TARGET_SECTION_TYPE_FLAGS
5175 @hook TARGET_ASM_RECORD_GCC_SWITCHES
5177 @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
5181 @subsection Output of Data
5184 @hook TARGET_ASM_BYTE_OP
5186 @hook TARGET_ASM_INTEGER
5188 @hook TARGET_ASM_DECL_END
5190 @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
5192 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5193 A C statement to output to the stdio stream @var{stream} an assembler
5194 instruction to assemble a string constant containing the @var{len}
5195 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5196 @code{char *} and @var{len} a C expression of type @code{int}.
5198 If the assembler has a @code{.ascii} pseudo-op as found in the
5199 Berkeley Unix assembler, do not define the macro
5200 @code{ASM_OUTPUT_ASCII}.
5203 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
5204 A C statement to output word @var{n} of a function descriptor for
5205 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
5206 is defined, and is otherwise unused.
5209 @defmac CONSTANT_POOL_BEFORE_FUNCTION
5210 You may define this macro as a C expression. You should define the
5211 expression to have a nonzero value if GCC should output the constant
5212 pool for a function before the code for the function, or a zero value if
5213 GCC should output the constant pool after the function. If you do
5214 not define this macro, the usual case, GCC will output the constant
5215 pool before the function.
5218 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5219 A C statement to output assembler commands to define the start of the
5220 constant pool for a function. @var{funname} is a string giving
5221 the name of the function. Should the return type of the function
5222 be required, it can be obtained via @var{fundecl}. @var{size}
5223 is the size, in bytes, of the constant pool that will be written
5224 immediately after this call.
5226 If no constant-pool prefix is required, the usual case, this macro need
5230 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5231 A C statement (with or without semicolon) to output a constant in the
5232 constant pool, if it needs special treatment. (This macro need not do
5233 anything for RTL expressions that can be output normally.)
5235 The argument @var{file} is the standard I/O stream to output the
5236 assembler code on. @var{x} is the RTL expression for the constant to
5237 output, and @var{mode} is the machine mode (in case @var{x} is a
5238 @samp{const_int}). @var{align} is the required alignment for the value
5239 @var{x}; you should output an assembler directive to force this much
5242 The argument @var{labelno} is a number to use in an internal label for
5243 the address of this pool entry. The definition of this macro is
5244 responsible for outputting the label definition at the proper place.
5245 Here is how to do this:
5248 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
5251 When you output a pool entry specially, you should end with a
5252 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5253 entry from being output a second time in the usual manner.
5255 You need not define this macro if it would do nothing.
5258 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5259 A C statement to output assembler commands to at the end of the constant
5260 pool for a function. @var{funname} is a string giving the name of the
5261 function. Should the return type of the function be required, you can
5262 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5263 constant pool that GCC wrote immediately before this call.
5265 If no constant-pool epilogue is required, the usual case, you need not
5269 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
5270 Define this macro as a C expression which is nonzero if @var{C} is
5271 used as a logical line separator by the assembler. @var{STR} points
5272 to the position in the string where @var{C} was found; this can be used if
5273 a line separator uses multiple characters.
5275 If you do not define this macro, the default is that only
5276 the character @samp{;} is treated as a logical line separator.
5279 @hook TARGET_ASM_OPEN_PAREN
5281 These macros are provided by @file{real.h} for writing the definitions
5282 of @code{ASM_OUTPUT_DOUBLE} and the like:
5284 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5285 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5286 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5287 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
5288 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
5289 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
5290 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
5291 target's floating point representation, and store its bit pattern in
5292 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
5293 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
5294 simple @code{long int}. For the others, it should be an array of
5295 @code{long int}. The number of elements in this array is determined
5296 by the size of the desired target floating point data type: 32 bits of
5297 it go in each @code{long int} array element. Each array element holds
5298 32 bits of the result, even if @code{long int} is wider than 32 bits
5299 on the host machine.
5301 The array element values are designed so that you can print them out
5302 using @code{fprintf} in the order they should appear in the target
5306 @node Uninitialized Data
5307 @subsection Output of Uninitialized Variables
5309 Each of the macros in this section is used to do the whole job of
5310 outputting a single uninitialized variable.
5312 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5313 A C statement (sans semicolon) to output to the stdio stream
5314 @var{stream} the assembler definition of a common-label named
5315 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5316 is the size rounded up to whatever alignment the caller wants. It is
5317 possible that @var{size} may be zero, for instance if a struct with no
5318 other member than a zero-length array is defined. In this case, the
5319 backend must output a symbol definition that allocates at least one
5320 byte, both so that the address of the resulting object does not compare
5321 equal to any other, and because some object formats cannot even express
5322 the concept of a zero-sized common symbol, as that is how they represent
5323 an ordinary undefined external.
5325 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5326 output the name itself; before and after that, output the additional
5327 assembler syntax for defining the name, and a newline.
5329 This macro controls how the assembler definitions of uninitialized
5330 common global variables are output.
5333 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5334 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5335 separate, explicit argument. If you define this macro, it is used in
5336 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5337 handling the required alignment of the variable. The alignment is specified
5338 as the number of bits.
5341 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5342 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5343 variable to be output, if there is one, or @code{NULL_TREE} if there
5344 is no corresponding variable. If you define this macro, GCC will use it
5345 in place of both @code{ASM_OUTPUT_COMMON} and
5346 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5347 the variable's decl in order to chose what to output.
5350 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5351 A C statement (sans semicolon) to output to the stdio stream
5352 @var{stream} the assembler definition of uninitialized global @var{decl} named
5353 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
5354 is the alignment specified as the number of bits.
5356 Try to use function @code{asm_output_aligned_bss} defined in file
5357 @file{varasm.cc} when defining this macro. If unable, use the expression
5358 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5359 before and after that, output the additional assembler syntax for defining
5360 the name, and a newline.
5362 There are two ways of handling global BSS@. One is to define this macro.
5363 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
5364 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
5365 You do not need to do both.
5367 Some languages do not have @code{common} data, and require a
5368 non-common form of global BSS in order to handle uninitialized globals
5369 efficiently. C++ is one example of this. However, if the target does
5370 not support global BSS, the front end may choose to make globals
5371 common in order to save space in the object file.
5374 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5375 A C statement (sans semicolon) to output to the stdio stream
5376 @var{stream} the assembler definition of a local-common-label named
5377 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5378 is the size rounded up to whatever alignment the caller wants.
5380 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5381 output the name itself; before and after that, output the additional
5382 assembler syntax for defining the name, and a newline.
5384 This macro controls how the assembler definitions of uninitialized
5385 static variables are output.
5388 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5389 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5390 separate, explicit argument. If you define this macro, it is used in
5391 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5392 handling the required alignment of the variable. The alignment is specified
5393 as the number of bits.
5396 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5397 Like @code{ASM_OUTPUT_ALIGNED_LOCAL} except that @var{decl} of the
5398 variable to be output, if there is one, or @code{NULL_TREE} if there
5399 is no corresponding variable. If you define this macro, GCC will use it
5400 in place of both @code{ASM_OUTPUT_LOCAL} and
5401 @code{ASM_OUTPUT_ALIGNED_LOCAL}. Define this macro when you need to see
5402 the variable's decl in order to chose what to output.
5406 @subsection Output and Generation of Labels
5408 @c prevent bad page break with this line
5409 This is about outputting labels.
5411 @findex assemble_name
5412 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5413 A C statement (sans semicolon) to output to the stdio stream
5414 @var{stream} the assembler definition of a label named @var{name}.
5415 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5416 output the name itself; before and after that, output the additional
5417 assembler syntax for defining the name, and a newline. A default
5418 definition of this macro is provided which is correct for most systems.
5421 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
5422 A C statement (sans semicolon) to output to the stdio stream
5423 @var{stream} the assembler definition of a label named @var{name} of
5425 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5426 output the name itself; before and after that, output the additional
5427 assembler syntax for defining the name, and a newline. A default
5428 definition of this macro is provided which is correct for most systems.
5430 If this macro is not defined, then the function name is defined in the
5431 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5434 @findex assemble_name_raw
5435 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
5436 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
5437 to refer to a compiler-generated label. The default definition uses
5438 @code{assemble_name_raw}, which is like @code{assemble_name} except
5439 that it is more efficient.
5443 A C string containing the appropriate assembler directive to specify the
5444 size of a symbol, without any arguments. On systems that use ELF, the
5445 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
5446 systems, the default is not to define this macro.
5448 Define this macro only if it is correct to use the default definitions
5449 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
5450 for your system. If you need your own custom definitions of those
5451 macros, or if you do not need explicit symbol sizes at all, do not
5455 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
5456 A C statement (sans semicolon) to output to the stdio stream
5457 @var{stream} a directive telling the assembler that the size of the
5458 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
5459 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5463 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
5464 A C statement (sans semicolon) to output to the stdio stream
5465 @var{stream} a directive telling the assembler to calculate the size of
5466 the symbol @var{name} by subtracting its address from the current
5469 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5470 provided. The default assumes that the assembler recognizes a special
5471 @samp{.} symbol as referring to the current address, and can calculate
5472 the difference between this and another symbol. If your assembler does
5473 not recognize @samp{.} or cannot do calculations with it, you will need
5474 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
5477 @defmac NO_DOLLAR_IN_LABEL
5478 Define this macro if the assembler does not accept the character
5479 @samp{$} in label names. By default constructors and destructors in
5480 G++ have @samp{$} in the identifiers. If this macro is defined,
5481 @samp{.} is used instead.
5484 @defmac NO_DOT_IN_LABEL
5485 Define this macro if the assembler does not accept the character
5486 @samp{.} in label names. By default constructors and destructors in G++
5487 have names that use @samp{.}. If this macro is defined, these names
5488 are rewritten to avoid @samp{.}.
5492 A C string containing the appropriate assembler directive to specify the
5493 type of a symbol, without any arguments. On systems that use ELF, the
5494 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
5495 systems, the default is not to define this macro.
5497 Define this macro only if it is correct to use the default definition of
5498 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5499 custom definition of this macro, or if you do not need explicit symbol
5500 types at all, do not define this macro.
5503 @defmac TYPE_OPERAND_FMT
5504 A C string which specifies (using @code{printf} syntax) the format of
5505 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
5506 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
5507 the default is not to define this macro.
5509 Define this macro only if it is correct to use the default definition of
5510 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5511 custom definition of this macro, or if you do not need explicit symbol
5512 types at all, do not define this macro.
5515 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
5516 A C statement (sans semicolon) to output to the stdio stream
5517 @var{stream} a directive telling the assembler that the type of the
5518 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
5519 that string is always either @samp{"function"} or @samp{"object"}, but
5520 you should not count on this.
5522 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
5523 definition of this macro is provided.
5526 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5527 A C statement (sans semicolon) to output to the stdio stream
5528 @var{stream} any text necessary for declaring the name @var{name} of a
5529 function which is being defined. This macro is responsible for
5530 outputting the label definition (perhaps using
5531 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5532 @code{FUNCTION_DECL} tree node representing the function.
5534 If this macro is not defined, then the function name is defined in the
5535 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
5537 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5541 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5542 A C statement (sans semicolon) to output to the stdio stream
5543 @var{stream} any text necessary for declaring the size of a function
5544 which is being defined. The argument @var{name} is the name of the
5545 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5546 representing the function.
5548 If this macro is not defined, then the function size is not defined.
5550 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5554 @defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5555 A C statement (sans semicolon) to output to the stdio stream
5556 @var{stream} any text necessary for declaring the name @var{name} of a
5557 cold function partition which is being defined. This macro is responsible
5558 for outputting the label definition (perhaps using
5559 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5560 @code{FUNCTION_DECL} tree node representing the function.
