1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Target Description Macros and Functions
9 @cindex machine description macros
10 @cindex target description macros
11 @cindex macros, target description
12 @cindex @file{tm.h} macros
14 In addition to the file @file{@var{machine}.md}, a machine description
15 includes a C header file conventionally given the name
16 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
17 The header file defines numerous macros that convey the information
18 about the target machine that does not fit into the scheme of the
19 @file{.md} file. The file @file{tm.h} should be a link to
20 @file{@var{machine}.h}. The header file @file{config.h} includes
21 @file{tm.h} and most compiler source files include @file{config.h}. The
22 source file defines a variable @code{targetm}, which is a structure
23 containing pointers to functions and data relating to the target
24 machine. @file{@var{machine}.c} should also contain their definitions,
25 if they are not defined elsewhere in GCC, and other functions called
26 through the macros defined in the @file{.h} file.
29 * Target Structure:: The @code{targetm} variable.
30 * Driver:: Controlling how the driver runs the compilation passes.
31 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
32 * Per-Function Data:: Defining data structures for per-function information.
33 * Storage Layout:: Defining sizes and alignments of data.
34 * Type Layout:: Defining sizes and properties of basic user data types.
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
37 * Old Constraints:: The old way to define machine-specific constraints.
38 * Stack and Calling:: Defining which way the stack grows and by how much.
39 * Varargs:: Defining the varargs macros.
40 * Trampolines:: Code set up at run time to enter a nested function.
41 * Library Calls:: Controlling how library routines are implicitly called.
42 * Addressing Modes:: Defining addressing modes valid for memory operands.
43 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
44 * Condition Code:: Defining how insns update the condition code.
45 * Costs:: Defining relative costs of different operations.
46 * Scheduling:: Adjusting the behavior of the instruction scheduler.
47 * Sections:: Dividing storage into text, data, and other sections.
48 * PIC:: Macros for position independent code.
49 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
50 * Debugging Info:: Defining the format of debugging output.
51 * Floating Point:: Handling floating point for cross-compilers.
52 * Mode Switching:: Insertion of mode-switching instructions.
53 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
54 * Emulated TLS:: Emulated TLS support.
55 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
56 * PCH Target:: Validity checking for precompiled headers.
57 * C++ ABI:: Controlling C++ ABI changes.
58 * Named Address Spaces:: Adding support for named address spaces
59 * Misc:: Everything else.
62 @node Target Structure
63 @section The Global @code{targetm} Variable
65 @cindex target functions
67 @deftypevar {struct gcc_target} targetm
68 The target @file{.c} file must define the global @code{targetm} variable
69 which contains pointers to functions and data relating to the target
70 machine. The variable is declared in @file{target.h};
71 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
72 used to initialize the variable, and macros for the default initializers
73 for elements of the structure. The @file{.c} file should override those
74 macros for which the default definition is inappropriate. For example:
77 #include "target-def.h"
79 /* @r{Initialize the GCC target structure.} */
81 #undef TARGET_COMP_TYPE_ATTRIBUTES
82 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
84 struct gcc_target targetm = TARGET_INITIALIZER;
88 Where a macro should be defined in the @file{.c} file in this manner to
89 form part of the @code{targetm} structure, it is documented below as a
90 ``Target Hook'' with a prototype. Many macros will change in future
91 from being defined in the @file{.h} file to being part of the
92 @code{targetm} structure.
94 Similarly, there is a @code{targetcm} variable for hooks that are
95 specific to front ends for C-family languages, documented as ``C
96 Target Hook''. This is declared in @file{c-family/c-target.h}, the
97 initializer @code{TARGETCM_INITIALIZER} in
98 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
99 themselves, they should set @code{target_has_targetcm=yes} in
100 @file{config.gcc}; otherwise a default definition is used.
102 Similarly, there is a @code{targetm_common} variable for hooks that
103 are shared between the compiler driver and the compilers proper,
104 documented as ``Common Target Hook''. This is declared in
105 @file{common/common-target.h}, the initializer
106 @code{TARGETM_COMMON_INITIALIZER} in
107 @file{common/common-target-def.h}. If targets initialize
108 @code{targetm_common} themselves, they should set
109 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
110 default definition is used.
113 @section Controlling the Compilation Driver, @file{gcc}
115 @cindex controlling the compilation driver
117 @c prevent bad page break with this line
118 You can control the compilation driver.
120 @defmac DRIVER_SELF_SPECS
121 A list of specs for the driver itself. It should be a suitable
122 initializer for an array of strings, with no surrounding braces.
124 The driver applies these specs to its own command line between loading
125 default @file{specs} files (but not command-line specified ones) and
126 choosing the multilib directory or running any subcommands. It
127 applies them in the order given, so each spec can depend on the
128 options added by earlier ones. It is also possible to remove options
129 using @samp{%<@var{option}} in the usual way.
131 This macro can be useful when a port has several interdependent target
132 options. It provides a way of standardizing the command line so
133 that the other specs are easier to write.
135 Do not define this macro if it does not need to do anything.
138 @defmac OPTION_DEFAULT_SPECS
139 A list of specs used to support configure-time default options (i.e.@:
140 @option{--with} options) in the driver. It should be a suitable initializer
141 for an array of structures, each containing two strings, without the
142 outermost pair of surrounding braces.
144 The first item in the pair is the name of the default. This must match
145 the code in @file{config.gcc} for the target. The second item is a spec
146 to apply if a default with this name was specified. The string
147 @samp{%(VALUE)} in the spec will be replaced by the value of the default
148 everywhere it occurs.
150 The driver will apply these specs to its own command line between loading
151 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
152 the same mechanism as @code{DRIVER_SELF_SPECS}.
154 Do not define this macro if it does not need to do anything.
158 A C string constant that tells the GCC driver program options to
159 pass to CPP@. It can also specify how to translate options you
160 give to GCC into options for GCC to pass to the CPP@.
162 Do not define this macro if it does not need to do anything.
165 @defmac CPLUSPLUS_CPP_SPEC
166 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
167 than C@. If you do not define this macro, then the value of
168 @code{CPP_SPEC} (if any) will be used instead.
172 A C string constant that tells the GCC driver program options to
173 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
175 It can also specify how to translate options you give to GCC into options
176 for GCC to pass to front ends.
178 Do not define this macro if it does not need to do anything.
182 A C string constant that tells the GCC driver program options to
183 pass to @code{cc1plus}. It can also specify how to translate options you
184 give to GCC into options for GCC to pass to the @code{cc1plus}.
186 Do not define this macro if it does not need to do anything.
187 Note that everything defined in CC1_SPEC is already passed to
188 @code{cc1plus} so there is no need to duplicate the contents of
189 CC1_SPEC in CC1PLUS_SPEC@.
193 A C string constant that tells the GCC driver program options to
194 pass to the assembler. It can also specify how to translate options
195 you give to GCC into options for GCC to pass to the assembler.
196 See the file @file{sun3.h} for an example of this.
198 Do not define this macro if it does not need to do anything.
201 @defmac ASM_FINAL_SPEC
202 A C string constant that tells the GCC driver program how to
203 run any programs which cleanup after the normal assembler.
204 Normally, this is not needed. See the file @file{mips.h} for
207 Do not define this macro if it does not need to do anything.
210 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
211 Define this macro, with no value, if the driver should give the assembler
212 an argument consisting of a single dash, @option{-}, to instruct it to
213 read from its standard input (which will be a pipe connected to the
214 output of the compiler proper). This argument is given after any
215 @option{-o} option specifying the name of the output file.
217 If you do not define this macro, the assembler is assumed to read its
218 standard input if given no non-option arguments. If your assembler
219 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
220 see @file{mips.h} for instance.
224 A C string constant that tells the GCC driver program options to
225 pass to the linker. It can also specify how to translate options you
226 give to GCC into options for GCC to pass to the linker.
228 Do not define this macro if it does not need to do anything.
232 Another C string constant used much like @code{LINK_SPEC}. The difference
233 between the two is that @code{LIB_SPEC} is used at the end of the
234 command given to the linker.
236 If this macro is not defined, a default is provided that
237 loads the standard C library from the usual place. See @file{gcc.c}.
241 Another C string constant that tells the GCC driver program
242 how and when to place a reference to @file{libgcc.a} into the
243 linker command line. This constant is placed both before and after
244 the value of @code{LIB_SPEC}.
246 If this macro is not defined, the GCC driver provides a default that
247 passes the string @option{-lgcc} to the linker.
250 @defmac REAL_LIBGCC_SPEC
251 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
252 @code{LIBGCC_SPEC} is not directly used by the driver program but is
253 instead modified to refer to different versions of @file{libgcc.a}
254 depending on the values of the command line flags @option{-static},
255 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
256 targets where these modifications are inappropriate, define
257 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
258 driver how to place a reference to @file{libgcc} on the link command
259 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
262 @defmac USE_LD_AS_NEEDED
263 A macro that controls the modifications to @code{LIBGCC_SPEC}
264 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
265 generated that uses --as-needed and the shared libgcc in place of the
266 static exception handler library, when linking without any of
267 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
271 If defined, this C string constant is added to @code{LINK_SPEC}.
272 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
273 the modifications to @code{LIBGCC_SPEC} mentioned in
274 @code{REAL_LIBGCC_SPEC}.
277 @defmac STARTFILE_SPEC
278 Another C string constant used much like @code{LINK_SPEC}. The
279 difference between the two is that @code{STARTFILE_SPEC} is used at
280 the very beginning of the command given to the linker.
282 If this macro is not defined, a default is provided that loads the
283 standard C startup file from the usual place. See @file{gcc.c}.
287 Another C string constant used much like @code{LINK_SPEC}. The
288 difference between the two is that @code{ENDFILE_SPEC} is used at
289 the very end of the command given to the linker.
291 Do not define this macro if it does not need to do anything.
294 @defmac THREAD_MODEL_SPEC
295 GCC @code{-v} will print the thread model GCC was configured to use.
296 However, this doesn't work on platforms that are multilibbed on thread
297 models, such as AIX 4.3. On such platforms, define
298 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
299 blanks that names one of the recognized thread models. @code{%*}, the
300 default value of this macro, will expand to the value of
301 @code{thread_file} set in @file{config.gcc}.