5562 If this macro is not defined, then the cold partition name is defined in the
5563 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5565 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5569 @defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5570 A C statement (sans semicolon) to output to the stdio stream
5571 @var{stream} any text necessary for declaring the size of a cold function
5572 partition which is being defined. The argument @var{name} is the name of the
5573 cold partition of the function. The argument @var{decl} is the
5574 @code{FUNCTION_DECL} tree node representing the function.
5576 If this macro is not defined, then the partition size is not defined.
5578 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5582 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5583 A C statement (sans semicolon) to output to the stdio stream
5584 @var{stream} any text necessary for declaring the name @var{name} of an
5585 initialized variable which is being defined. This macro must output the
5586 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5587 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5589 If this macro is not defined, then the variable name is defined in the
5590 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5592 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
5593 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
5596 @hook TARGET_ASM_DECLARE_CONSTANT_NAME
5598 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5599 A C statement (sans semicolon) to output to the stdio stream
5600 @var{stream} any text necessary for claiming a register @var{regno}
5601 for a global variable @var{decl} with name @var{name}.
5603 If you don't define this macro, that is equivalent to defining it to do
5607 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5608 A C statement (sans semicolon) to finish up declaring a variable name
5609 once the compiler has processed its initializer fully and thus has had a
5610 chance to determine the size of an array when controlled by an
5611 initializer. This is used on systems where it's necessary to declare
5612 something about the size of the object.
5614 If you don't define this macro, that is equivalent to defining it to do
5617 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
5618 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
5621 @hook TARGET_ASM_GLOBALIZE_LABEL
5623 @hook TARGET_ASM_GLOBALIZE_DECL_NAME
5625 @hook TARGET_ASM_ASSEMBLE_UNDEFINED_DECL
5627 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
5628 A C statement (sans semicolon) to output to the stdio stream
5629 @var{stream} some commands that will make the label @var{name} weak;
5630 that is, available for reference from other files but only used if
5631 no other definition is available. Use the expression
5632 @code{assemble_name (@var{stream}, @var{name})} to output the name
5633 itself; before and after that, output the additional assembler syntax
5634 for making that name weak, and a newline.
5636 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
5637 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
5641 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
5642 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
5643 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
5644 or variable decl. If @var{value} is not @code{NULL}, this C statement
5645 should output to the stdio stream @var{stream} assembler code which
5646 defines (equates) the weak symbol @var{name} to have the value
5647 @var{value}. If @var{value} is @code{NULL}, it should output commands
5648 to make @var{name} weak.
5651 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
5652 Outputs a directive that enables @var{name} to be used to refer to
5653 symbol @var{value} with weak-symbol semantics. @code{decl} is the
5654 declaration of @code{name}.
5657 @defmac SUPPORTS_WEAK
5658 A preprocessor constant expression which evaluates to true if the target
5659 supports weak symbols.
5661 If you don't define this macro, @file{defaults.h} provides a default
5662 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
5663 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
5666 @defmac TARGET_SUPPORTS_WEAK
5667 A C expression which evaluates to true if the target supports weak symbols.
5669 If you don't define this macro, @file{defaults.h} provides a default
5670 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
5671 this macro if you want to control weak symbol support with a compiler
5672 flag such as @option{-melf}.
5675 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
5676 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5677 public symbol such that extra copies in multiple translation units will
5678 be discarded by the linker. Define this macro if your object file
5679 format provides support for this concept, such as the @samp{COMDAT}
5680 section flags in the Microsoft Windows PE/COFF format, and this support
5681 requires changes to @var{decl}, such as putting it in a separate section.
5684 @defmac SUPPORTS_ONE_ONLY
5685 A C expression which evaluates to true if the target supports one-only
5688 If you don't define this macro, @file{varasm.cc} provides a default
5689 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5690 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5691 you want to control one-only symbol support with a compiler flag, or if
5692 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5693 be emitted as one-only.
5696 @hook TARGET_ASM_ASSEMBLE_VISIBILITY
5698 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
5699 A C expression that evaluates to true if the target's linker expects
5700 that weak symbols do not appear in a static archive's table of contents.
5701 The default is @code{0}.
5703 Leaving weak symbols out of an archive's table of contents means that,
5704 if a symbol will only have a definition in one translation unit and
5705 will have undefined references from other translation units, that
5706 symbol should not be weak. Defining this macro to be nonzero will
5707 thus have the effect that certain symbols that would normally be weak
5708 (explicit template instantiations, and vtables for polymorphic classes
5709 with noninline key methods) will instead be nonweak.
5711 The C++ ABI requires this macro to be zero. Define this macro for
5712 targets where full C++ ABI compliance is impossible and where linker
5713 restrictions require weak symbols to be left out of a static archive's
5717 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5718 A C statement (sans semicolon) to output to the stdio stream
5719 @var{stream} any text necessary for declaring the name of an external
5720 symbol named @var{name} which is referenced in this compilation but
5721 not defined. The value of @var{decl} is the tree node for the
5724 This macro need not be defined if it does not need to output anything.
5725 The GNU assembler and most Unix assemblers don't require anything.
5728 @hook TARGET_ASM_EXTERNAL_LIBCALL
5730 @hook TARGET_ASM_MARK_DECL_PRESERVED
5732 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5733 A C statement (sans semicolon) to output to the stdio stream
5734 @var{stream} a reference in assembler syntax to a label named
5735 @var{name}. This should add @samp{_} to the front of the name, if that
5736 is customary on your operating system, as it is in most Berkeley Unix
5737 systems. This macro is used in @code{assemble_name}.
5740 @hook TARGET_MANGLE_ASSEMBLER_NAME
5742 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
5743 A C statement (sans semicolon) to output a reference to
5744 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
5745 will be used to output the name of the symbol. This macro may be used
5746 to modify the way a symbol is referenced depending on information
5747 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
5750 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
5751 A C statement (sans semicolon) to output a reference to @var{buf}, the
5752 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
5753 @code{assemble_name} will be used to output the name of the symbol.
5754 This macro is not used by @code{output_asm_label}, or the @code{%l}
5755 specifier that calls it; the intention is that this macro should be set
5756 when it is necessary to output a label differently when its address is
5760 @hook TARGET_ASM_INTERNAL_LABEL
5762 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
5763 A C statement to output to the stdio stream @var{stream} a debug info
5764 label whose name is made from the string @var{prefix} and the number
5765 @var{num}. This is useful for VLIW targets, where debug info labels
5766 may need to be treated differently than branch target labels. On some
5767 systems, branch target labels must be at the beginning of instruction
5768 bundles, but debug info labels can occur in the middle of instruction
5771 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
5775 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5776 A C statement to store into the string @var{string} a label whose name
5777 is made from the string @var{prefix} and the number @var{num}.
5779 This string, when output subsequently by @code{assemble_name}, should
5780 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
5781 with the same @var{prefix} and @var{num}.
5783 If the string begins with @samp{*}, then @code{assemble_name} will
5784 output the rest of the string unchanged. It is often convenient for
5785 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5786 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5787 to output the string, and may change it. (Of course,
5788 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5789 you should know what it does on your machine.)
5792 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5793 A C expression to assign to @var{outvar} (which is a variable of type
5794 @code{char *}) a newly allocated string made from the string
5795 @var{name} and the number @var{number}, with some suitable punctuation
5796 added. Use @code{alloca} to get space for the string.
5798 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5799 produce an assembler label for an internal static variable whose name is
5800 @var{name}. Therefore, the string must be such as to result in valid
5801 assembler code. The argument @var{number} is different each time this
5802 macro is executed; it prevents conflicts between similarly-named
5803 internal static variables in different scopes.
5805 Ideally this string should not be a valid C identifier, to prevent any
5806 conflict with the user's own symbols. Most assemblers allow periods
5807 or percent signs in assembler symbols; putting at least one of these
5808 between the name and the number will suffice.
5810 If this macro is not defined, a default definition will be provided
5811 which is correct for most systems.
5814 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5815 A C statement to output to the stdio stream @var{stream} assembler code
5816 which defines (equates) the symbol @var{name} to have the value @var{value}.
5819 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5820 correct for most systems.
5823 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
5824 A C statement to output to the stdio stream @var{stream} assembler code
5825 which defines (equates) the symbol whose tree node is @var{decl_of_name}
5826 to have the value of the tree node @var{decl_of_value}. This macro will
5827 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
5828 the tree nodes are available.
5831 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5832 correct for most systems.
5835 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
5836 A C statement that evaluates to true if the assembler code which defines
5837 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
5838 of the tree node @var{decl_of_value} should be emitted near the end of the
5839 current compilation unit. The default is to not defer output of defines.
5840 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
5841 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
5844 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5845 A C statement to output to the stdio stream @var{stream} assembler code
5846 which defines (equates) the weak symbol @var{name} to have the value
5847 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
5848 an undefined weak symbol.
5850 Define this macro if the target only supports weak aliases; define
5851 @code{ASM_OUTPUT_DEF} instead if possible.
5854 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5855 Define this macro to override the default assembler names used for
5856 Objective-C methods.
5858 The default name is a unique method number followed by the name of the
5859 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
5860 the category is also included in the assembler name (e.g.@:
5863 These names are safe on most systems, but make debugging difficult since
5864 the method's selector is not present in the name. Therefore, particular
5865 systems define other ways of computing names.
5867 @var{buf} is an expression of type @code{char *} which gives you a
5868 buffer in which to store the name; its length is as long as
5869 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
5870 50 characters extra.
5872 The argument @var{is_inst} specifies whether the method is an instance
5873 method or a class method; @var{class_name} is the name of the class;
5874 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
5875 in a category); and @var{sel_name} is the name of the selector.
5877 On systems where the assembler can handle quoted names, you can use this
5878 macro to provide more human-readable names.
5881 @node Initialization
5882 @subsection How Initialization Functions Are Handled
5883 @cindex initialization routines
5884 @cindex termination routines
5885 @cindex constructors, output of
5886 @cindex destructors, output of
5888 The compiled code for certain languages includes @dfn{constructors}
5889 (also called @dfn{initialization routines})---functions to initialize
5890 data in the program when the program is started. These functions need
5891 to be called before the program is ``started''---that is to say, before
5892 @code{main} is called.
5894 Compiling some languages generates @dfn{destructors} (also called
5895 @dfn{termination routines}) that should be called when the program
5898 To make the initialization and termination functions work, the compiler
5899 must output something in the assembler code to cause those functions to
5900 be called at the appropriate time. When you port the compiler to a new
5901 system, you need to specify how to do this.
5903 There are two major ways that GCC currently supports the execution of
5904 initialization and termination functions. Each way has two variants.
5905 Much of the structure is common to all four variations.
5907 @findex __CTOR_LIST__
5908 @findex __DTOR_LIST__
5909 The linker must build two lists of these functions---a list of
5910 initialization functions, called @code{__CTOR_LIST__}, and a list of
5911 termination functions, called @code{__DTOR_LIST__}.
5913 Each list always begins with an ignored function pointer (which may hold
5914 0, @minus{}1, or a count of the function pointers after it, depending on
5915 the environment). This is followed by a series of zero or more function
5916 pointers to constructors (or destructors), followed by a function
5917 pointer containing zero.
5919 Depending on the operating system and its executable file format, either
5920 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
5921 time and exit time. Constructors are called in reverse order of the
5922 list; destructors in forward order.
5924 The best way to handle static constructors works only for object file
5925 formats which provide arbitrarily-named sections. A section is set
5926 aside for a list of constructors, and another for a list of destructors.
5927 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
5928 object file that defines an initialization function also puts a word in
5929 the constructor section to point to that function. The linker
5930 accumulates all these words into one contiguous @samp{.ctors} section.
5931 Termination functions are handled similarly.
5933 This method will be chosen as the default by @file{target-def.h} if
5934 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
5935 support arbitrary sections, but does support special designated
5936 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
5937 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
5939 When arbitrary sections are available, there are two variants, depending
5940 upon how the code in @file{crtstuff.c} is called. On systems that
5941 support a @dfn{.init} section which is executed at program startup,
5942 parts of @file{crtstuff.c} are compiled into that section. The
5943 program is linked by the @command{gcc} driver like this:
5946 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
5949 The prologue of a function (@code{__init}) appears in the @code{.init}
5950 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
5951 for the function @code{__fini} in the @dfn{.fini} section. Normally these
5952 files are provided by the operating system or by the GNU C library, but
5953 are provided by GCC for a few targets.
5955 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
5956 compiled from @file{crtstuff.c}. They contain, among other things, code
5957 fragments within the @code{.init} and @code{.fini} sections that branch
5958 to routines in the @code{.text} section. The linker will pull all parts
5959 of a section together, which results in a complete @code{__init} function
5960 that invokes the routines we need at startup.
5962 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
5965 If no init section is available, when GCC compiles any function called
5966 @code{main} (or more accurately, any function designated as a program
5967 entry point by the language front end calling @code{expand_main_function}),
5968 it inserts a procedure call to @code{__main} as the first executable code
5969 after the function prologue. The @code{__main} function is defined
5970 in @file{libgcc2.c} and runs the global constructors.
5972 In file formats that don't support arbitrary sections, there are again
5973 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
5974 and an `a.out' format must be used. In this case,
5975 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
5976 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
5977 and with the address of the void function containing the initialization
5978 code as its value. The GNU linker recognizes this as a request to add
5979 the value to a @dfn{set}; the values are accumulated, and are eventually
5980 placed in the executable as a vector in the format described above, with
5981 a leading (ignored) count and a trailing zero element.
5982 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
5983 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
5984 the compilation of @code{main} to call @code{__main} as above, starting
5985 the initialization process.
5987 The last variant uses neither arbitrary sections nor the GNU linker.
5988 This is preferable when you want to do dynamic linking and when using
5989 file formats which the GNU linker does not support, such as `ECOFF'@. In
5990 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
5991 termination functions are recognized simply by their names. This requires
5992 an extra program in the linkage step, called @command{collect2}. This program
5993 pretends to be the linker, for use with GCC; it does its job by running
5994 the ordinary linker, but also arranges to include the vectors of
5995 initialization and termination functions. These functions are called
5996 via @code{__main} as described above. In order to use this method,
5997 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6000 The following section describes the specific macros that control and
6001 customize the handling of initialization and termination functions.
6004 @node Macros for Initialization
6005 @subsection Macros Controlling Initialization Routines
6007 Here are the macros that control how the compiler handles initialization
6008 and termination functions:
6010 @defmac INIT_SECTION_ASM_OP
6011 If defined, a C string constant, including spacing, for the assembler
6012 operation to identify the following data as initialization code. If not
6013 defined, GCC will assume such a section does not exist. When you are
6014 using special sections for initialization and termination functions, this
6015 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6016 run the initialization functions.
6019 @defmac HAS_INIT_SECTION
6020 If defined, @code{main} will not call @code{__main} as described above.
6021 This macro should be defined for systems that control start-up code
6022 on a symbol-by-symbol basis, such as OSF/1, and should not
6023 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
6026 @defmac LD_INIT_SWITCH
6027 If defined, a C string constant for a switch that tells the linker that
6028 the following symbol is an initialization routine.
6031 @defmac LD_FINI_SWITCH
6032 If defined, a C string constant for a switch that tells the linker that
6033 the following symbol is a finalization routine.
6036 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
6037 If defined, a C statement that will write a function that can be
6038 automatically called when a shared library is loaded. The function
6039 should call @var{func}, which takes no arguments. If not defined, and
6040 the object format requires an explicit initialization function, then a
6041 function called @code{_GLOBAL__DI} will be generated.
6043 This function and the following one are used by collect2 when linking a
6044 shared library that needs constructors or destructors, or has DWARF2
6045 exception tables embedded in the code.
6048 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
6049 If defined, a C statement that will write a function that can be
6050 automatically called when a shared library is unloaded. The function
6051 should call @var{func}, which takes no arguments. If not defined, and
6052 the object format requires an explicit finalization function, then a
6053 function called @code{_GLOBAL__DD} will be generated.
6056 @defmac INVOKE__main
6057 If defined, @code{main} will call @code{__main} despite the presence of
6058 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6059 where the init section is not actually run automatically, but is still
6060 useful for collecting the lists of constructors and destructors.
6063 @defmac SUPPORTS_INIT_PRIORITY
6064 If nonzero, the C++ @code{init_priority} attribute is supported and the
6065 compiler should emit instructions to control the order of initialization
6066 of objects. If zero, the compiler will issue an error message upon
6067 encountering an @code{init_priority} attribute.
6070 @hook TARGET_HAVE_CTORS_DTORS
6072 @hook TARGET_DTORS_FROM_CXA_ATEXIT
6074 @hook TARGET_ASM_CONSTRUCTOR
6076 @hook TARGET_ASM_DESTRUCTOR
6078 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
6079 generated for the generated object file will have static linkage.
6081 If your system uses @command{collect2} as the means of processing
6082 constructors, then that program normally uses @command{nm} to scan
6083 an object file for constructor functions to be called.
6085 On certain kinds of systems, you can define this macro to make
6086 @command{collect2} work faster (and, in some cases, make it work at all):
6088 @defmac OBJECT_FORMAT_COFF
6089 Define this macro if the system uses COFF (Common Object File Format)
6090 object files, so that @command{collect2} can assume this format and scan
6091 object files directly for dynamic constructor/destructor functions.
6093 This macro is effective only in a native compiler; @command{collect2} as
6094 part of a cross compiler always uses @command{nm} for the target machine.
6097 @defmac REAL_NM_FILE_NAME
6098 Define this macro as a C string constant containing the file name to use
6099 to execute @command{nm}. The default is to search the path normally for
6104 @command{collect2} calls @command{nm} to scan object files for static
6105 constructors and destructors and LTO info. By default, @option{-n} is
6106 passed. Define @code{NM_FLAGS} to a C string constant if other options
6107 are needed to get the same output format as GNU @command{nm -n}
6111 If your system supports shared libraries and has a program to list the
6112 dynamic dependencies of a given library or executable, you can define
6113 these macros to enable support for running initialization and
6114 termination functions in shared libraries:
6117 Define this macro to a C string constant containing the name of the program
6118 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
6121 @defmac PARSE_LDD_OUTPUT (@var{ptr})
6122 Define this macro to be C code that extracts filenames from the output
6123 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
6124 of type @code{char *} that points to the beginning of a line of output
6125 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6126 code must advance @var{ptr} to the beginning of the filename on that
6127 line. Otherwise, it must set @var{ptr} to @code{NULL}.
6130 @defmac SHLIB_SUFFIX
6131 Define this macro to a C string constant containing the default shared
6132 library extension of the target (e.g., @samp{".so"}). @command{collect2}
6133 strips version information after this suffix when generating global
6134 constructor and destructor names. This define is only needed on targets
6135 that use @command{collect2} to process constructors and destructors.
6138 @node Instruction Output
6139 @subsection Output of Assembler Instructions
6141 @c prevent bad page break with this line
6142 This describes assembler instruction output.
6144 @defmac REGISTER_NAMES
6145 A C initializer containing the assembler's names for the machine
6146 registers, each one as a C string constant. This is what translates
6147 register numbers in the compiler into assembler language.
6150 @defmac ADDITIONAL_REGISTER_NAMES
6151 If defined, a C initializer for an array of structures containing a name
6152 and a register number. This macro defines additional names for hard
6153 registers, thus allowing the @code{asm} option in declarations to refer
6154 to registers using alternate names.
6157 @defmac OVERLAPPING_REGISTER_NAMES
6158 If defined, a C initializer for an array of structures containing a
6159 name, a register number and a count of the number of consecutive
6160 machine registers the name overlaps. This macro defines additional
6161 names for hard registers, thus allowing the @code{asm} option in
6162 declarations to refer to registers using alternate names. Unlike
6163 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
6164 register name implies multiple underlying registers.
6166 This macro should be used when it is important that a clobber in an
6167 @code{asm} statement clobbers all the underlying values implied by the
6168 register name. For example, on ARM, clobbering the double-precision
6169 VFP register ``d0'' implies clobbering both single-precision registers
6173 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6174 Define this macro if you are using an unusual assembler that
6175 requires different names for the machine instructions.
6177 The definition is a C statement or statements which output an
6178 assembler instruction opcode to the stdio stream @var{stream}. The
6179 macro-operand @var{ptr} is a variable of type @code{char *} which
6180 points to the opcode name in its ``internal'' form---the form that is
6181 written in the machine description. The definition should output the
6182 opcode name to @var{stream}, performing any translation you desire, and
6183 increment the variable @var{ptr} to point at the end of the opcode
6184 so that it will not be output twice.
6186 In fact, your macro definition may process less than the entire opcode
6187 name, or more than the opcode name; but if you want to process text
6188 that includes @samp{%}-sequences to substitute operands, you must take
6189 care of the substitution yourself. Just be sure to increment
6190 @var{ptr} over whatever text should not be output normally.
6192 @findex recog_data.operand
6193 If you need to look at the operand values, they can be found as the
6194 elements of @code{recog_data.operand}.
6196 If the macro definition does nothing, the instruction is output
6200 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6201 If defined, a C statement to be executed just prior to the output of
6202 assembler code for @var{insn}, to modify the extracted operands so
6203 they will be output differently.
6205 Here the argument @var{opvec} is the vector containing the operands
6206 extracted from @var{insn}, and @var{noperands} is the number of
6207 elements of the vector which contain meaningful data for this insn.
6208 The contents of this vector are what will be used to convert the insn
6209 template into assembler code, so you can change the assembler output
6210 by changing the contents of the vector.
6212 This macro is useful when various assembler syntaxes share a single
6213 file of instruction patterns; by defining this macro differently, you
6214 can cause a large class of instructions to be output differently (such
6215 as with rearranged operands). Naturally, variations in assembler
6216 syntax affecting individual insn patterns ought to be handled by
6217 writing conditional output routines in those patterns.
6219 If this macro is not defined, it is equivalent to a null statement.
6222 @hook TARGET_ASM_FINAL_POSTSCAN_INSN
6224 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6225 A C compound statement to output to stdio stream @var{stream} the
6226 assembler syntax for an instruction operand @var{x}. @var{x} is an
6229 @var{code} is a value that can be used to specify one of several ways
6230 of printing the operand. It is used when identical operands must be
6231 printed differently depending on the context. @var{code} comes from
6232 the @samp{%} specification that was used to request printing of the
6233 operand. If the specification was just @samp{%@var{digit}} then
6234 @var{code} is 0; if the specification was @samp{%@var{ltr}
6235 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6238 If @var{x} is a register, this macro should print the register's name.
6239 The names can be found in an array @code{reg_names} whose type is
6240 @code{char *[]}. @code{reg_names} is initialized from
6241 @code{REGISTER_NAMES}.