304 @defmac SYSROOT_SUFFIX_SPEC
305 Define this macro to add a suffix to the target sysroot when GCC is
306 configured with a sysroot. This will cause GCC to search for usr/lib,
307 et al, within sysroot+suffix.
310 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
311 Define this macro to add a headers_suffix to the target sysroot when
312 GCC is configured with a sysroot. This will cause GCC to pass the
313 updated sysroot+headers_suffix to CPP, causing it to search for
314 usr/include, et al, within sysroot+headers_suffix.
318 Define this macro to provide additional specifications to put in the
319 @file{specs} file that can be used in various specifications like
322 The definition should be an initializer for an array of structures,
323 containing a string constant, that defines the specification name, and a
324 string constant that provides the specification.
326 Do not define this macro if it does not need to do anything.
328 @code{EXTRA_SPECS} is useful when an architecture contains several
329 related targets, which have various @code{@dots{}_SPECS} which are similar
330 to each other, and the maintainer would like one central place to keep
333 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
334 define either @code{_CALL_SYSV} when the System V calling sequence is
335 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
338 The @file{config/rs6000/rs6000.h} target file defines:
341 #define EXTRA_SPECS \
342 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
344 #define CPP_SYS_DEFAULT ""
347 The @file{config/rs6000/sysv.h} target file defines:
351 "%@{posix: -D_POSIX_SOURCE @} \
352 %@{mcall-sysv: -D_CALL_SYSV @} \
353 %@{!mcall-sysv: %(cpp_sysv_default) @} \
354 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
356 #undef CPP_SYSV_DEFAULT
357 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
360 while the @file{config/rs6000/eabiaix.h} target file defines
361 @code{CPP_SYSV_DEFAULT} as:
364 #undef CPP_SYSV_DEFAULT
365 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
369 @defmac LINK_LIBGCC_SPECIAL_1
370 Define this macro if the driver program should find the library
371 @file{libgcc.a}. If you do not define this macro, the driver program will pass
372 the argument @option{-lgcc} to tell the linker to do the search.
375 @defmac LINK_GCC_C_SEQUENCE_SPEC
376 The sequence in which libgcc and libc are specified to the linker.
377 By default this is @code{%G %L %G}.
380 @defmac LINK_COMMAND_SPEC
381 A C string constant giving the complete command line need to execute the
382 linker. When you do this, you will need to update your port each time a
383 change is made to the link command line within @file{gcc.c}. Therefore,
384 define this macro only if you need to completely redefine the command
385 line for invoking the linker and there is no other way to accomplish
386 the effect you need. Overriding this macro may be avoidable by overriding
387 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
390 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
391 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
392 directories from linking commands. Do not give it a nonzero value if
393 removing duplicate search directories changes the linker's semantics.
396 @deftypevr {Common Target Hook} bool TARGET_ALWAYS_STRIP_DOTDOT
397 True if @file{..} components should always be removed from directory names computed relative to GCC's internal directories, false (default) if such components should be preserved and directory names containing them passed to other tools such as the linker.
400 @defmac MULTILIB_DEFAULTS
401 Define this macro as a C expression for the initializer of an array of
402 string to tell the driver program which options are defaults for this
403 target and thus do not need to be handled specially when using
404 @code{MULTILIB_OPTIONS}.
406 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
407 the target makefile fragment or if none of the options listed in
408 @code{MULTILIB_OPTIONS} are set by default.
409 @xref{Target Fragment}.
412 @defmac RELATIVE_PREFIX_NOT_LINKDIR
413 Define this macro to tell @command{gcc} that it should only translate
414 a @option{-B} prefix into a @option{-L} linker option if the prefix
415 indicates an absolute file name.
418 @defmac MD_EXEC_PREFIX
419 If defined, this macro is an additional prefix to try after
420 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
421 when the compiler is built as a cross
422 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
423 to the list of directories used to find the assembler in @file{configure.in}.
426 @defmac STANDARD_STARTFILE_PREFIX
427 Define this macro as a C string constant if you wish to override the
428 standard choice of @code{libdir} as the default prefix to
429 try when searching for startup files such as @file{crt0.o}.
430 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
431 is built as a cross compiler.
434 @defmac STANDARD_STARTFILE_PREFIX_1
435 Define this macro as a C string constant if you wish to override the
436 standard choice of @code{/lib} as a prefix to try after the default prefix
437 when searching for startup files such as @file{crt0.o}.
438 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
439 is built as a cross compiler.
442 @defmac STANDARD_STARTFILE_PREFIX_2
443 Define this macro as a C string constant if you wish to override the
444 standard choice of @code{/lib} as yet another prefix to try after the
445 default prefix when searching for startup files such as @file{crt0.o}.
446 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
447 is built as a cross compiler.
450 @defmac MD_STARTFILE_PREFIX
451 If defined, this macro supplies an additional prefix to try after the
452 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
453 compiler is built as a cross compiler.
456 @defmac MD_STARTFILE_PREFIX_1
457 If defined, this macro supplies yet another prefix to try after the
458 standard prefixes. It is not searched when the compiler is built as a
462 @defmac INIT_ENVIRONMENT
463 Define this macro as a C string constant if you wish to set environment
464 variables for programs called by the driver, such as the assembler and
465 loader. The driver passes the value of this macro to @code{putenv} to
466 initialize the necessary environment variables.
469 @defmac LOCAL_INCLUDE_DIR
470 Define this macro as a C string constant if you wish to override the
471 standard choice of @file{/usr/local/include} as the default prefix to
472 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
473 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
475 Cross compilers do not search either @file{/usr/local/include} or its
479 @defmac SYSTEM_INCLUDE_DIR
480 Define this macro as a C string constant if you wish to specify a
481 system-specific directory to search for header files before the standard
482 directory. @code{SYSTEM_INCLUDE_DIR} comes before
483 @code{STANDARD_INCLUDE_DIR} in the search order.
485 Cross compilers do not use this macro and do not search the directory
489 @defmac STANDARD_INCLUDE_DIR
490 Define this macro as a C string constant if you wish to override the
491 standard choice of @file{/usr/include} as the default prefix to
492 try when searching for header files.
494 Cross compilers ignore this macro and do not search either
495 @file{/usr/include} or its replacement.
498 @defmac STANDARD_INCLUDE_COMPONENT
499 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
500 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
501 If you do not define this macro, no component is used.
504 @defmac INCLUDE_DEFAULTS
505 Define this macro if you wish to override the entire default search path
506 for include files. For a native compiler, the default search path
507 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
508 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
509 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
510 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
511 and specify private search areas for GCC@. The directory
512 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
514 The definition should be an initializer for an array of structures.
515 Each array element should have four elements: the directory name (a
516 string constant), the component name (also a string constant), a flag
517 for C++-only directories,
518 and a flag showing that the includes in the directory don't need to be
519 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
520 the array with a null element.
522 The component name denotes what GNU package the include file is part of,
523 if any, in all uppercase letters. For example, it might be @samp{GCC}
524 or @samp{BINUTILS}. If the package is part of a vendor-supplied
525 operating system, code the component name as @samp{0}.
527 For example, here is the definition used for VAX/VMS:
530 #define INCLUDE_DEFAULTS \
532 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
533 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
534 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
541 Here is the order of prefixes tried for exec files:
545 Any prefixes specified by the user with @option{-B}.
548 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
549 is not set and the compiler has not been installed in the configure-time
550 @var{prefix}, the location in which the compiler has actually been installed.
553 The directories specified by the environment variable @code{COMPILER_PATH}.
556 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
557 in the configured-time @var{prefix}.
560 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
563 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
566 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
570 Here is the order of prefixes tried for startfiles:
574 Any prefixes specified by the user with @option{-B}.
577 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
578 value based on the installed toolchain location.
581 The directories specified by the environment variable @code{LIBRARY_PATH}
582 (or port-specific name; native only, cross compilers do not use this).
585 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
586 in the configured @var{prefix} or this is a native compiler.
589 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
592 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
596 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
597 native compiler, or we have a target system root.
600 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
601 native compiler, or we have a target system root.
604 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
605 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
606 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
609 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
610 compiler, or we have a target system root. The default for this macro is
614 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
615 compiler, or we have a target system root. The default for this macro is
619 @node Run-time Target
620 @section Run-time Target Specification
621 @cindex run-time target specification
622 @cindex predefined macros
623 @cindex target specifications
625 @c prevent bad page break with this line
626 Here are run-time target specifications.
628 @defmac TARGET_CPU_CPP_BUILTINS ()
629 This function-like macro expands to a block of code that defines
630 built-in preprocessor macros and assertions for the target CPU, using
631 the functions @code{builtin_define}, @code{builtin_define_std} and
632 @code{builtin_assert}. When the front end
633 calls this macro it provides a trailing semicolon, and since it has
634 finished command line option processing your code can use those
637 @code{builtin_assert} takes a string in the form you pass to the
638 command-line option @option{-A}, such as @code{cpu=mips}, and creates
639 the assertion. @code{builtin_define} takes a string in the form
640 accepted by option @option{-D} and unconditionally defines the macro.
642 @code{builtin_define_std} takes a string representing the name of an
643 object-like macro. If it doesn't lie in the user's namespace,
644 @code{builtin_define_std} defines it unconditionally. Otherwise, it
645 defines a version with two leading underscores, and another version
646 with two leading and trailing underscores, and defines the original
647 only if an ISO standard was not requested on the command line. For
648 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
649 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
650 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
651 defines only @code{_ABI64}.
653 You can also test for the C dialect being compiled. The variable
654 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
655 or @code{clk_objective_c}. Note that if we are preprocessing
656 assembler, this variable will be @code{clk_c} but the function-like
657 macro @code{preprocessing_asm_p()} will return true, so you might want
658 to check for that first. If you need to check for strict ANSI, the
659 variable @code{flag_iso} can be used. The function-like macro
660 @code{preprocessing_trad_p()} can be used to check for traditional
664 @defmac TARGET_OS_CPP_BUILTINS ()
665 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
666 and is used for the target operating system instead.