6243 When the machine description has a specification @samp{%@var{punct}}
6244 (a @samp{%} followed by a punctuation character), this macro is called
6245 with a null pointer for @var{x} and the punctuation character for
6249 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6250 A C expression which evaluates to true if @var{code} is a valid
6251 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6252 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6253 punctuation characters (except for the standard one, @samp{%}) are used
6257 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6258 A C compound statement to output to stdio stream @var{stream} the
6259 assembler syntax for an instruction operand that is a memory reference
6260 whose address is @var{x}. @var{x} is an RTL expression.
6262 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
6263 On some machines, the syntax for a symbolic address depends on the
6264 section that the address refers to. On these machines, define the hook
6265 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
6266 @code{symbol_ref}, and then check for it here. @xref{Assembler
6270 @findex dbr_sequence_length
6271 @defmac DBR_OUTPUT_SEQEND (@var{file})
6272 A C statement, to be executed after all slot-filler instructions have
6273 been output. If necessary, call @code{dbr_sequence_length} to
6274 determine the number of slots filled in a sequence (zero if not
6275 currently outputting a sequence), to decide how many no-ops to output,
6278 Don't define this macro if it has nothing to do, but it is helpful in
6279 reading assembly output if the extent of the delay sequence is made
6280 explicit (e.g.@: with white space).
6283 @findex final_sequence
6284 Note that output routines for instructions with delay slots must be
6285 prepared to deal with not being output as part of a sequence
6286 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
6287 found.) The variable @code{final_sequence} is null when not
6288 processing a sequence, otherwise it contains the @code{sequence} rtx
6292 @defmac REGISTER_PREFIX
6293 @defmacx LOCAL_LABEL_PREFIX
6294 @defmacx USER_LABEL_PREFIX
6295 @defmacx IMMEDIATE_PREFIX
6296 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6297 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6298 @file{final.cc}). These are useful when a single @file{md} file must
6299 support multiple assembler formats. In that case, the various @file{tm.h}
6300 files can define these macros differently.
6303 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
6304 If defined this macro should expand to a series of @code{case}
6305 statements which will be parsed inside the @code{switch} statement of
6306 the @code{asm_fprintf} function. This allows targets to define extra
6307 printf formats which may useful when generating their assembler
6308 statements. Note that uppercase letters are reserved for future
6309 generic extensions to asm_fprintf, and so are not available to target
6310 specific code. The output file is given by the parameter @var{file}.
6311 The varargs input pointer is @var{argptr} and the rest of the format
6312 string, starting the character after the one that is being switched
6313 upon, is pointed to by @var{format}.
6316 @defmac ASSEMBLER_DIALECT
6317 If your target supports multiple dialects of assembler language (such as
6318 different opcodes), define this macro as a C expression that gives the
6319 numeric index of the assembler language dialect to use, with zero as the
6322 If this macro is defined, you may use constructs of the form
6324 @samp{@{option0|option1|option2@dots{}@}}
6327 in the output templates of patterns (@pxref{Output Template}) or in the
6328 first argument of @code{asm_fprintf}. This construct outputs
6329 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
6330 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
6331 within these strings retain their usual meaning. If there are fewer
6332 alternatives within the braces than the value of
6333 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
6334 to print curly braces or @samp{|} character in assembler output directly,
6335 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
6337 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6338 @samp{@}} do not have any special meaning when used in templates or
6339 operands to @code{asm_fprintf}.
6341 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6342 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6343 the variations in assembler language syntax with that mechanism. Define
6344 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6345 if the syntax variant are larger and involve such things as different
6346 opcodes or operand order.
6349 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6350 A C expression to output to @var{stream} some assembler code
6351 which will push hard register number @var{regno} onto the stack.
6352 The code need not be optimal, since this macro is used only when
6356 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6357 A C expression to output to @var{stream} some assembler code
6358 which will pop hard register number @var{regno} off of the stack.
6359 The code need not be optimal, since this macro is used only when
6363 @node Dispatch Tables
6364 @subsection Output of Dispatch Tables
6366 @c prevent bad page break with this line
6367 This concerns dispatch tables.
6369 @cindex dispatch table
6370 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6371 A C statement to output to the stdio stream @var{stream} an assembler
6372 pseudo-instruction to generate a difference between two labels.
6373 @var{value} and @var{rel} are the numbers of two internal labels. The
6374 definitions of these labels are output using
6375 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
6376 way here. For example,
6379 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6380 @var{value}, @var{rel})
6383 You must provide this macro on machines where the addresses in a
6384 dispatch table are relative to the table's own address. If defined, GCC
6385 will also use this macro on all machines when producing PIC@.
6386 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
6387 mode and flags can be read.
6390 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6391 This macro should be provided on machines where the addresses
6392 in a dispatch table are absolute.
6394 The definition should be a C statement to output to the stdio stream
6395 @var{stream} an assembler pseudo-instruction to generate a reference to
6396 a label. @var{value} is the number of an internal label whose
6397 definition is output using @code{(*targetm.asm_out.internal_label)}.
6401 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6405 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6406 Define this if the label before a jump-table needs to be output
6407 specially. The first three arguments are the same as for
6408 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
6409 jump-table which follows (a @code{jump_table_data} containing an
6410 @code{addr_vec} or @code{addr_diff_vec}).
6412 This feature is used on system V to output a @code{swbeg} statement
6415 If this macro is not defined, these labels are output with
6416 @code{(*targetm.asm_out.internal_label)}.
6419 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6420 Define this if something special must be output at the end of a
6421 jump-table. The definition should be a C statement to be executed
6422 after the assembler code for the table is written. It should write
6423 the appropriate code to stdio stream @var{stream}. The argument
6424 @var{table} is the jump-table insn, and @var{num} is the label-number
6425 of the preceding label.
6427 If this macro is not defined, nothing special is output at the end of
6431 @hook TARGET_ASM_POST_CFI_STARTPROC
6433 @hook TARGET_ASM_EMIT_UNWIND_LABEL
6435 @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
6437 @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
6439 @hook TARGET_ASM_UNWIND_EMIT
6441 @hook TARGET_ASM_MAKE_EH_SYMBOL_INDIRECT
6443 @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
6445 @hook TARGET_ASM_SHOULD_RESTORE_CFA_STATE
6447 @node Exception Region Output
6448 @subsection Assembler Commands for Exception Regions
6450 @c prevent bad page break with this line
6452 This describes commands marking the start and the end of an exception
6455 @defmac EH_FRAME_SECTION_NAME
6456 If defined, a C string constant for the name of the section containing
6457 exception handling frame unwind information. If not defined, GCC will
6458 provide a default definition if the target supports named sections.
6459 @file{crtstuff.c} uses this macro to switch to the appropriate section.
6461 You should define this symbol if your target supports DWARF 2 frame
6462 unwind information and the default definition does not work.
6465 @defmac EH_FRAME_THROUGH_COLLECT2
6466 If defined, DWARF 2 frame unwind information will identified by
6467 specially named labels. The collect2 process will locate these
6468 labels and generate code to register the frames.
6470 This might be necessary, for instance, if the system linker will not
6471 place the eh_frames in-between the sentinals from @file{crtstuff.c},
6472 or if the system linker does garbage collection and sections cannot
6473 be marked as not to be collected.
6476 @defmac EH_TABLES_CAN_BE_READ_ONLY
6477 Define this macro to 1 if your target is such that no frame unwind
6478 information encoding used with non-PIC code will ever require a
6479 runtime relocation, but the linker may not support merging read-only
6480 and read-write sections into a single read-write section.
6483 @defmac MASK_RETURN_ADDR
6484 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
6485 that it does not contain any extraneous set bits in it.
6488 @defmac DWARF2_UNWIND_INFO
6489 Define this macro to 0 if your target supports DWARF 2 frame unwind
6490 information, but it does not yet work with exception handling.
6491 Otherwise, if your target supports this information (if it defines
6492 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
6493 GCC will provide a default definition of 1.
6496 @hook TARGET_EXCEPT_UNWIND_INFO
6497 This hook defines the mechanism that will be used for exception handling
6498 by the target. If the target has ABI specified unwind tables, the hook
6499 should return @code{UI_TARGET}. If the target is to use the
6500 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
6501 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
6502 information, the hook should return @code{UI_DWARF2}.
6504 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
6505 This may end up simplifying other parts of target-specific code. The
6506 default implementation of this hook never returns @code{UI_NONE}.
6508 Note that the value returned by this hook should be constant. It should
6509 not depend on anything except the command-line switches described by
6510 @var{opts}. In particular, the
6511 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
6512 macros and builtin functions related to exception handling are set up
6513 depending on this setting.
6515 The default implementation of the hook first honors the
6516 @option{--enable-sjlj-exceptions} configure option, then
6517 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
6518 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
6519 must define this hook so that @var{opts} is used correctly.
6522 @hook TARGET_UNWIND_TABLES_DEFAULT
6523 This variable should be set to @code{true} if the target ABI requires unwinding
6524 tables even when exceptions are not used. It must not be modified by
6525 command-line option processing.
6528 @defmac DONT_USE_BUILTIN_SETJMP
6529 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
6530 should use the @code{setjmp}/@code{longjmp} functions from the C library
6531 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
6534 @defmac JMP_BUF_SIZE
6535 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
6536 defined. Define this macro if the default size of @code{jmp_buf} buffer
6537 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
6538 is not large enough, or if it is much too large.
6539 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
6542 @defmac DWARF_CIE_DATA_ALIGNMENT
6543 This macro need only be defined if the target might save registers in the
6544 function prologue at an offset to the stack pointer that is not aligned to
6545 @code{UNITS_PER_WORD}. The definition should be the negative minimum
6546 alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive
6547 minimum alignment otherwise. @xref{DWARF}. Only applicable if
6548 the target supports DWARF 2 frame unwind information.
6551 @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
6553 @hook TARGET_DWARF_REGISTER_SPAN
6555 @hook TARGET_DWARF_FRAME_REG_MODE
6557 @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
6559 @hook TARGET_ASM_TTYPE
6561 @hook TARGET_ARM_EABI_UNWINDER
6563 @node Alignment Output
6564 @subsection Assembler Commands for Alignment
6566 @c prevent bad page break with this line
6567 This describes commands for alignment.
6569 @defmac JUMP_ALIGN (@var{label})
6570 The alignment (log base 2) to put in front of @var{label}, which is
6571 a common destination of jumps and has no fallthru incoming edge.
6573 This macro need not be defined if you don't want any special alignment
6574 to be done at such a time. Most machine descriptions do not currently
6577 Unless it's necessary to inspect the @var{label} parameter, it is better
6578 to set the variable @var{align_jumps} in the target's
6579 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6580 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
6583 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
6584 The alignment (log base 2) to put in front of @var{label}, which follows
6587 This macro need not be defined if you don't want any special alignment
6588 to be done at such a time. Most machine descriptions do not currently
6592 @defmac LOOP_ALIGN (@var{label})
6593 The alignment (log base 2) to put in front of @var{label} that heads
6594 a frequently executed basic block (usually the header of a loop).
6596 This macro need not be defined if you don't want any special alignment
6597 to be done at such a time. Most machine descriptions do not currently
6600 Unless it's necessary to inspect the @var{label} parameter, it is better
6601 to set the variable @code{align_loops} in the target's
6602 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6603 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
6606 @defmac LABEL_ALIGN (@var{label})
6607 The alignment (log base 2) to put in front of @var{label}.
6608 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
6609 the maximum of the specified values is used.
6611 Unless it's necessary to inspect the @var{label} parameter, it is better
6612 to set the variable @code{align_labels} in the target's
6613 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6614 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
6617 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6618 A C statement to output to the stdio stream @var{stream} an assembler
6619 instruction to advance the location counter by @var{nbytes} bytes.
6620 Those bytes should be zero when loaded. @var{nbytes} will be a C
6621 expression of type @code{unsigned HOST_WIDE_INT}.
6624 @defmac ASM_NO_SKIP_IN_TEXT
6625 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6626 text section because it fails to put zeros in the bytes that are skipped.
6627 This is true on many Unix systems, where the pseudo--op to skip bytes
6628 produces no-op instructions rather than zeros when used in the text
6632 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6633 A C statement to output to the stdio stream @var{stream} an assembler
6634 command to advance the location counter to a multiple of 2 to the
6635 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6638 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
6639 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
6640 for padding, if necessary.
6643 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6644 A C statement to output to the stdio stream @var{stream} an assembler
6645 command to advance the location counter to a multiple of 2 to the
6646 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6647 satisfy the alignment request. @var{power} and @var{max_skip} will be
6648 a C expression of type @code{int}.
6652 @node Debugging Info
6653 @section Controlling Debugging Information Format
6655 @c prevent bad page break with this line
6656 This describes how to specify debugging information.
6659 * All Debuggers:: Macros that affect all debugging formats uniformly.
6660 * DWARF:: Macros for DWARF format.
6661 * VMS Debug:: Macros for VMS debug format.
6662 * CTF Debug:: Macros for CTF debug format.
6663 * BTF Debug:: Macros for BTF debug format.
6667 @subsection Macros Affecting All Debugging Formats
6669 @c prevent bad page break with this line
6670 These macros affect all debugging formats.
6672 @defmac DEBUGGER_REGNO (@var{regno})
6673 A C expression that returns the debugger register number for the compiler
6674 register number @var{regno}. In the default macro provided, the value
6675 of this expression will be @var{regno} itself. But sometimes there are
6676 some registers that the compiler knows about and debugger does not, or vice
6677 versa. In such cases, some register may need to have one number in the
6678 compiler and another for debugger@.
6680 If two registers have consecutive numbers inside GCC, and they can be
6681 used as a pair to hold a multiword value, then they @emph{must} have
6682 consecutive numbers after renumbering with @code{DEBUGGER_REGNO}.
6683 Otherwise, debuggers will be unable to access such a pair, because they
6684 expect register pairs to be consecutive in their own numbering scheme.
6686 If you find yourself defining @code{DEBUGGER_REGNO} in way that
6687 does not preserve register pairs, then what you must do instead is
6688 redefine the actual register numbering scheme.
6691 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
6692 A C expression that returns the integer offset value for an automatic
6693 variable having address @var{x} (an RTL expression). The default
6694 computation assumes that @var{x} is based on the frame-pointer and
6695 gives the offset from the frame-pointer. This is required for targets
6696 that produce debugging output for debugger and allow the frame-pointer to be
6697 eliminated when the @option{-g} option is used.
6700 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6701 A C expression that returns the integer offset value for an argument
6702 having address @var{x} (an RTL expression). The nominal offset is
6706 @defmac PREFERRED_DEBUGGING_TYPE
6707 A C expression that returns the type of debugging output GCC should
6708 produce when the user specifies just @option{-g}. Define
6709 this if you have arranged for GCC to support more than one format of
6710 debugging output. Currently, the allowable values are
6711 @code{DWARF2_DEBUG}, @code{VMS_DEBUG},
6712 and @code{VMS_AND_DWARF2_DEBUG}.
6714 When the user specifies @option{-ggdb}, GCC normally also uses the
6715 value of this macro to select the debugging output format, but with two
6716 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
6717 value @code{DWARF2_DEBUG}.
6719 The value of this macro only affects the default debugging output; the
6720 user can always get a specific type of output by using @option{-gdwarf-2},
6724 @defmac DEFAULT_GDB_EXTENSIONS
6725 Define this macro to control whether GCC should by default generate
6726 GDB's extended version of debugging information. If you don't define the
6727 macro, the default is 1: always generate the extended information
6728 if there is any occasion to.
6733 @subsection Macros for DWARF Output
6735 @c prevent bad page break with this line
6736 Here are macros for DWARF output.
6738 @defmac DWARF2_DEBUGGING_INFO
6739 Define this macro if GCC should produce dwarf version 2 format
6740 debugging output in response to the @option{-g} option.
6742 To support optional call frame debugging information, you must also
6743 define @code{INCOMING_RETURN_ADDR_RTX} and either set
6744 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
6745 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
6746 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
6749 @hook TARGET_DWARF_CALLING_CONVENTION
6751 @defmac DWARF2_FRAME_INFO
6752 Define this macro to a nonzero value if GCC should always output
6753 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
6754 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
6755 exceptions are enabled, GCC will output this information not matter
6756 how you define @code{DWARF2_FRAME_INFO}.
6759 @hook TARGET_DEBUG_UNWIND_INFO
6761 @defmac DWARF2_ASM_LINE_DEBUG_INFO
6762 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
6763 line debug info sections. This will result in much more compact line number
6764 tables, and hence is desirable if it works.
6767 @defmac DWARF2_ASM_VIEW_DEBUG_INFO
6768 Define this macro to be a nonzero value if the assembler supports view
6769 assignment and verification in @code{.loc}. If it does not, but the
6770 user enables location views, the compiler may have to fallback to
6771 internal line number tables.
6774 @hook TARGET_RESET_LOCATION_VIEW
6776 @hook TARGET_WANT_DEBUG_PUB_SECTIONS
6778 @hook TARGET_DELAY_SCHED2
6780 @hook TARGET_DELAY_VARTRACK
6782 @hook TARGET_NO_REGISTER_ALLOCATION
6784 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
6785 A C statement to issue assembly directives that create a difference
6786 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
6789 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
6790 A C statement to issue assembly directives that create a difference
6791 between the two given labels in system defined units, e.g.@: instruction
6792 slots on IA64 VMS, using an integer of the given size.
6795 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{offset}, @var{section})
6796 A C statement to issue assembly directives that create a
6797 section-relative reference to the given @var{label} plus @var{offset}, using
6798 an integer of the given @var{size}. The label is known to be defined in the
6799 given @var{section}.
6802 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
6803 A C statement to issue assembly directives that create a self-relative
6804 reference to the given @var{label}, using an integer of the given @var{size}.
6807 @defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label})
6808 A C statement to issue assembly directives that create a reference to the
6809 given @var{label} relative to the dbase, using an integer of the given @var{size}.
6812 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
6813 A C statement to issue assembly directives that create a reference to
6814 the DWARF table identifier @var{label} from the current section. This
6815 is used on some systems to avoid garbage collecting a DWARF table which
6816 is referenced by a function.
6819 @hook TARGET_ASM_OUTPUT_DWARF_DTPREL
6823 @subsection Macros for VMS Debug Format
6825 @c prevent bad page break with this line
6826 Here are macros for VMS debug format.
6828 @defmac VMS_DEBUGGING_INFO
6829 Define this macro if GCC should produce debugging output for VMS
6830 in response to the @option{-g} option. The default behavior for VMS
6831 is to generate minimal debug info for a traceback in the absence of
6832 @option{-g} unless explicitly overridden with @option{-g0}. This
6833 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
6834 @code{TARGET_OPTION_OVERRIDE}.
6839 @subsection Macros for CTF Debug Format
6841 @c prevent bad page break with this line
6842 Here are macros for CTF debug format.
6844 @defmac CTF_DEBUGGING_INFO
6845 Define this macro if GCC should produce debugging output in CTF debug
6846 format in response to the @option{-gctf} option.
6851 @subsection Macros for BTF Debug Format
6853 @c prevent bad page break with this line
6854 Here are macros for BTF debug format.
6856 @defmac BTF_DEBUGGING_INFO
6857 Define this macro if GCC should produce debugging output in BTF debug
6858 format in response to the @option{-gbtf} option.
6861 @node Floating Point
6862 @section Cross Compilation and Floating Point
6863 @cindex cross compilation and floating point
6864 @cindex floating point and cross compilation
6866 While all modern machines use twos-complement representation for integers,
6867 there are a variety of representations for floating point numbers. This
6868 means that in a cross-compiler the representation of floating point numbers
6869 in the compiled program may be different from that used in the machine
6870 doing the compilation.
6872 Because different representation systems may offer different amounts of
6873 range and precision, all floating point constants must be represented in
6874 the target machine's format. Therefore, the cross compiler cannot
6875 safely use the host machine's floating point arithmetic; it must emulate
6876 the target's arithmetic. To ensure consistency, GCC always uses
6877 emulation to work with floating point values, even when the host and
6878 target floating point formats are identical.
6880 The following macros are provided by @file{real.h} for the compiler to
6881 use. All parts of the compiler which generate or optimize
6882 floating-point calculations must use these macros. They may evaluate
6883 their operands more than once, so operands must not have side effects.
6885 @defmac REAL_VALUE_TYPE
6886 The C data type to be used to hold a floating point value in the target
6887 machine's format. Typically this is a @code{struct} containing an
6888 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
6892 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
6893 Truncates @var{x} to a signed integer, rounding toward zero.
6896 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
6897 Truncates @var{x} to an unsigned integer, rounding toward zero. If
6898 @var{x} is negative, returns zero.
6901 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode})
6902 Converts @var{string} into a floating point number in the target machine's
6903 representation for mode @var{mode}. This routine can handle both
6904 decimal and hexadecimal floating point constants, using the syntax
6905 defined by the C language for both.
6908 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
6909 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
6912 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
6913 Determines whether @var{x} represents infinity (positive or negative).
6916 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
6917 Determines whether @var{x} represents a ``NaN'' (not-a-number).
6920 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
6921 Returns the negative of the floating point value @var{x}.
6924 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
6925 Returns the absolute value of @var{x}.
6928 @node Mode Switching
6929 @section Mode Switching Instructions
6930 @cindex mode switching
6931 The following macros control mode switching optimizations:
6933 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
6934 Define this macro if the port needs extra instructions inserted for mode
6935 switching in an optimizing compilation.
6937 For an example, the SH4 can perform both single and double precision
6938 floating point operations, but to perform a single precision operation,
6939 the FPSCR PR bit has to be cleared, while for a double precision
6940 operation, this bit has to be set. Changing the PR bit requires a general
6941 purpose register as a scratch register, hence these FPSCR sets have to
6942 be inserted before reload, i.e.@: you cannot put this into instruction emitting
6943 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
6945 You can have multiple entities that are mode-switched, and select at run time
6946 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
6947 return nonzero for any @var{entity} that needs mode-switching.
6948 If you define this macro, you also have to define
6949 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
6950 @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
6951 @code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
6955 @defmac NUM_MODES_FOR_MODE_SWITCHING
6956 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
6957 initializer for an array of integers. Each initializer element
6958 N refers to an entity that needs mode switching, and specifies the number
6959 of different modes that might need to be set for this entity.
6960 The position of the initializer in the initializer---starting counting at
6961 zero---determines the integer that is used to refer to the mode-switched
6963 In macros that take mode arguments / yield a mode result, modes are
6964 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
6965 switch is needed / supplied.
6968 @hook TARGET_MODE_EMIT
6970 @hook TARGET_MODE_NEEDED
6972 @hook TARGET_MODE_AFTER
6974 @hook TARGET_MODE_ENTRY
6976 @hook TARGET_MODE_EXIT
6978 @hook TARGET_MODE_PRIORITY
6980 @node Target Attributes
6981 @section Defining target-specific uses of @code{__attribute__}
6982 @cindex target attributes
6983 @cindex machine attributes
6984 @cindex attributes, target-specific
6986 Target-specific attributes may be defined for functions, data and types.
6987 These are described using the following target hooks; they also need to
6988 be documented in @file{extend.texi}.