669 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
670 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
671 and is used for the target object format. @file{elfos.h} uses this
672 macro to define @code{__ELF__}, so you probably do not need to define
676 @deftypevar {extern int} target_flags
677 This variable is declared in @file{options.h}, which is included before
678 any target-specific headers.
681 @deftypevr {Common Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
682 This variable specifies the initial value of @code{target_flags}.
683 Its default setting is 0.
686 @cindex optional hardware or system features
687 @cindex features, optional, in system conventions
689 @deftypefn {Common Target Hook} bool TARGET_HANDLE_OPTION (struct gcc_options *@var{opts}, struct gcc_options *@var{opts_set}, const struct cl_decoded_option *@var{decoded}, location_t @var{loc})
690 This hook is called whenever the user specifies one of the
691 target-specific options described by the @file{.opt} definition files
692 (@pxref{Options}). It has the opportunity to do some option-specific
693 processing and should return true if the option is valid. The default
694 definition does nothing but return true.
696 @var{decoded} specifies the option and its arguments. @var{opts} and
697 @var{opts_set} are the @code{gcc_options} structures to be used for
698 storing option state, and @var{loc} is the location at which the
699 option was passed (@code{UNKNOWN_LOCATION} except for options passed
703 @deftypefn {C Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
704 This target hook is called whenever the user specifies one of the
705 target-specific C language family options described by the @file{.opt}
706 definition files(@pxref{Options}). It has the opportunity to do some
707 option-specific processing and should return true if the option is
708 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
709 default definition does nothing but return false.
711 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
712 options. However, if processing an option requires routines that are
713 only available in the C (and related language) front ends, then you
714 should use @code{TARGET_HANDLE_C_OPTION} instead.
717 @deftypefn {C Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING_OBJECT (tree @var{string})
718 Targets may provide a string object type that can be used within and between C, C++ and their respective Objective-C dialects. A string object might, for example, embed encoding and length information. These objects are considered opaque to the compiler and handled as references. An ideal implementation makes the composition of the string object match that of the Objective-C @code{NSString} (@code{NXString} for GNUStep), allowing efficient interworking between C-only and Objective-C code. If a target implements string objects then this hook should return a reference to such an object constructed from the normal `C' string representation provided in @var{string}. At present, the hook is used by Objective-C only, to obtain a common-format string object when the target provides one.
721 @deftypefn {C Target Hook} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref})
722 If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object.
725 @deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list})
726 If a target implements string objects then this hook should should provide a facility to check the function arguments in @var{args_list} against the format specifiers in @var{format_arg} where the type of @var{format_arg} is one recognized as a valid string reference type.
729 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
730 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
731 but is called when the optimize level is changed via an attribute or
732 pragma or when it is reset at the end of the code affected by the
733 attribute or pragma. It is not called at the beginning of compilation
734 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
735 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
736 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
739 @defmac C_COMMON_OVERRIDE_OPTIONS
740 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
741 but is only used in the C
742 language frontends (C, Objective-C, C++, Objective-C++) and so can be
743 used to alter option flag variables which only exist in those
747 @deftypevr {Common Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
748 Some machines may desire to change what optimizations are performed for
749 various optimization levels. This variable, if defined, describes
750 options to enable at particular sets of optimization levels. These
751 options are processed once
752 just after the optimization level is determined and before the remainder
753 of the command options have been parsed, so may be overridden by other
754 options passed explicitly.
756 This processing is run once at program startup and when the optimization
757 options are changed via @code{#pragma GCC optimize} or by using the
758 @code{optimize} attribute.
761 @deftypefn {Common Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
762 Set target-dependent initial values of fields in @var{opts}.
765 @deftypefn {Common Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
766 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
769 @defmac SWITCHABLE_TARGET
770 Some targets need to switch between substantially different subtargets
771 during compilation. For example, the MIPS target has one subtarget for
772 the traditional MIPS architecture and another for MIPS16. Source code
773 can switch between these two subarchitectures using the @code{mips16}
774 and @code{nomips16} attributes.
776 Such subtargets can differ in things like the set of available
777 registers, the set of available instructions, the costs of various
778 operations, and so on. GCC caches a lot of this type of information
779 in global variables, and recomputing them for each subtarget takes a
780 significant amount of time. The compiler therefore provides a facility
781 for maintaining several versions of the global variables and quickly
782 switching between them; see @file{target-globals.h} for details.
784 Define this macro to 1 if your target needs this facility. The default
788 @node Per-Function Data
789 @section Defining data structures for per-function information.
790 @cindex per-function data
791 @cindex data structures
793 If the target needs to store information on a per-function basis, GCC
794 provides a macro and a couple of variables to allow this. Note, just
795 using statics to store the information is a bad idea, since GCC supports
796 nested functions, so you can be halfway through encoding one function
797 when another one comes along.
799 GCC defines a data structure called @code{struct function} which
800 contains all of the data specific to an individual function. This
801 structure contains a field called @code{machine} whose type is
802 @code{struct machine_function *}, which can be used by targets to point
803 to their own specific data.
805 If a target needs per-function specific data it should define the type
806 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
807 This macro should be used to initialize the function pointer
808 @code{init_machine_status}. This pointer is explained below.
810 One typical use of per-function, target specific data is to create an
811 RTX to hold the register containing the function's return address. This
812 RTX can then be used to implement the @code{__builtin_return_address}
813 function, for level 0.
815 Note---earlier implementations of GCC used a single data area to hold
816 all of the per-function information. Thus when processing of a nested
817 function began the old per-function data had to be pushed onto a
818 stack, and when the processing was finished, it had to be popped off the
819 stack. GCC used to provide function pointers called
820 @code{save_machine_status} and @code{restore_machine_status} to handle
821 the saving and restoring of the target specific information. Since the
822 single data area approach is no longer used, these pointers are no
825 @defmac INIT_EXPANDERS
826 Macro called to initialize any target specific information. This macro
827 is called once per function, before generation of any RTL has begun.
828 The intention of this macro is to allow the initialization of the
829 function pointer @code{init_machine_status}.
832 @deftypevar {void (*)(struct function *)} init_machine_status
833 If this function pointer is non-@code{NULL} it will be called once per
834 function, before function compilation starts, in order to allow the
835 target to perform any target specific initialization of the
836 @code{struct function} structure. It is intended that this would be
837 used to initialize the @code{machine} of that structure.
839 @code{struct machine_function} structures are expected to be freed by GC@.
840 Generally, any memory that they reference must be allocated by using
841 GC allocation, including the structure itself.
845 @section Storage Layout
846 @cindex storage layout
848 Note that the definitions of the macros in this table which are sizes or
849 alignments measured in bits do not need to be constant. They can be C
850 expressions that refer to static variables, such as the @code{target_flags}.
851 @xref{Run-time Target}.
853 @defmac BITS_BIG_ENDIAN
854 Define this macro to have the value 1 if the most significant bit in a
855 byte has the lowest number; otherwise define it to have the value zero.
856 This means that bit-field instructions count from the most significant
857 bit. If the machine has no bit-field instructions, then this must still
858 be defined, but it doesn't matter which value it is defined to. This
859 macro need not be a constant.
861 This macro does not affect the way structure fields are packed into
862 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
865 @defmac BYTES_BIG_ENDIAN
866 Define this macro to have the value 1 if the most significant byte in a
867 word has the lowest number. This macro need not be a constant.
870 @defmac WORDS_BIG_ENDIAN
871 Define this macro to have the value 1 if, in a multiword object, the
872 most significant word has the lowest number. This applies to both
873 memory locations and registers; GCC fundamentally assumes that the
874 order of words in memory is the same as the order in registers. This
875 macro need not be a constant.
878 @defmac FLOAT_WORDS_BIG_ENDIAN
879 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
880 @code{TFmode} floating point numbers are stored in memory with the word
881 containing the sign bit at the lowest address; otherwise define it to
882 have the value 0. This macro need not be a constant.
884 You need not define this macro if the ordering is the same as for
888 @defmac BITS_PER_UNIT
889 Define this macro to be the number of bits in an addressable storage
890 unit (byte). If you do not define this macro the default is 8.
893 @defmac BITS_PER_WORD
894 Number of bits in a word. If you do not define this macro, the default
895 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
898 @defmac MAX_BITS_PER_WORD
899 Maximum number of bits in a word. If this is undefined, the default is
900 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
901 largest value that @code{BITS_PER_WORD} can have at run-time.
904 @defmac UNITS_PER_WORD
905 Number of storage units in a word; normally the size of a general-purpose
906 register, a power of two from 1 or 8.
909 @defmac MIN_UNITS_PER_WORD
910 Minimum number of units in a word. If this is undefined, the default is
911 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
912 smallest value that @code{UNITS_PER_WORD} can have at run-time.
916 Width of a pointer, in bits. You must specify a value no wider than the
917 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
918 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
919 a value the default is @code{BITS_PER_WORD}.
922 @defmac POINTERS_EXTEND_UNSIGNED
923 A C expression that determines how pointers should be extended from
924 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
925 greater than zero if pointers should be zero-extended, zero if they
926 should be sign-extended, and negative if some other sort of conversion
927 is needed. In the last case, the extension is done by the target's
928 @code{ptr_extend} instruction.
930 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
931 and @code{word_mode} are all the same width.
934 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
935 A macro to update @var{m} and @var{unsignedp} when an object whose type
936 is @var{type} and which has the specified mode and signedness is to be
937 stored in a register. This macro is only called when @var{type} is a
940 On most RISC machines, which only have operations that operate on a full
941 register, define this macro to set @var{m} to @code{word_mode} if
942 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
943 cases, only integer modes should be widened because wider-precision
944 floating-point operations are usually more expensive than their narrower
947 For most machines, the macro definition does not change @var{unsignedp}.
948 However, some machines, have instructions that preferentially handle
949 either signed or unsigned quantities of certain modes. For example, on
950 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
951 sign-extend the result to 64 bits. On such machines, set
952 @var{unsignedp} according to which kind of extension is more efficient.