6990 @hook TARGET_ATTRIBUTE_TABLE
6992 @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
6994 @hook TARGET_COMP_TYPE_ATTRIBUTES
6996 @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
6998 @hook TARGET_MERGE_TYPE_ATTRIBUTES
7000 @hook TARGET_MERGE_DECL_ATTRIBUTES
7002 @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
7004 @defmac TARGET_DECLSPEC
7005 Define this macro to a nonzero value if you want to treat
7006 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
7007 default, this behavior is enabled only for targets that define
7008 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
7009 of @code{__declspec} is via a built-in macro, but you should not rely
7010 on this implementation detail.
7013 @hook TARGET_INSERT_ATTRIBUTES
7015 @hook TARGET_HANDLE_GENERIC_ATTRIBUTE
7017 @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
7019 @hook TARGET_OPTION_VALID_ATTRIBUTE_P
7021 @hook TARGET_OPTION_SAVE
7023 @hook TARGET_OPTION_RESTORE
7025 @hook TARGET_OPTION_POST_STREAM_IN
7027 @hook TARGET_OPTION_PRINT
7029 @hook TARGET_OPTION_PRAGMA_PARSE
7031 @hook TARGET_OPTION_OVERRIDE
7033 @hook TARGET_OPTION_FUNCTION_VERSIONS
7035 @hook TARGET_CAN_INLINE_P
7037 @hook TARGET_UPDATE_IPA_FN_TARGET_INFO
7039 @hook TARGET_NEED_IPA_FN_TARGET_INFO
7041 @hook TARGET_RELAYOUT_FUNCTION
7044 @section Emulating TLS
7045 @cindex Emulated TLS
7047 For targets whose psABI does not provide Thread Local Storage via
7048 specific relocations and instruction sequences, an emulation layer is
7049 used. A set of target hooks allows this emulation layer to be
7050 configured for the requirements of a particular target. For instance
7051 the psABI may in fact specify TLS support in terms of an emulation
7054 The emulation layer works by creating a control object for every TLS
7055 object. To access the TLS object, a lookup function is provided
7056 which, when given the address of the control object, will return the
7057 address of the current thread's instance of the TLS object.
7059 @hook TARGET_EMUTLS_GET_ADDRESS
7061 @hook TARGET_EMUTLS_REGISTER_COMMON
7063 @hook TARGET_EMUTLS_VAR_SECTION
7065 @hook TARGET_EMUTLS_TMPL_SECTION
7067 @hook TARGET_EMUTLS_VAR_PREFIX
7069 @hook TARGET_EMUTLS_TMPL_PREFIX
7071 @hook TARGET_EMUTLS_VAR_FIELDS
7073 @hook TARGET_EMUTLS_VAR_INIT
7075 @hook TARGET_EMUTLS_VAR_ALIGN_FIXED
7077 @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
7079 @node MIPS Coprocessors
7080 @section Defining coprocessor specifics for MIPS targets.
7081 @cindex MIPS coprocessor-definition macros
7083 The MIPS specification allows MIPS implementations to have as many as 4
7084 coprocessors, each with as many as 32 private registers. GCC supports
7085 accessing these registers and transferring values between the registers
7086 and memory using asm-ized variables. For example:
7089 register unsigned int cp0count asm ("c0r1");
7095 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
7096 names may be added as described below, or the default names may be
7097 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
7099 Coprocessor registers are assumed to be epilogue-used; sets to them will
7100 be preserved even if it does not appear that the register is used again
7101 later in the function.
7103 Another note: according to the MIPS spec, coprocessor 1 (if present) is
7104 the FPU@. One accesses COP1 registers through standard mips
7105 floating-point support; they are not included in this mechanism.
7108 @section Parameters for Precompiled Header Validity Checking
7109 @cindex parameters, precompiled headers
7111 @hook TARGET_GET_PCH_VALIDITY
7113 @hook TARGET_PCH_VALID_P
7115 @hook TARGET_CHECK_PCH_TARGET_FLAGS
7117 @hook TARGET_PREPARE_PCH_SAVE
7120 @section C++ ABI parameters
7121 @cindex parameters, c++ abi
7123 @hook TARGET_CXX_GUARD_TYPE
7125 @hook TARGET_CXX_GUARD_MASK_BIT
7127 @hook TARGET_CXX_GET_COOKIE_SIZE
7129 @hook TARGET_CXX_COOKIE_HAS_SIZE
7131 @hook TARGET_CXX_IMPORT_EXPORT_CLASS
7133 @hook TARGET_CXX_CDTOR_RETURNS_THIS
7135 @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
7137 @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
7139 @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
7141 @hook TARGET_CXX_LIBRARY_RTTI_COMDAT
7143 @hook TARGET_CXX_USE_AEABI_ATEXIT
7145 @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
7147 @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
7149 @hook TARGET_CXX_DECL_MANGLING_CONTEXT
7151 @node D Language and ABI
7152 @section D ABI parameters
7153 @cindex parameters, d abi
7155 @hook TARGET_D_CPU_VERSIONS
7157 @hook TARGET_D_OS_VERSIONS
7159 @hook TARGET_D_REGISTER_CPU_TARGET_INFO
7161 @hook TARGET_D_REGISTER_OS_TARGET_INFO
7163 @hook TARGET_D_MINFO_SECTION
7165 @hook TARGET_D_MINFO_SECTION_START
7167 @hook TARGET_D_MINFO_SECTION_END
7169 @hook TARGET_D_HAS_STDCALL_CONVENTION
7171 @hook TARGET_D_TEMPLATES_ALWAYS_COMDAT
7173 @node Rust Language and ABI
7174 @section Rust ABI parameters
7175 @cindex parameters, rust abi
7177 @hook TARGET_RUST_CPU_INFO
7179 @hook TARGET_RUST_OS_INFO
7181 @node Named Address Spaces
7182 @section Adding support for named address spaces
7183 @cindex named address spaces
7185 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
7186 standards committee, @cite{Programming Languages - C - Extensions to
7187 support embedded processors}, specifies a syntax for embedded
7188 processors to specify alternate address spaces. You can configure a
7189 GCC port to support section 5.1 of the draft report to add support for
7190 address spaces other than the default address space. These address
7191 spaces are new keywords that are similar to the @code{volatile} and
7192 @code{const} type attributes.
7194 Pointers to named address spaces can have a different size than
7195 pointers to the generic address space.
7197 For example, the SPU port uses the @code{__ea} address space to refer
7198 to memory in the host processor, rather than memory local to the SPU
7199 processor. Access to memory in the @code{__ea} address space involves
7200 issuing DMA operations to move data between the host processor and the
7201 local processor memory address space. Pointers in the @code{__ea}
7202 address space are either 32 bits or 64 bits based on the
7203 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
7206 Internally, address spaces are represented as a small integer in the
7207 range 0 to 15 with address space 0 being reserved for the generic
7210 To register a named address space qualifier keyword with the C front end,
7211 the target may call the @code{c_register_addr_space} routine. For example,
7212 the SPU port uses the following to declare @code{__ea} as the keyword for
7213 named address space #1:
7215 #define ADDR_SPACE_EA 1
7216 c_register_addr_space ("__ea", ADDR_SPACE_EA);
7219 @hook TARGET_ADDR_SPACE_POINTER_MODE
7221 @hook TARGET_ADDR_SPACE_ADDRESS_MODE
7223 @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
7225 @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
7227 @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
7229 @hook TARGET_ADDR_SPACE_SUBSET_P
7231 @hook TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID
7233 @hook TARGET_ADDR_SPACE_CONVERT
7235 @hook TARGET_ADDR_SPACE_DEBUG
7237 @hook TARGET_ADDR_SPACE_DIAGNOSE_USAGE
7240 @section Miscellaneous Parameters
7241 @cindex parameters, miscellaneous
7243 @c prevent bad page break with this line
7244 Here are several miscellaneous parameters.
7246 @defmac HAS_LONG_COND_BRANCH
7247 Define this boolean macro to indicate whether or not your architecture
7248 has conditional branches that can span all of memory. It is used in
7249 conjunction with an optimization that partitions hot and cold basic
7250 blocks into separate sections of the executable. If this macro is
7251 set to false, gcc will convert any conditional branches that attempt
7252 to cross between sections into unconditional branches or indirect jumps.
7255 @defmac HAS_LONG_UNCOND_BRANCH
7256 Define this boolean macro to indicate whether or not your architecture
7257 has unconditional branches that can span all of memory. It is used in
7258 conjunction with an optimization that partitions hot and cold basic
7259 blocks into separate sections of the executable. If this macro is
7260 set to false, gcc will convert any unconditional branches that attempt
7261 to cross between sections into indirect jumps.
7264 @defmac CASE_VECTOR_MODE
7265 An alias for a machine mode name. This is the machine mode that
7266 elements of a jump-table should have.
7269 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7270 Optional: return the preferred mode for an @code{addr_diff_vec}
7271 when the minimum and maximum offset are known. If you define this,
7272 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7273 To make this work, you also have to define @code{INSN_ALIGN} and
7274 make the alignment for @code{addr_diff_vec} explicit.
7275 The @var{body} argument is provided so that the offset_unsigned and scale
7276 flags can be updated.
7279 @defmac CASE_VECTOR_PC_RELATIVE
7280 Define this macro to be a C expression to indicate when jump-tables
7281 should contain relative addresses. You need not define this macro if
7282 jump-tables never contain relative addresses, or jump-tables should
7283 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
7287 @hook TARGET_CASE_VALUES_THRESHOLD
7289 @defmac WORD_REGISTER_OPERATIONS
7290 Define this macro to 1 if operations between registers with integral mode
7291 smaller than a word are always performed on the entire register. To be
7292 more explicit, if you start with a pair of @code{word_mode} registers with
7293 known values and you do a subword, for example @code{QImode}, addition on
7294 the low part of the registers, then the compiler may consider that the
7295 result has a known value in @code{word_mode} too if the macro is defined
7296 to 1. Most RISC machines have this property and most CISC machines do not.
7299 @hook TARGET_MIN_ARITHMETIC_PRECISION
7301 @defmac LOAD_EXTEND_OP (@var{mem_mode})
7302 Define this macro to be a C expression indicating when insns that read
7303 memory in @var{mem_mode}, an integral mode narrower than a word, set the
7304 bits outside of @var{mem_mode} to be either the sign-extension or the
7305 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7306 of @var{mem_mode} for which the
7307 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7308 @code{UNKNOWN} for other modes.
7310 This macro is not called with @var{mem_mode} non-integral or with a width
7311 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7312 value in this case. Do not define this macro if it would always return
7313 @code{UNKNOWN}. On machines where this macro is defined, you will normally
7314 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7316 You may return a non-@code{UNKNOWN} value even if for some hard registers
7317 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
7318 of these hard registers @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false
7319 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
7320 integral mode larger than this but not larger than @code{word_mode}.
7322 You must return @code{UNKNOWN} if for some hard registers that allow this
7323 mode, @code{TARGET_CAN_CHANGE_MODE_CLASS} says that they cannot change to
7324 @code{word_mode}, but that they can change to another integral mode that
7325 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
7328 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
7329 Define this macro to 1 if loading short immediate values into registers sign
7333 @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
7336 The maximum number of bytes that a single instruction can move quickly
7337 between memory and registers or between two memory locations.
7340 @defmac MAX_MOVE_MAX
7341 The maximum number of bytes that a single instruction can move quickly
7342 between memory and registers or between two memory locations. If this
7343 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7344 constant value that is the largest value that @code{MOVE_MAX} can have
7348 @defmac SHIFT_COUNT_TRUNCATED
7349 A C expression that is nonzero if on this machine the number of bits
7350 actually used for the count of a shift operation is equal to the number
7351 of bits needed to represent the size of the object being shifted. When
7352 this macro is nonzero, the compiler will assume that it is safe to omit
7353 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7354 truncates the count of a shift operation. On machines that have
7355 instructions that act on bit-fields at variable positions, which may
7356 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7357 also enables deletion of truncations of the values that serve as
7358 arguments to bit-field instructions.
7360 If both types of instructions truncate the count (for shifts) and
7361 position (for bit-field operations), or if no variable-position bit-field
7362 instructions exist, you should define this macro.
7364 However, on some machines, such as the 80386 and the 680x0, truncation
7365 only applies to shift operations and not the (real or pretended)
7366 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7367 such machines. Instead, add patterns to the @file{md} file that include
7368 the implied truncation of the shift instructions.
7370 You need not define this macro if it would always have the value of zero.
7373 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
7374 @hook TARGET_SHIFT_TRUNCATION_MASK
7376 @hook TARGET_TRULY_NOOP_TRUNCATION
7378 @hook TARGET_MODE_REP_EXTENDED
7380 @hook TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P
7382 @defmac STORE_FLAG_VALUE
7383 A C expression describing the value returned by a comparison operator
7384 with an integral mode and stored by a store-flag instruction
7385 (@samp{cstore@var{mode}4}) when the condition is true. This description must
7386 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
7387 comparison operators whose results have a @code{MODE_INT} mode.
7389 A value of 1 or @minus{}1 means that the instruction implementing the
7390 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
7391 and 0 when the comparison is false. Otherwise, the value indicates
7392 which bits of the result are guaranteed to be 1 when the comparison is
7393 true. This value is interpreted in the mode of the comparison
7394 operation, which is given by the mode of the first operand in the
7395 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
7396 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7399 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
7400 generate code that depends only on the specified bits. It can also
7401 replace comparison operators with equivalent operations if they cause
7402 the required bits to be set, even if the remaining bits are undefined.
7403 For example, on a machine whose comparison operators return an
7404 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7405 @samp{0x80000000}, saying that just the sign bit is relevant, the
7409 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7416 (ashift:SI @var{x} (const_int @var{n}))
7420 where @var{n} is the appropriate shift count to move the bit being
7421 tested into the sign bit.
7423 There is no way to describe a machine that always sets the low-order bit
7424 for a true value, but does not guarantee the value of any other bits,
7425 but we do not know of any machine that has such an instruction. If you
7426 are trying to port GCC to such a machine, include an instruction to
7427 perform a logical-and of the result with 1 in the pattern for the
7428 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
7430 Often, a machine will have multiple instructions that obtain a value
7431 from a comparison (or the condition codes). Here are rules to guide the
7432 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7437 Use the shortest sequence that yields a valid definition for
7438 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7439 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7440 comparison operators to do so because there may be opportunities to
7441 combine the normalization with other operations.
7444 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
7445 slightly preferred on machines with expensive jumps and 1 preferred on
7449 As a second choice, choose a value of @samp{0x80000001} if instructions
7450 exist that set both the sign and low-order bits but do not define the
7454 Otherwise, use a value of @samp{0x80000000}.
7457 Many machines can produce both the value chosen for
7458 @code{STORE_FLAG_VALUE} and its negation in the same number of
7459 instructions. On those machines, you should also define a pattern for
7460 those cases, e.g., one matching
7463 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7466 Some machines can also perform @code{and} or @code{plus} operations on
7467 condition code values with less instructions than the corresponding
7468 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
7469 machines, define the appropriate patterns. Use the names @code{incscc}
7470 and @code{decscc}, respectively, for the patterns which perform
7471 @code{plus} or @code{minus} operations on condition code values. See
7472 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
7473 find such instruction sequences on other machines.
7475 If this macro is not defined, the default value, 1, is used. You need
7476 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7477 instructions, or if the value generated by these instructions is 1.
7480 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
7481 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
7482 returned when comparison operators with floating-point results are true.
7483 Define this macro on machines that have comparison operations that return
7484 floating-point values. If there are no such operations, do not define
7488 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
7489 A C expression that gives an rtx representing the nonzero true element
7490 for vector comparisons. The returned rtx should be valid for the inner
7491 mode of @var{mode} which is guaranteed to be a vector mode. Define
7492 this macro on machines that have vector comparison operations that
7493 return a vector result. If there are no such operations, do not define
7494 this macro. Typically, this macro is defined as @code{const1_rtx} or
7495 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
7496 the compiler optimizing such vector comparison operations for the
7500 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7501 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7502 A C expression that indicates whether the architecture defines a value
7503 for @code{clz} or @code{ctz} with a zero operand.
7504 A result of @code{0} indicates the value is undefined.
7505 If the value is defined for only the RTL expression, the macro should
7506 evaluate to @code{1}; if the value applies also to the corresponding optab
7507 entry (which is normally the case if it expands directly into
7508 the corresponding RTL), then the macro should evaluate to @code{2}.
7509 In the cases where the value is defined, @var{value} should be set to
7512 If this macro is not defined, the value of @code{clz} or
7513 @code{ctz} at zero is assumed to be undefined.
7515 This macro must be defined if the target's expansion for @code{ffs}
7516 relies on a particular value to get correct results. Otherwise it
7517 is not necessary, though it may be used to optimize some corner cases, and
7518 to provide a default expansion for the @code{ffs} optab.
7520 Note that regardless of this macro the ``definedness'' of @code{clz}
7521 and @code{ctz} at zero do @emph{not} extend to the builtin functions
7522 visible to the user. Thus one may be free to adjust the value at will
7523 to match the target expansion of these operations without fear of
7528 An alias for the machine mode for pointers. On most machines, define
7529 this to be the integer mode corresponding to the width of a hardware
7530 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7531 On some machines you must define this to be one of the partial integer
7532 modes, such as @code{PSImode}.
7534 The width of @code{Pmode} must be at least as large as the value of
7535 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7536 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7540 @defmac FUNCTION_MODE
7541 An alias for the machine mode used for memory references to functions
7542 being called, in @code{call} RTL expressions. On most CISC machines,
7543 where an instruction can begin at any byte address, this should be
7544 @code{QImode}. On most RISC machines, where all instructions have fixed
7545 size and alignment, this should be a mode with the same size and alignment
7546 as the machine instruction words - typically @code{SImode} or @code{HImode}.
7549 @defmac STDC_0_IN_SYSTEM_HEADERS
7550 In normal operation, the preprocessor expands @code{__STDC__} to the
7551 constant 1, to signify that GCC conforms to ISO Standard C@. On some
7552 hosts, like Solaris, the system compiler uses a different convention,
7553 where @code{__STDC__} is normally 0, but is 1 if the user specifies
7554 strict conformance to the C Standard.
7556 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
7557 convention when processing system header files, but when processing user
7558 files @code{__STDC__} will always expand to 1.
7561 @hook TARGET_C_PREINCLUDE
7563 @hook TARGET_CXX_IMPLICIT_EXTERN_C
7565 @defmac SYSTEM_IMPLICIT_EXTERN_C
7566 Define this macro if the system header files do not support C++@.
7567 This macro handles system header files by pretending that system
7568 header files are enclosed in @samp{extern "C" @{@dots{}@}}.
7573 @defmac REGISTER_TARGET_PRAGMAS ()
7574 Define this macro if you want to implement any target-specific pragmas.
7575 If defined, it is a C expression which makes a series of calls to
7576 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
7577 for each pragma. The macro may also do any
7578 setup required for the pragmas.
7580 The primary reason to define this macro is to provide compatibility with
7581 other compilers for the same target. In general, we discourage
7582 definition of target-specific pragmas for GCC@.
7584 If the pragma can be implemented by attributes then you should consider
7585 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
7587 Preprocessor macros that appear on pragma lines are not expanded. All
7588 @samp{#pragma} directives that do not match any registered pragma are
7589 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
7592 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7593 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7595 Each call to @code{c_register_pragma} or
7596 @code{c_register_pragma_with_expansion} establishes one pragma. The
7597 @var{callback} routine will be called when the preprocessor encounters a
7601 #pragma [@var{space}] @var{name} @dots{}
7604 @var{space} is the case-sensitive namespace of the pragma, or
7605 @code{NULL} to put the pragma in the global namespace. The callback
7606 routine receives @var{pfile} as its first argument, which can be passed
7607 on to cpplib's functions if necessary. You can lex tokens after the
7608 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
7609 callback will be silently ignored. The end of the line is indicated by
7610 a token of type @code{CPP_EOF}. Macro expansion occurs on the
7611 arguments of pragmas registered with
7612 @code{c_register_pragma_with_expansion} but not on the arguments of
7613 pragmas registered with @code{c_register_pragma}.
7615 Note that the use of @code{pragma_lex} is specific to the C and C++
7616 compilers. It will not work in the Java or Fortran compilers, or any
7617 other language compilers for that matter. Thus if @code{pragma_lex} is going
7618 to be called from target-specific code, it must only be done so when
7619 building the C and C++ compilers. This can be done by defining the
7620 variables @code{c_target_objs} and @code{cxx_target_objs} in the
7621 target entry in the @file{config.gcc} file. These variables should name
7622 the target-specific, language-specific object file which contains the
7623 code that uses @code{pragma_lex}. Note it will also be necessary to add a
7624 rule to the makefile fragment pointed to by @code{tmake_file} that shows
7625 how to build this object file.
7628 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
7629 Define this macro if macros should be expanded in the
7630 arguments of @samp{#pragma pack}.
7633 @defmac TARGET_DEFAULT_PACK_STRUCT
7634 If your target requires a structure packing default other than 0 (meaning
7635 the machine default), define this macro to the necessary value (in bytes).
7636 This must be a value that would also be valid to use with
7637 @samp{#pragma pack()} (that is, a small power of two).
7640 @defmac DOLLARS_IN_IDENTIFIERS
7641 Define this macro to control use of the character @samp{$} in
7642 identifier names for the C family of languages. 0 means @samp{$} is
7643 not allowed by default; 1 means it is allowed. 1 is the default;
7644 there is no need to define this macro in that case.
7647 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
7648 Define this macro as a C expression that is nonzero if it is safe for the
7649 delay slot scheduler to place instructions in the delay slot of @var{insn},
7650 even if they appear to use a resource set or clobbered in @var{insn}.
7651 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
7652 every @code{call_insn} has this behavior. On machines where some @code{insn}
7653 or @code{jump_insn} is really a function call and hence has this behavior,
7654 you should define this macro.
7656 You need not define this macro if it would always return zero.
7659 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
7660 Define this macro as a C expression that is nonzero if it is safe for the
7661 delay slot scheduler to place instructions in the delay slot of @var{insn},
7662 even if they appear to set or clobber a resource referenced in @var{insn}.
7663 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
7664 some @code{insn} or @code{jump_insn} is really a function call and its operands
7665 are registers whose use is actually in the subroutine it calls, you should
7666 define this macro. Doing so allows the delay slot scheduler to move
7667 instructions which copy arguments into the argument registers into the delay
7670 You need not define this macro if it would always return zero.
7673 @defmac MULTIPLE_SYMBOL_SPACES
7674 Define this macro as a C expression that is nonzero if, in some cases,
7675 global symbols from one translation unit may not be bound to undefined
7676 symbols in another translation unit without user intervention. For
7677 instance, under Microsoft Windows symbols must be explicitly imported
7678 from shared libraries (DLLs).