954 Do not define this macro if it would never modify @var{m}.
957 @deftypefn {Target Hook} {enum machine_mode} TARGET_PROMOTE_FUNCTION_MODE (const_tree @var{type}, enum machine_mode @var{mode}, int *@var{punsignedp}, const_tree @var{funtype}, int @var{for_return})
958 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
959 function return values. The target hook should return the new mode
960 and possibly change @code{*@var{punsignedp}} if the promotion should
961 change signedness. This function is called only for scalar @emph{or
964 @var{for_return} allows to distinguish the promotion of arguments and
965 return values. If it is @code{1}, a return value is being promoted and
966 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
967 If it is @code{2}, the returned mode should be that of the register in
968 which an incoming parameter is copied, or the outgoing result is computed;
969 then the hook should return the same mode as @code{promote_mode}, though
970 the signedness may be different.
972 @var{type} can be NULL when promoting function arguments of libcalls.
974 The default is to not promote arguments and return values. You can
975 also define the hook to @code{default_promote_function_mode_always_promote}
976 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
979 @defmac PARM_BOUNDARY
980 Normal alignment required for function parameters on the stack, in
981 bits. All stack parameters receive at least this much alignment
982 regardless of data type. On most machines, this is the same as the
986 @defmac STACK_BOUNDARY
987 Define this macro to the minimum alignment enforced by hardware for the
988 stack pointer on this machine. The definition is a C expression for the
989 desired alignment (measured in bits). This value is used as a default
990 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
991 this should be the same as @code{PARM_BOUNDARY}.
994 @defmac PREFERRED_STACK_BOUNDARY
995 Define this macro if you wish to preserve a certain alignment for the
996 stack pointer, greater than what the hardware enforces. The definition
997 is a C expression for the desired alignment (measured in bits). This
998 macro must evaluate to a value equal to or larger than
999 @code{STACK_BOUNDARY}.
1002 @defmac INCOMING_STACK_BOUNDARY
1003 Define this macro if the incoming stack boundary may be different
1004 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
1005 to a value equal to or larger than @code{STACK_BOUNDARY}.
1008 @defmac FUNCTION_BOUNDARY
1009 Alignment required for a function entry point, in bits.
1012 @defmac BIGGEST_ALIGNMENT
1013 Biggest alignment that any data type can require on this machine, in
1014 bits. Note that this is not the biggest alignment that is supported,
1015 just the biggest alignment that, when violated, may cause a fault.
1018 @defmac MALLOC_ABI_ALIGNMENT
1019 Alignment, in bits, a C conformant malloc implementation has to
1020 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1023 @defmac ATTRIBUTE_ALIGNED_VALUE
1024 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1025 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1028 @defmac MINIMUM_ATOMIC_ALIGNMENT
1029 If defined, the smallest alignment, in bits, that can be given to an
1030 object that can be referenced in one operation, without disturbing any
1031 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1032 on machines that don't have byte or half-word store operations.
1035 @defmac BIGGEST_FIELD_ALIGNMENT
1036 Biggest alignment that any structure or union field can require on this
1037 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1038 structure and union fields only, unless the field alignment has been set
1039 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1042 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1043 An expression for the alignment of a structure field @var{field} if the
1044 alignment computed in the usual way (including applying of
1045 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1046 alignment) is @var{computed}. It overrides alignment only if the
1047 field alignment has not been set by the
1048 @code{__attribute__ ((aligned (@var{n})))} construct.
1051 @defmac MAX_STACK_ALIGNMENT
1052 Biggest stack alignment guaranteed by the backend. Use this macro
1053 to specify the maximum alignment of a variable on stack.
1055 If not defined, the default value is @code{STACK_BOUNDARY}.
1057 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1058 @c But the fix for PR 32893 indicates that we can only guarantee
1059 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1060 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1063 @defmac MAX_OFILE_ALIGNMENT
1064 Biggest alignment supported by the object file format of this machine.
1065 Use this macro to limit the alignment which can be specified using the
1066 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1067 the default value is @code{BIGGEST_ALIGNMENT}.
1069 On systems that use ELF, the default (in @file{config/elfos.h}) is
1070 the largest supported 32-bit ELF section alignment representable on
1071 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1072 On 32-bit ELF the largest supported section alignment in bits is
1073 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1076 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1077 If defined, a C expression to compute the alignment for a variable in
1078 the static store. @var{type} is the data type, and @var{basic-align} is
1079 the alignment that the object would ordinarily have. The value of this
1080 macro is used instead of that alignment to align the object.
1082 If this macro is not defined, then @var{basic-align} is used.
1085 One use of this macro is to increase alignment of medium-size data to
1086 make it all fit in fewer cache lines. Another is to cause character
1087 arrays to be word-aligned so that @code{strcpy} calls that copy
1088 constants to character arrays can be done inline.
1091 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1092 If defined, a C expression to compute the alignment given to a constant
1093 that is being placed in memory. @var{constant} is the constant and
1094 @var{basic-align} is the alignment that the object would ordinarily
1095 have. The value of this macro is used instead of that alignment to
1098 If this macro is not defined, then @var{basic-align} is used.
1100 The typical use of this macro is to increase alignment for string
1101 constants to be word aligned so that @code{strcpy} calls that copy
1102 constants can be done inline.
1105 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1106 If defined, a C expression to compute the alignment for a variable in
1107 the local store. @var{type} is the data type, and @var{basic-align} is
1108 the alignment that the object would ordinarily have. The value of this
1109 macro is used instead of that alignment to align the object.
1111 If this macro is not defined, then @var{basic-align} is used.
1113 One use of this macro is to increase alignment of medium-size data to
1114 make it all fit in fewer cache lines.
1116 If the value of this macro has a type, it should be an unsigned type.
1119 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1120 If defined, a C expression to compute the alignment for stack slot.
1121 @var{type} is the data type, @var{mode} is the widest mode available,
1122 and @var{basic-align} is the alignment that the slot would ordinarily
1123 have. The value of this macro is used instead of that alignment to
1126 If this macro is not defined, then @var{basic-align} is used when
1127 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1130 This macro is to set alignment of stack slot to the maximum alignment
1131 of all possible modes which the slot may have.
1133 If the value of this macro has a type, it should be an unsigned type.
1136 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1137 If defined, a C expression to compute the alignment for a local
1138 variable @var{decl}.
1140 If this macro is not defined, then
1141 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1144 One use of this macro is to increase alignment of medium-size data to
1145 make it all fit in fewer cache lines.
1147 If the value of this macro has a type, it should be an unsigned type.
1150 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1151 If defined, a C expression to compute the minimum required alignment
1152 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1153 @var{mode}, assuming normal alignment @var{align}.
1155 If this macro is not defined, then @var{align} will be used.
1158 @defmac EMPTY_FIELD_BOUNDARY
1159 Alignment in bits to be given to a structure bit-field that follows an
1160 empty field such as @code{int : 0;}.
1162 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1165 @defmac STRUCTURE_SIZE_BOUNDARY
1166 Number of bits which any structure or union's size must be a multiple of.
1167 Each structure or union's size is rounded up to a multiple of this.
1169 If you do not define this macro, the default is the same as
1170 @code{BITS_PER_UNIT}.
1173 @defmac STRICT_ALIGNMENT
1174 Define this macro to be the value 1 if instructions will fail to work
1175 if given data not on the nominal alignment. If instructions will merely
1176 go slower in that case, define this macro as 0.
1179 @defmac PCC_BITFIELD_TYPE_MATTERS
1180 Define this if you wish to imitate the way many other C compilers handle
1181 alignment of bit-fields and the structures that contain them.
1183 The behavior is that the type written for a named bit-field (@code{int},
1184 @code{short}, or other integer type) imposes an alignment for the entire
1185 structure, as if the structure really did contain an ordinary field of
1186 that type. In addition, the bit-field is placed within the structure so
1187 that it would fit within such a field, not crossing a boundary for it.
1189 Thus, on most machines, a named bit-field whose type is written as
1190 @code{int} would not cross a four-byte boundary, and would force
1191 four-byte alignment for the whole structure. (The alignment used may
1192 not be four bytes; it is controlled by the other alignment parameters.)
1194 An unnamed bit-field will not affect the alignment of the containing
1197 If the macro is defined, its definition should be a C expression;
1198 a nonzero value for the expression enables this behavior.
1200 Note that if this macro is not defined, or its value is zero, some
1201 bit-fields may cross more than one alignment boundary. The compiler can
1202 support such references if there are @samp{insv}, @samp{extv}, and
1203 @samp{extzv} insns that can directly reference memory.
1205 The other known way of making bit-fields work is to define
1206 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1207 Then every structure can be accessed with fullwords.
1209 Unless the machine has bit-field instructions or you define
1210 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1211 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1213 If your aim is to make GCC use the same conventions for laying out
1214 bit-fields as are used by another compiler, here is how to investigate
1215 what the other compiler does. Compile and run this program:
1234 printf ("Size of foo1 is %d\n",
1235 sizeof (struct foo1));
1236 printf ("Size of foo2 is %d\n",
1237 sizeof (struct foo2));
1242 If this prints 2 and 5, then the compiler's behavior is what you would
1243 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1246 @defmac BITFIELD_NBYTES_LIMITED
1247 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1248 to aligning a bit-field within the structure.
1251 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1252 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1253 whether unnamed bitfields affect the alignment of the containing
1254 structure. The hook should return true if the structure should inherit
1255 the alignment requirements of an unnamed bitfield's type.
1258 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1259 This target hook should return @code{true} if accesses to volatile bitfields
1260 should use the narrowest mode possible. It should return @code{false} if
1261 these accesses should use the bitfield container type.
1263 The default is @code{!TARGET_STRICT_ALIGN}.
1266 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1267 Return 1 if a structure or array containing @var{field} should be accessed using
1270 If @var{field} is the only field in the structure, @var{mode} is its
1271 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1272 case where structures of one field would require the structure's mode to
1273 retain the field's mode.
1275 Normally, this is not needed.
1278 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1279 Define this macro as an expression for the alignment of a type (given
1280 by @var{type} as a tree node) if the alignment computed in the usual
1281 way is @var{computed} and the alignment explicitly specified was
1284 The default is to use @var{specified} if it is larger; otherwise, use
1285 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1288 @defmac MAX_FIXED_MODE_SIZE
1289 An integer expression for the size in bits of the largest integer
1290 machine mode that should actually be used. All integer machine modes of
1291 this size or smaller can be used for structures and unions with the
1292 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1293 (DImode)} is assumed.
1296 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1297 If defined, an expression of type @code{enum machine_mode} that
1298 specifies the mode of the save area operand of a
1299 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1300 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1301 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1302 having its mode specified.
1304 You need not define this macro if it always returns @code{Pmode}. You
1305 would most commonly define this macro if the
1306 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1310 @defmac STACK_SIZE_MODE
1311 If defined, an expression of type @code{enum machine_mode} that
1312 specifies the mode of the size increment operand of an
1313 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1315 You need not define this macro if it always returns @code{word_mode}.
1316 You would most commonly define this macro if the @code{allocate_stack}
1317 pattern needs to support both a 32- and a 64-bit mode.
1320 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE (void)
1321 This target hook should return the mode to be used for the return value
1322 of compare instructions expanded to libgcc calls. If not defined
1323 @code{word_mode} is returned which is the right choice for a majority of
1327 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1328 This target hook should return the mode to be used for the shift count operand
1329 of shift instructions expanded to libgcc calls. If not defined
1330 @code{word_mode} is returned which is the right choice for a majority of
1334 @deftypefn {Target Hook} {enum machine_mode} TARGET_UNWIND_WORD_MODE (void)
1335 Return machine mode to be used for @code{_Unwind_Word} type.
1336 The default is to use @code{word_mode}.
1339 @defmac ROUND_TOWARDS_ZERO
1340 If defined, this macro should be true if the prevailing rounding
1341 mode is towards zero.
1343 Defining this macro only affects the way @file{libgcc.a} emulates
1344 floating-point arithmetic.
1346 Not defining this macro is equivalent to returning zero.
1349 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1350 This macro should return true if floats with @var{size}
1351 bits do not have a NaN or infinity representation, but use the largest
1352 exponent for normal numbers instead.
1354 Defining this macro only affects the way @file{libgcc.a} emulates
1355 floating-point arithmetic.
1357 The default definition of this macro returns false for all sizes.
1360 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1361 This target hook returns @code{true} if bit-fields in the given
1362 @var{record_type} are to be laid out following the rules of Microsoft
1363 Visual C/C++, namely: (i) a bit-field won't share the same storage
1364 unit with the previous bit-field if their underlying types have
1365 different sizes, and the bit-field will be aligned to the highest
1366 alignment of the underlying types of itself and of the previous
1367 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1368 the whole enclosing structure, even if it is unnamed; except that
1369 (iii) a zero-sized bit-field will be disregarded unless it follows
1370 another bit-field of nonzero size. If this hook returns @code{true},
1371 other macros that control bit-field layout are ignored.
1373 When a bit-field is inserted into a packed record, the whole size
1374 of the underlying type is used by one or more same-size adjacent
1375 bit-fields (that is, if its long:3, 32 bits is used in the record,
1376 and any additional adjacent long bit-fields are packed into the same
1377 chunk of 32 bits. However, if the size changes, a new field of that
1378 size is allocated). In an unpacked record, this is the same as using
1379 alignment, but not equivalent when packing.
1381 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1382 the latter will take precedence. If @samp{__attribute__((packed))} is
1383 used on a single field when MS bit-fields are in use, it will take
1384 precedence for that field, but the alignment of the rest of the structure
1385 may affect its placement.
1388 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1389 Returns true if the target supports decimal floating point.
1392 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1393 Returns true if the target supports fixed-point arithmetic.
1396 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1397 This hook is called just before expansion into rtl, allowing the target
1398 to perform additional initializations or analysis before the expansion.
1399 For example, the rs6000 port uses it to allocate a scratch stack slot
1400 for use in copying SDmode values between memory and floating point
1401 registers whenever the function being expanded has any SDmode
1405 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1406 This hook allows the backend to perform additional instantiations on rtl
1407 that are not actually in any insns yet, but will be later.
1410 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1411 If your target defines any fundamental types, or any types your target
1412 uses should be mangled differently from the default, define this hook
1413 to return the appropriate encoding for these types as part of a C++
1414 mangled name. The @var{type} argument is the tree structure representing
1415 the type to be mangled. The hook may be applied to trees which are
1416 not target-specific fundamental types; it should return @code{NULL}
1417 for all such types, as well as arguments it does not recognize. If the
1418 return value is not @code{NULL}, it must point to a statically-allocated
1421 Target-specific fundamental types might be new fundamental types or
1422 qualified versions of ordinary fundamental types. Encode new
1423 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1424 is the name used for the type in source code, and @var{n} is the
1425 length of @var{name} in decimal. Encode qualified versions of
1426 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1427 @var{name} is the name used for the type qualifier in source code,
1428 @var{n} is the length of @var{name} as above, and @var{code} is the
1429 code used to represent the unqualified version of this type. (See
1430 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1431 codes.) In both cases the spaces are for clarity; do not include any
1432 spaces in your string.
1434 This hook is applied to types prior to typedef resolution. If the mangled
1435 name for a particular type depends only on that type's main variant, you
1436 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1439 The default version of this hook always returns @code{NULL}, which is
1440 appropriate for a target that does not define any new fundamental
1445 @section Layout of Source Language Data Types
1447 These macros define the sizes and other characteristics of the standard
1448 basic data types used in programs being compiled. Unlike the macros in
1449 the previous section, these apply to specific features of C and related
1450 languages, rather than to fundamental aspects of storage layout.
1452 @defmac INT_TYPE_SIZE
1453 A C expression for the size in bits of the type @code{int} on the
1454 target machine. If you don't define this, the default is one word.
1457 @defmac SHORT_TYPE_SIZE
1458 A C expression for the size in bits of the type @code{short} on the
1459 target machine. If you don't define this, the default is half a word.
1460 (If this would be less than one storage unit, it is rounded up to one
1464 @defmac LONG_TYPE_SIZE
1465 A C expression for the size in bits of the type @code{long} on the
1466 target machine. If you don't define this, the default is one word.
1469 @defmac ADA_LONG_TYPE_SIZE
1470 On some machines, the size used for the Ada equivalent of the type
1471 @code{long} by a native Ada compiler differs from that used by C@. In
1472 that situation, define this macro to be a C expression to be used for
1473 the size of that type. If you don't define this, the default is the
1474 value of @code{LONG_TYPE_SIZE}.
1477 @defmac LONG_LONG_TYPE_SIZE
1478 A C expression for the size in bits of the type @code{long long} on the
1479 target machine. If you don't define this, the default is two
1480 words. If you want to support GNU Ada on your machine, the value of this
1481 macro must be at least 64.
1484 @defmac CHAR_TYPE_SIZE
1485 A C expression for the size in bits of the type @code{char} on the
1486 target machine. If you don't define this, the default is
1487 @code{BITS_PER_UNIT}.
1490 @defmac BOOL_TYPE_SIZE
1491 A C expression for the size in bits of the C++ type @code{bool} and
1492 C99 type @code{_Bool} on the target machine. If you don't define
1493 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1496 @defmac FLOAT_TYPE_SIZE
1497 A C expression for the size in bits of the type @code{float} on the
1498 target machine. If you don't define this, the default is one word.
1501 @defmac DOUBLE_TYPE_SIZE
1502 A C expression for the size in bits of the type @code{double} on the
1503 target machine. If you don't define this, the default is two
1507 @defmac LONG_DOUBLE_TYPE_SIZE
1508 A C expression for the size in bits of the type @code{long double} on
1509 the target machine. If you don't define this, the default is two
1513 @defmac SHORT_FRACT_TYPE_SIZE
1514 A C expression for the size in bits of the type @code{short _Fract} on
1515 the target machine. If you don't define this, the default is
1516 @code{BITS_PER_UNIT}.
1519 @defmac FRACT_TYPE_SIZE
1520 A C expression for the size in bits of the type @code{_Fract} on
1521 the target machine. If you don't define this, the default is
1522 @code{BITS_PER_UNIT * 2}.
1525 @defmac LONG_FRACT_TYPE_SIZE
1526 A C expression for the size in bits of the type @code{long _Fract} on
1527 the target machine. If you don't define this, the default is
1528 @code{BITS_PER_UNIT * 4}.
1531 @defmac LONG_LONG_FRACT_TYPE_SIZE
1532 A C expression for the size in bits of the type @code{long long _Fract} on
1533 the target machine. If you don't define this, the default is
1534 @code{BITS_PER_UNIT * 8}.
1537 @defmac SHORT_ACCUM_TYPE_SIZE
1538 A C expression for the size in bits of the type @code{short _Accum} on
1539 the target machine. If you don't define this, the default is
1540 @code{BITS_PER_UNIT * 2}.
1543 @defmac ACCUM_TYPE_SIZE
1544 A C expression for the size in bits of the type @code{_Accum} on
1545 the target machine. If you don't define this, the default is
1546 @code{BITS_PER_UNIT * 4}.
1549 @defmac LONG_ACCUM_TYPE_SIZE
1550 A C expression for the size in bits of the type @code{long _Accum} on
1551 the target machine. If you don't define this, the default is
1552 @code{BITS_PER_UNIT * 8}.
1555 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1556 A C expression for the size in bits of the type @code{long long _Accum} on
1557 the target machine. If you don't define this, the default is
1558 @code{BITS_PER_UNIT * 16}.
1561 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1562 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1563 if you want routines in @file{libgcc2.a} for a size other than
1564 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1565 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1568 @defmac LIBGCC2_HAS_DF_MODE
1569 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1570 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1571 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1572 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1573 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1577 @defmac LIBGCC2_HAS_XF_MODE
1578 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1579 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1580 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1581 is 80 then the default is 1, otherwise it is 0.