7680 You need not define this macro if it would always evaluate to zero.
7683 @hook TARGET_MD_ASM_ADJUST
7685 @defmac MATH_LIBRARY
7686 Define this macro as a C string constant for the linker argument to link
7687 in the system math library, minus the initial @samp{"-l"}, or
7688 @samp{""} if the target does not have a
7689 separate math library.
7691 You need only define this macro if the default of @samp{"m"} is wrong.
7694 @defmac LIBRARY_PATH_ENV
7695 Define this macro as a C string constant for the environment variable that
7696 specifies where the linker should look for libraries.
7698 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
7702 @defmac TARGET_POSIX_IO
7703 Define this macro if the target supports the following POSIX@ file
7704 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
7705 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
7706 to use file locking when exiting a program, which avoids race conditions
7707 if the program has forked. It will also create directories at run-time
7708 for cross-profiling.
7711 @defmac MAX_CONDITIONAL_EXECUTE
7713 A C expression for the maximum number of instructions to execute via
7714 conditional execution instructions instead of a branch. A value of
7715 @code{BRANCH_COST}+1 is the default.
7718 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
7719 Used if the target needs to perform machine-dependent modifications on the
7720 conditionals used for turning basic blocks into conditionally executed code.
7721 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
7722 contains information about the currently processed blocks. @var{true_expr}
7723 and @var{false_expr} are the tests that are used for converting the
7724 then-block and the else-block, respectively. Set either @var{true_expr} or
7725 @var{false_expr} to a null pointer if the tests cannot be converted.
7728 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
7729 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
7730 if-statements into conditions combined by @code{and} and @code{or} operations.
7731 @var{bb} contains the basic block that contains the test that is currently
7732 being processed and about to be turned into a condition.
7735 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
7736 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
7737 be converted to conditional execution format. @var{ce_info} points to
7738 a data structure, @code{struct ce_if_block}, which contains information
7739 about the currently processed blocks.
7742 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
7743 A C expression to perform any final machine dependent modifications in
7744 converting code to conditional execution. The involved basic blocks
7745 can be found in the @code{struct ce_if_block} structure that is pointed
7746 to by @var{ce_info}.
7749 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
7750 A C expression to cancel any machine dependent modifications in
7751 converting code to conditional execution. The involved basic blocks
7752 can be found in the @code{struct ce_if_block} structure that is pointed
7753 to by @var{ce_info}.
7756 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
7757 A C expression to initialize any machine specific data for if-conversion
7758 of the if-block in the @code{struct ce_if_block} structure that is pointed
7759 to by @var{ce_info}.
7762 @hook TARGET_MACHINE_DEPENDENT_REORG
7764 @hook TARGET_INIT_BUILTINS
7766 @hook TARGET_BUILTIN_DECL
7768 @hook TARGET_EXPAND_BUILTIN
7770 @hook TARGET_RESOLVE_OVERLOADED_BUILTIN
7772 @hook TARGET_CHECK_BUILTIN_CALL
7774 @hook TARGET_FOLD_BUILTIN
7776 @hook TARGET_GIMPLE_FOLD_BUILTIN
7778 @hook TARGET_COMPARE_VERSION_PRIORITY
7780 @hook TARGET_GET_FUNCTION_VERSIONS_DISPATCHER
7782 @hook TARGET_GENERATE_VERSION_DISPATCHER_BODY
7784 @hook TARGET_PREDICT_DOLOOP_P
7786 @hook TARGET_HAVE_COUNT_REG_DECR_P
7788 @hook TARGET_DOLOOP_COST_FOR_GENERIC
7790 @hook TARGET_DOLOOP_COST_FOR_ADDRESS
7792 @hook TARGET_CAN_USE_DOLOOP_P
7794 @hook TARGET_INVALID_WITHIN_DOLOOP
7796 @hook TARGET_PREFERRED_DOLOOP_MODE
7798 @hook TARGET_LEGITIMATE_COMBINED_INSN
7800 @hook TARGET_CAN_FOLLOW_JUMP
7802 @hook TARGET_COMMUTATIVE_P
7804 @hook TARGET_ALLOCATE_INITIAL_VALUE
7806 @hook TARGET_UNSPEC_MAY_TRAP_P
7808 @hook TARGET_SET_CURRENT_FUNCTION
7810 @defmac TARGET_OBJECT_SUFFIX
7811 Define this macro to be a C string representing the suffix for object
7812 files on your target machine. If you do not define this macro, GCC will
7813 use @samp{.o} as the suffix for object files.
7816 @defmac TARGET_EXECUTABLE_SUFFIX
7817 Define this macro to be a C string representing the suffix to be
7818 automatically added to executable files on your target machine. If you
7819 do not define this macro, GCC will use the null string as the suffix for
7823 @defmac COLLECT_EXPORT_LIST
7824 If defined, @code{collect2} will scan the individual object files
7825 specified on its command line and create an export list for the linker.
7826 Define this macro for systems like AIX, where the linker discards
7827 object files that are not referenced from @code{main} and uses export
7831 @hook TARGET_CANNOT_MODIFY_JUMPS_P
7833 @hook TARGET_HAVE_CONDITIONAL_EXECUTION
7835 @hook TARGET_GEN_CCMP_FIRST
7837 @hook TARGET_GEN_CCMP_NEXT
7839 @hook TARGET_LOOP_UNROLL_ADJUST
7841 @defmac POWI_MAX_MULTS
7842 If defined, this macro is interpreted as a signed integer C expression
7843 that specifies the maximum number of floating point multiplications
7844 that should be emitted when expanding exponentiation by an integer
7845 constant inline. When this value is defined, exponentiation requiring
7846 more than this number of multiplications is implemented by calling the
7847 system library's @code{pow}, @code{powf} or @code{powl} routines.
7848 The default value places no upper bound on the multiplication count.
7851 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
7852 This target hook should register any extra include files for the
7853 target. The parameter @var{stdinc} indicates if normal include files
7854 are present. The parameter @var{sysroot} is the system root directory.
7855 The parameter @var{iprefix} is the prefix for the gcc directory.
7858 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
7859 This target hook should register any extra include files for the
7860 target before any standard headers. The parameter @var{stdinc}
7861 indicates if normal include files are present. The parameter
7862 @var{sysroot} is the system root directory. The parameter
7863 @var{iprefix} is the prefix for the gcc directory.
7866 @deftypefn Macro void TARGET_OPTF (char *@var{path})
7867 This target hook should register special include paths for the target.
7868 The parameter @var{path} is the include to register. On Darwin
7869 systems, this is used for Framework includes, which have semantics
7870 that are different from @option{-I}.
7873 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
7874 This target macro returns @code{true} if it is safe to use a local alias
7875 for a virtual function @var{fndecl} when constructing thunks,
7876 @code{false} otherwise. By default, the macro returns @code{true} for all
7877 functions, if a target supports aliases (i.e.@: defines
7878 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
7881 @defmac TARGET_FORMAT_TYPES
7882 If defined, this macro is the name of a global variable containing
7883 target-specific format checking information for the @option{-Wformat}
7884 option. The default is to have no target-specific format checks.
7887 @defmac TARGET_N_FORMAT_TYPES
7888 If defined, this macro is the number of entries in
7889 @code{TARGET_FORMAT_TYPES}.
7892 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
7893 If defined, this macro is the name of a global variable containing
7894 target-specific format overrides for the @option{-Wformat} option. The
7895 default is to have no target-specific format overrides. If defined,
7896 @code{TARGET_FORMAT_TYPES} and @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT}
7897 must be defined, too.
7900 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
7901 If defined, this macro specifies the number of entries in
7902 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
7905 @defmac TARGET_OVERRIDES_FORMAT_INIT
7906 If defined, this macro specifies the optional initialization
7907 routine for target specific customizations of the system printf
7908 and scanf formatter settings.
7911 @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
7913 @hook TARGET_INVALID_CONVERSION
7915 @hook TARGET_INVALID_UNARY_OP
7917 @hook TARGET_INVALID_BINARY_OP
7919 @hook TARGET_PROMOTED_TYPE
7921 @hook TARGET_CONVERT_TO_TYPE
7923 @hook TARGET_VERIFY_TYPE_CONTEXT
7926 This macro determines the size of the objective C jump buffer for the
7927 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
7930 @defmac LIBGCC2_UNWIND_ATTRIBUTE
7931 Define this macro if any target-specific attributes need to be attached
7932 to the functions in @file{libgcc} that provide low-level support for
7933 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
7934 and the associated definitions of those functions.
7937 @hook TARGET_UPDATE_STACK_BOUNDARY
7939 @hook TARGET_GET_DRAP_RTX
7941 @hook TARGET_ZERO_CALL_USED_REGS
7943 @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
7945 @hook TARGET_CONST_ANCHOR
7947 @hook TARGET_ASAN_SHADOW_OFFSET
7949 @hook TARGET_MEMMODEL_CHECK
7951 @hook TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
7953 @hook TARGET_HAS_IFUNC_P
7955 @hook TARGET_IFUNC_REF_LOCAL_OK
7957 @hook TARGET_ATOMIC_ALIGN_FOR_MODE
7959 @hook TARGET_ATOMIC_ASSIGN_EXPAND_FENV
7961 @hook TARGET_RECORD_OFFLOAD_SYMBOL
7963 @hook TARGET_OFFLOAD_OPTIONS
7965 @defmac TARGET_SUPPORTS_WIDE_INT
7967 On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
7968 objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
7969 to indicate that large integers are stored in
7970 @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows
7971 very large integer constants to be represented. @code{CONST_DOUBLE}
7972 is limited to twice the size of the host's @code{HOST_WIDE_INT}
7975 Converting a port mostly requires looking for the places where
7976 @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
7977 code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i
7978 const_double"} at the port level gets you to 95% of the changes that
7979 need to be made. There are a few places that require a deeper look.
7983 There is no equivalent to @code{hval} and @code{lval} for
7984 @code{CONST_WIDE_INT}s. This would be difficult to express in the md
7985 language since there are a variable number of elements.
7987 Most ports only check that @code{hval} is either 0 or -1 to see if the
7988 value is small. As mentioned above, this will no longer be necessary
7989 since small constants are always @code{CONST_INT}. Of course there
7990 are still a few exceptions, the alpha's constraint used by the zap
7991 instruction certainly requires careful examination by C code.
7992 However, all the current code does is pass the hval and lval to C
7993 code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
7994 not really a large change.
7997 Because there is no standard template that ports use to materialize
7998 constants, there is likely to be some futzing that is unique to each
8002 The rtx costs may have to be adjusted to properly account for larger
8003 constants that are represented as @code{CONST_WIDE_INT}.
8006 All and all it does not take long to convert ports that the
8007 maintainer is familiar with.
8011 @hook TARGET_HAVE_SPECULATION_SAFE_VALUE
8013 @hook TARGET_SPECULATION_SAFE_VALUE
8015 @hook TARGET_RUN_TARGET_SELFTESTS
8017 @hook TARGET_MEMTAG_CAN_TAG_ADDRESSES
8019 @hook TARGET_MEMTAG_TAG_SIZE
8021 @hook TARGET_MEMTAG_GRANULE_SIZE
8023 @hook TARGET_MEMTAG_INSERT_RANDOM_TAG
8025 @hook TARGET_MEMTAG_ADD_TAG
8027 @hook TARGET_MEMTAG_SET_TAG
8029 @hook TARGET_MEMTAG_EXTRACT_TAG
8031 @hook TARGET_MEMTAG_UNTAGGED_POINTER
8033 @hook TARGET_GCOV_TYPE_SIZE
8035 @hook TARGET_HAVE_SHADOW_CALL_STACK