1584 @defmac LIBGCC2_HAS_TF_MODE
1585 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1586 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1587 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1588 is 128 then the default is 1, otherwise it is 0.
1591 @defmac LIBGCC2_GNU_PREFIX
1592 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1593 hook and should be defined if that hook is overriden to be true. It
1594 causes function names in libgcc to be changed to use a @code{__gnu_}
1595 prefix for their name rather than the default @code{__}. A port which
1596 uses this macro should also arrange to use @file{t-gnu-prefix} in
1597 the libgcc @file{config.host}.
1604 Define these macros to be the size in bits of the mantissa of
1605 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1606 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1607 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1608 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1609 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1610 @code{DOUBLE_TYPE_SIZE} or
1611 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1614 @defmac TARGET_FLT_EVAL_METHOD
1615 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1616 assuming, if applicable, that the floating-point control word is in its
1617 default state. If you do not define this macro the value of
1618 @code{FLT_EVAL_METHOD} will be zero.
1621 @defmac WIDEST_HARDWARE_FP_SIZE
1622 A C expression for the size in bits of the widest floating-point format
1623 supported by the hardware. If you define this macro, you must specify a
1624 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1625 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1629 @defmac DEFAULT_SIGNED_CHAR
1630 An expression whose value is 1 or 0, according to whether the type
1631 @code{char} should be signed or unsigned by default. The user can
1632 always override this default with the options @option{-fsigned-char}
1633 and @option{-funsigned-char}.
1636 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1637 This target hook should return true if the compiler should give an
1638 @code{enum} type only as many bytes as it takes to represent the range
1639 of possible values of that type. It should return false if all
1640 @code{enum} types should be allocated like @code{int}.
1642 The default is to return false.
1646 A C expression for a string describing the name of the data type to use
1647 for size values. The typedef name @code{size_t} is defined using the
1648 contents of the string.
1650 The string can contain more than one keyword. If so, separate them with
1651 spaces, and write first any length keyword, then @code{unsigned} if
1652 appropriate, and finally @code{int}. The string must exactly match one
1653 of the data type names defined in the function
1654 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1655 omit @code{int} or change the order---that would cause the compiler to
1658 If you don't define this macro, the default is @code{"long unsigned
1662 @defmac PTRDIFF_TYPE
1663 A C expression for a string describing the name of the data type to use
1664 for the result of subtracting two pointers. The typedef name
1665 @code{ptrdiff_t} is defined using the contents of the string. See
1666 @code{SIZE_TYPE} above for more information.
1668 If you don't define this macro, the default is @code{"long int"}.
1672 A C expression for a string describing the name of the data type to use
1673 for wide characters. The typedef name @code{wchar_t} is defined using
1674 the contents of the string. See @code{SIZE_TYPE} above for more
1677 If you don't define this macro, the default is @code{"int"}.
1680 @defmac WCHAR_TYPE_SIZE
1681 A C expression for the size in bits of the data type for wide
1682 characters. This is used in @code{cpp}, which cannot make use of
1687 A C expression for a string describing the name of the data type to
1688 use for wide characters passed to @code{printf} and returned from
1689 @code{getwc}. The typedef name @code{wint_t} is defined using the
1690 contents of the string. See @code{SIZE_TYPE} above for more
1693 If you don't define this macro, the default is @code{"unsigned int"}.
1697 A C expression for a string describing the name of the data type that
1698 can represent any value of any standard or extended signed integer type.
1699 The typedef name @code{intmax_t} is defined using the contents of the
1700 string. See @code{SIZE_TYPE} above for more information.
1702 If you don't define this macro, the default is the first of
1703 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1704 much precision as @code{long long int}.
1707 @defmac UINTMAX_TYPE
1708 A C expression for a string describing the name of the data type that
1709 can represent any value of any standard or extended unsigned integer
1710 type. The typedef name @code{uintmax_t} is defined using the contents
1711 of the string. See @code{SIZE_TYPE} above for more information.
1713 If you don't define this macro, the default is the first of
1714 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1715 unsigned int"} that has as much precision as @code{long long unsigned
1719 @defmac SIG_ATOMIC_TYPE
1725 @defmacx UINT16_TYPE
1726 @defmacx UINT32_TYPE
1727 @defmacx UINT64_TYPE
1728 @defmacx INT_LEAST8_TYPE
1729 @defmacx INT_LEAST16_TYPE
1730 @defmacx INT_LEAST32_TYPE
1731 @defmacx INT_LEAST64_TYPE
1732 @defmacx UINT_LEAST8_TYPE
1733 @defmacx UINT_LEAST16_TYPE
1734 @defmacx UINT_LEAST32_TYPE
1735 @defmacx UINT_LEAST64_TYPE
1736 @defmacx INT_FAST8_TYPE
1737 @defmacx INT_FAST16_TYPE
1738 @defmacx INT_FAST32_TYPE
1739 @defmacx INT_FAST64_TYPE
1740 @defmacx UINT_FAST8_TYPE
1741 @defmacx UINT_FAST16_TYPE
1742 @defmacx UINT_FAST32_TYPE
1743 @defmacx UINT_FAST64_TYPE
1744 @defmacx INTPTR_TYPE
1745 @defmacx UINTPTR_TYPE
1746 C expressions for the standard types @code{sig_atomic_t},
1747 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1748 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1749 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1750 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1751 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1752 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1753 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1754 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1755 @code{SIZE_TYPE} above for more information.
1757 If any of these macros evaluates to a null pointer, the corresponding
1758 type is not supported; if GCC is configured to provide
1759 @code{<stdint.h>} in such a case, the header provided may not conform
1760 to C99, depending on the type in question. The defaults for all of
1761 these macros are null pointers.
1764 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1765 The C++ compiler represents a pointer-to-member-function with a struct
1772 ptrdiff_t vtable_index;
1779 The C++ compiler must use one bit to indicate whether the function that
1780 will be called through a pointer-to-member-function is virtual.
1781 Normally, we assume that the low-order bit of a function pointer must
1782 always be zero. Then, by ensuring that the vtable_index is odd, we can
1783 distinguish which variant of the union is in use. But, on some
1784 platforms function pointers can be odd, and so this doesn't work. In
1785 that case, we use the low-order bit of the @code{delta} field, and shift
1786 the remainder of the @code{delta} field to the left.
1788 GCC will automatically make the right selection about where to store
1789 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1790 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1791 set such that functions always start at even addresses, but the lowest
1792 bit of pointers to functions indicate whether the function at that
1793 address is in ARM or Thumb mode. If this is the case of your
1794 architecture, you should define this macro to
1795 @code{ptrmemfunc_vbit_in_delta}.
1797 In general, you should not have to define this macro. On architectures
1798 in which function addresses are always even, according to
1799 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1800 @code{ptrmemfunc_vbit_in_pfn}.
1803 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1804 Normally, the C++ compiler uses function pointers in vtables. This
1805 macro allows the target to change to use ``function descriptors''
1806 instead. Function descriptors are found on targets for whom a
1807 function pointer is actually a small data structure. Normally the
1808 data structure consists of the actual code address plus a data
1809 pointer to which the function's data is relative.
1811 If vtables are used, the value of this macro should be the number
1812 of words that the function descriptor occupies.
1815 @defmac TARGET_VTABLE_ENTRY_ALIGN
1816 By default, the vtable entries are void pointers, the so the alignment
1817 is the same as pointer alignment. The value of this macro specifies
1818 the alignment of the vtable entry in bits. It should be defined only
1819 when special alignment is necessary. */
1822 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1823 There are a few non-descriptor entries in the vtable at offsets below
1824 zero. If these entries must be padded (say, to preserve the alignment
1825 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1826 of words in each data entry.
1830 @section Register Usage
1831 @cindex register usage
1833 This section explains how to describe what registers the target machine
1834 has, and how (in general) they can be used.
1836 The description of which registers a specific instruction can use is
1837 done with register classes; see @ref{Register Classes}. For information
1838 on using registers to access a stack frame, see @ref{Frame Registers}.
1839 For passing values in registers, see @ref{Register Arguments}.
1840 For returning values in registers, see @ref{Scalar Return}.
1843 * Register Basics:: Number and kinds of registers.
1844 * Allocation Order:: Order in which registers are allocated.
1845 * Values in Registers:: What kinds of values each reg can hold.
1846 * Leaf Functions:: Renumbering registers for leaf functions.
1847 * Stack Registers:: Handling a register stack such as 80387.
1850 @node Register Basics
1851 @subsection Basic Characteristics of Registers
1853 @c prevent bad page break with this line
1854 Registers have various characteristics.
1856 @defmac FIRST_PSEUDO_REGISTER
1857 Number of hardware registers known to the compiler. They receive
1858 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1859 pseudo register's number really is assigned the number
1860 @code{FIRST_PSEUDO_REGISTER}.
1863 @defmac FIXED_REGISTERS
1864 @cindex fixed register
1865 An initializer that says which registers are used for fixed purposes
1866 all throughout the compiled code and are therefore not available for
1867 general allocation. These would include the stack pointer, the frame
1868 pointer (except on machines where that can be used as a general
1869 register when no frame pointer is needed), the program counter on
1870 machines where that is considered one of the addressable registers,
1871 and any other numbered register with a standard use.
1873 This information is expressed as a sequence of numbers, separated by
1874 commas and surrounded by braces. The @var{n}th number is 1 if
1875 register @var{n} is fixed, 0 otherwise.
1877 The table initialized from this macro, and the table initialized by
1878 the following one, may be overridden at run time either automatically,
1879 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1880 the user with the command options @option{-ffixed-@var{reg}},
1881 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1884 @defmac CALL_USED_REGISTERS
1885 @cindex call-used register
1886 @cindex call-clobbered register
1887 @cindex call-saved register
1888 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1889 clobbered (in general) by function calls as well as for fixed
1890 registers. This macro therefore identifies the registers that are not
1891 available for general allocation of values that must live across
1894 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1895 automatically saves it on function entry and restores it on function
1896 exit, if the register is used within the function.
1899 @defmac CALL_REALLY_USED_REGISTERS
1900 @cindex call-used register
1901 @cindex call-clobbered register
1902 @cindex call-saved register
1903 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1904 that the entire set of @code{FIXED_REGISTERS} be included.
1905 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1906 This macro is optional. If not specified, it defaults to the value
1907 of @code{CALL_USED_REGISTERS}.
1910 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1911 @cindex call-used register
1912 @cindex call-clobbered register
1913 @cindex call-saved register
1914 A C expression that is nonzero if it is not permissible to store a
1915 value of mode @var{mode} in hard register number @var{regno} across a
1916 call without some part of it being clobbered. For most machines this
1917 macro need not be defined. It is only required for machines that do not
1918 preserve the entire contents of a register across a call.
1922 @findex call_used_regs
1925 @findex reg_class_contents
1926 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1927 This hook may conditionally modify five variables
1928 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1929 @code{reg_names}, and @code{reg_class_contents}, to take into account
1930 any dependence of these register sets on target flags. The first three
1931 of these are of type @code{char []} (interpreted as Boolean vectors).
1932 @code{global_regs} is a @code{const char *[]}, and
1933 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1934 called, @code{fixed_regs}, @code{call_used_regs},
1935 @code{reg_class_contents}, and @code{reg_names} have been initialized
1936 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1937 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1938 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1939 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1940 command options have been applied.
1942 @cindex disabling certain registers
1943 @cindex controlling register usage
1944 If the usage of an entire class of registers depends on the target
1945 flags, you may indicate this to GCC by using this macro to modify
1946 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1947 registers in the classes which should not be used by GCC@. Also define
1948 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1949 to return @code{NO_REGS} if it
1950 is called with a letter for a class that shouldn't be used.
1952 (However, if this class is not included in @code{GENERAL_REGS} and all
1953 of the insn patterns whose constraints permit this class are
1954 controlled by target switches, then GCC will automatically avoid using
1955 these registers when the target switches are opposed to them.)
1958 @defmac INCOMING_REGNO (@var{out})
1959 Define this macro if the target machine has register windows. This C
1960 expression returns the register number as seen by the called function
1961 corresponding to the register number @var{out} as seen by the calling
1962 function. Return @var{out} if register number @var{out} is not an
1966 @defmac OUTGOING_REGNO (@var{in})
1967 Define this macro if the target machine has register windows. This C
1968 expression returns the register number as seen by the calling function
1969 corresponding to the register number @var{in} as seen by the called
1970 function. Return @var{in} if register number @var{in} is not an inbound
1974 @defmac LOCAL_REGNO (@var{regno})
1975 Define this macro if the target machine has register windows. This C
1976 expression returns true if the register is call-saved but is in the
1977 register window. Unlike most call-saved registers, such registers
1978 need not be explicitly restored on function exit or during non-local
1983 If the program counter has a register number, define this as that
1984 register number. Otherwise, do not define it.
1987 @node Allocation Order
1988 @subsection Order of Allocation of Registers
1989 @cindex order of register allocation
1990 @cindex register allocation order
1992 @c prevent bad page break with this line
1993 Registers are allocated in order.
1995 @defmac REG_ALLOC_ORDER
1996 If defined, an initializer for a vector of integers, containing the
1997 numbers of hard registers in the order in which GCC should prefer
1998 to use them (from most preferred to least).
2000 If this macro is not defined, registers are used lowest numbered first
2001 (all else being equal).
2003 One use of this macro is on machines where the highest numbered
2004 registers must always be saved and the save-multiple-registers
2005 instruction supports only sequences of consecutive registers. On such
2006 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2007 the highest numbered allocable register first.
2010 @defmac ADJUST_REG_ALLOC_ORDER
2011 A C statement (sans semicolon) to choose the order in which to allocate
2012 hard registers for pseudo-registers local to a basic block.
2014 Store the desired register order in the array @code{reg_alloc_order}.
2015 Element 0 should be the register to allocate first; element 1, the next
2016 register; and so on.
2018 The macro body should not assume anything about the contents of
2019 @code{reg_alloc_order} before execution of the macro.
2021 On most machines, it is not necessary to define this macro.
2024 @defmac HONOR_REG_ALLOC_ORDER
2025 Normally, IRA tries to estimate the costs for saving a register in the
2026 prologue and restoring it in the epilogue. This discourages it from
2027 using call-saved registers. If a machine wants to ensure that IRA
2028 allocates registers in the order given by REG_ALLOC_ORDER even if some
2029 call-saved registers appear earlier than call-used ones, this macro
2033 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2034 In some case register allocation order is not enough for the
2035 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2036 If this macro is defined, it should return a floating point value
2037 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2038 be increased by approximately the pseudo's usage frequency times the
2039 value returned by this macro. Not defining this macro is equivalent
2040 to having it always return @code{0.0}.
2042 On most machines, it is not necessary to define this macro.
2045 @node Values in Registers
2046 @subsection How Values Fit in Registers
2048 This section discusses the macros that describe which kinds of values
2049 (specifically, which machine modes) each register can hold, and how many
2050 consecutive registers are needed for a given mode.
2052 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2053 A C expression for the number of consecutive hard registers, starting
2054 at register number @var{regno}, required to hold a value of mode
2055 @var{mode}. This macro must never return zero, even if a register
2056 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2057 and/or CANNOT_CHANGE_MODE_CLASS instead.
2059 On a machine where all registers are exactly one word, a suitable
2060 definition of this macro is
2063 #define HARD_REGNO_NREGS(REGNO, MODE) \
2064 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2069 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2070 A C expression that is nonzero if a value of mode @var{mode}, stored
2071 in memory, ends with padding that causes it to take up more space than
2072 in registers starting at register number @var{regno} (as determined by
2073 multiplying GCC's notion of the size of the register when containing
2074 this mode by the number of registers returned by
2075 @code{HARD_REGNO_NREGS}). By default this is zero.
2077 For example, if a floating-point value is stored in three 32-bit
2078 registers but takes up 128 bits in memory, then this would be
2081 This macros only needs to be defined if there are cases where
2082 @code{subreg_get_info}
2083 would otherwise wrongly determine that a @code{subreg} can be
2084 represented by an offset to the register number, when in fact such a
2085 @code{subreg} would contain some of the padding not stored in
2086 registers and so not be representable.
2089 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2090 For values of @var{regno} and @var{mode} for which
2091 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2092 returning the greater number of registers required to hold the value
2093 including any padding. In the example above, the value would be four.
2096 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2097 Define this macro if the natural size of registers that hold values
2098 of mode @var{mode} is not the word size. It is a C expression that
2099 should give the natural size in bytes for the specified mode. It is
2100 used by the register allocator to try to optimize its results. This
2101 happens for example on SPARC 64-bit where the natural size of
2102 floating-point registers is still 32-bit.
2105 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2106 A C expression that is nonzero if it is permissible to store a value
2107 of mode @var{mode} in hard register number @var{regno} (or in several
2108 registers starting with that one). For a machine where all registers
2109 are equivalent, a suitable definition is
2112 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2115 You need not include code to check for the numbers of fixed registers,
2116 because the allocation mechanism considers them to be always occupied.
2118 @cindex register pairs
2119 On some machines, double-precision values must be kept in even/odd
2120 register pairs. You can implement that by defining this macro to reject
2121 odd register numbers for such modes.
2123 The minimum requirement for a mode to be OK in a register is that the
2124 @samp{mov@var{mode}} instruction pattern support moves between the
2125 register and other hard register in the same class and that moving a
2126 value into the register and back out not alter it.
2128 Since the same instruction used to move @code{word_mode} will work for
2129 all narrower integer modes, it is not necessary on any machine for
2130 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2131 you define patterns @samp{movhi}, etc., to take advantage of this. This
2132 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2133 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2136 Many machines have special registers for floating point arithmetic.
2137 Often people assume that floating point machine modes are allowed only
2138 in floating point registers. This is not true. Any registers that
2139 can hold integers can safely @emph{hold} a floating point machine
2140 mode, whether or not floating arithmetic can be done on it in those
2141 registers. Integer move instructions can be used to move the values.
2143 On some machines, though, the converse is true: fixed-point machine
2144 modes may not go in floating registers. This is true if the floating
2145 registers normalize any value stored in them, because storing a
2146 non-floating value there would garble it. In this case,
2147 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2148 floating registers. But if the floating registers do not automatically
2149 normalize, if you can store any bit pattern in one and retrieve it
2150 unchanged without a trap, then any machine mode may go in a floating
2151 register, so you can define this macro to say so.
2153 The primary significance of special floating registers is rather that
2154 they are the registers acceptable in floating point arithmetic
2155 instructions. However, this is of no concern to
2156 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2157 constraints for those instructions.
2159 On some machines, the floating registers are especially slow to access,
2160 so that it is better to store a value in a stack frame than in such a
2161 register if floating point arithmetic is not being done. As long as the
2162 floating registers are not in class @code{GENERAL_REGS}, they will not
2163 be used unless some pattern's constraint asks for one.
2166 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2167 A C expression that is nonzero if it is OK to rename a hard register
2168 @var{from} to another hard register @var{to}.
2170 One common use of this macro is to prevent renaming of a register to
2171 another register that is not saved by a prologue in an interrupt
2174 The default is always nonzero.
2177 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2178 A C expression that is nonzero if a value of mode
2179 @var{mode1} is accessible in mode @var{mode2} without copying.
2181 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2182 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2183 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2184 should be nonzero. If they differ for any @var{r}, you should define
2185 this macro to return zero unless some other mechanism ensures the
2186 accessibility of the value in a narrower mode.
2188 You should define this macro to return nonzero in as many cases as
2189 possible since doing so will allow GCC to perform better register
2193 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2194 This target hook should return @code{true} if it is OK to use a hard register
2195 @var{regno} as scratch reg in peephole2.
2197 One common use of this macro is to prevent using of a register that
2198 is not saved by a prologue in an interrupt handler.
2200 The default version of this hook always returns @code{true}.
2203 @defmac AVOID_CCMODE_COPIES
2204 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2205 registers. You should only define this macro if support for copying to/from
2206 @code{CCmode} is incomplete.
2209 @node Leaf Functions
2210 @subsection Handling Leaf Functions
2212 @cindex leaf functions
2213 @cindex functions, leaf
2214 On some machines, a leaf function (i.e., one which makes no calls) can run
2215 more efficiently if it does not make its own register window. Often this
2216 means it is required to receive its arguments in the registers where they
2217 are passed by the caller, instead of the registers where they would
2220 The special treatment for leaf functions generally applies only when
2221 other conditions are met; for example, often they may use only those
2222 registers for its own variables and temporaries. We use the term ``leaf
2223 function'' to mean a function that is suitable for this special
2224 handling, so that functions with no calls are not necessarily ``leaf
2227 GCC assigns register numbers before it knows whether the function is
2228 suitable for leaf function treatment. So it needs to renumber the
2229 registers in order to output a leaf function. The following macros
2232 @defmac LEAF_REGISTERS
2233 Name of a char vector, indexed by hard register number, which
2234 contains 1 for a register that is allowable in a candidate for leaf
2237 If leaf function treatment involves renumbering the registers, then the
2238 registers marked here should be the ones before renumbering---those that
2239 GCC would ordinarily allocate. The registers which will actually be
2240 used in the assembler code, after renumbering, should not be marked with 1
2243 Define this macro only if the target machine offers a way to optimize
2244 the treatment of leaf functions.
2247 @defmac LEAF_REG_REMAP (@var{regno})
2248 A C expression whose value is the register number to which @var{regno}
2249 should be renumbered, when a function is treated as a leaf function.
2251 If @var{regno} is a register number which should not appear in a leaf
2252 function before renumbering, then the expression should yield @minus{}1, which
2253 will cause the compiler to abort.
2255 Define this macro only if the target machine offers a way to optimize the
2256 treatment of leaf functions, and registers need to be renumbered to do
2260 @findex current_function_is_leaf
2261 @findex current_function_uses_only_leaf_regs
2262 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2263 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2264 specially. They can test the C variable @code{current_function_is_leaf}
2265 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2266 set prior to local register allocation and is valid for the remaining
2267 compiler passes. They can also test the C variable
2268 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2269 functions which only use leaf registers.
2270 @code{current_function_uses_only_leaf_regs} is valid after all passes
2271 that modify the instructions have been run and is only useful if
2272 @code{LEAF_REGISTERS} is defined.
2273 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2274 @c of the next paragraph?! --mew 2feb93
2276 @node Stack Registers
2277 @subsection Registers That Form a Stack
2279 There are special features to handle computers where some of the
2280 ``registers'' form a stack. Stack registers are normally written by
2281 pushing onto the stack, and are numbered relative to the top of the
2284 Currently, GCC can only handle one group of stack-like registers, and
2285 they must be consecutively numbered. Furthermore, the existing
2286 support for stack-like registers is specific to the 80387 floating
2287 point coprocessor. If you have a new architecture that uses
2288 stack-like registers, you will need to do substantial work on
2289 @file{reg-stack.c} and write your machine description to cooperate
2290 with it, as well as defining these macros.
2293 Define this if the machine has any stack-like registers.
2296 @defmac STACK_REG_COVER_CLASS
2297 This is a cover class containing the stack registers. Define this if
2298 the machine has any stack-like registers.
2301 @defmac FIRST_STACK_REG
2302 The number of the first stack-like register. This one is the top
2306 @defmac LAST_STACK_REG
2307 The number of the last stack-like register. This one is the bottom of
2311 @node Register Classes
2312 @section Register Classes
2313 @cindex register class definitions
2314 @cindex class definitions, register
2316 On many machines, the numbered registers are not all equivalent.
2317 For example, certain registers may not be allowed for indexed addressing;
2318 certain registers may not be allowed in some instructions. These machine
2319 restrictions are described to the compiler using @dfn{register classes}.
2321 You define a number of register classes, giving each one a name and saying
2322 which of the registers belong to it. Then you can specify register classes
2323 that are allowed as operands to particular instruction patterns.
2327 In general, each register will belong to several classes. In fact, one
2328 class must be named @code{ALL_REGS} and contain all the registers. Another
2329 class must be named @code{NO_REGS} and contain no registers. Often the
2330 union of two classes will be another class; however, this is not required.
2332 @findex GENERAL_REGS
2333 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2334 terribly special about the name, but the operand constraint letters
2335 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2336 the same as @code{ALL_REGS}, just define it as a macro which expands
2339 Order the classes so that if class @var{x} is contained in class @var{y}
2340 then @var{x} has a lower class number than @var{y}.
2342 The way classes other than @code{GENERAL_REGS} are specified in operand
2343 constraints is through machine-dependent operand constraint letters.
2344 You can define such letters to correspond to various classes, then use
2345 them in operand constraints.
2347 You must define the narrowest register classes for allocatable
2348 registers, so that each class either has no subclasses, or that for
2349 some mode, the move cost between registers within the class is
2350 cheaper than moving a register in the class to or from memory
2353 You should define a class for the union of two classes whenever some
2354 instruction allows both classes. For example, if an instruction allows
2355 either a floating point (coprocessor) register or a general register for a
2356 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2357 which includes both of them. Otherwise you will get suboptimal code,
2358 or even internal compiler errors when reload cannot find a register in the
2359 class computed via @code{reg_class_subunion}.
2361 You must also specify certain redundant information about the register
2362 classes: for each class, which classes contain it and which ones are
2363 contained in it; for each pair of classes, the largest class contained
2366 When a value occupying several consecutive registers is expected in a
2367 certain class, all the registers used must belong to that class.
2368 Therefore, register classes cannot be used to enforce a requirement for
2369 a register pair to start with an even-numbered register. The way to
2370 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2372 Register classes used for input-operands of bitwise-and or shift
2373 instructions have a special requirement: each such class must have, for
2374 each fixed-point machine mode, a subclass whose registers can transfer that
2375 mode to or from memory. For example, on some machines, the operations for
2376 single-byte values (@code{QImode}) are limited to certain registers. When
2377 this is so, each register class that is used in a bitwise-and or shift
2378 instruction must have a subclass consisting of registers from which
2379 single-byte values can be loaded or stored. This is so that
2380 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2382 @deftp {Data type} {enum reg_class}
2383 An enumerated type that must be defined with all the register class names
2384 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2385 must be the last register class, followed by one more enumerated value,
2386 @code{LIM_REG_CLASSES}, which is not a register class but rather
2387 tells how many classes there are.
2389 Each register class has a number, which is the value of casting
2390 the class name to type @code{int}. The number serves as an index
2391 in many of the tables described below.
2394 @defmac N_REG_CLASSES
2395 The number of distinct register classes, defined as follows:
2398 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2402 @defmac REG_CLASS_NAMES
2403 An initializer containing the names of the register classes as C string
2404 constants. These names are used in writing some of the debugging dumps.
2407 @defmac REG_CLASS_CONTENTS
2408 An initializer containing the contents of the register classes, as integers
2409 which are bit masks. The @var{n}th integer specifies the contents of class
2410 @var{n}. The way the integer @var{mask} is interpreted is that
2411 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2413 When the machine has more than 32 registers, an integer does not suffice.
2414 Then the integers are replaced by sub-initializers, braced groupings containing
2415 several integers. Each sub-initializer must be suitable as an initializer
2416 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2417 In this situation, the first integer in each sub-initializer corresponds to
2418 registers 0 through 31, the second integer to registers 32 through 63, and
2422 @defmac REGNO_REG_CLASS (@var{regno})
2423 A C expression whose value is a register class containing hard register
2424 @var{regno}. In general there is more than one such class; choose a class
2425 which is @dfn{minimal}, meaning that no smaller class also contains the
2429 @defmac BASE_REG_CLASS
2430 A macro whose definition is the name of the class to which a valid
2431 base register must belong. A base register is one used in an address
2432 which is the register value plus a displacement.
2435 @defmac MODE_BASE_REG_CLASS (@var{mode})
2436 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2437 the selection of a base register in a mode dependent manner. If
2438 @var{mode} is VOIDmode then it should return the same value as
2439 @code{BASE_REG_CLASS}.
2442 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2443 A C expression whose value is the register class to which a valid
2444 base register must belong in order to be used in a base plus index
2445 register address. You should define this macro if base plus index
2446 addresses have different requirements than other base register uses.
2449 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2450 A C expression whose value is the register class to which a valid
2451 base register must belong. @var{outer_code} and @var{index_code} define the
2452 context in which the base register occurs. @var{outer_code} is the code of
2453 the immediately enclosing expression (@code{MEM} for the top level of an
2454 address, @code{ADDRESS} for something that occurs in an
2455 @code{address_operand}). @var{index_code} is the code of the corresponding
2456 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2459 @defmac INDEX_REG_CLASS
2460 A macro whose definition is the name of the class to which a valid
2461 index register must belong. An index register is one used in an
2462 address where its value is either multiplied by a scale factor or
2463 added to another register (as well as added to a displacement).
2466 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2467 A C expression which is nonzero if register number @var{num} is
2468 suitable for use as a base register in operand addresses.
2471 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2472 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2473 that expression may examine the mode of the memory reference in
2474 @var{mode}. You should define this macro if the mode of the memory
2475 reference affects whether a register may be used as a base register. If
2476 you define this macro, the compiler will use it instead of
2477 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2478 addresses that appear outside a @code{MEM}, i.e., as an
2479 @code{address_operand}.
2482 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2483 A C expression which is nonzero if register number @var{num} is suitable for
2484 use as a base register in base plus index operand addresses, accessing
2485 memory in mode @var{mode}. It may be either a suitable hard register or a
2486 pseudo register that has been allocated such a hard register. You should
2487 define this macro if base plus index addresses have different requirements
2488 than other base register uses.
2490 Use of this macro is deprecated; please use the more general
2491 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2494 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2495 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2496 that that expression may examine the context in which the register
2497 appears in the memory reference. @var{outer_code} is the code of the
2498 immediately enclosing expression (@code{MEM} if at the top level of the
2499 address, @code{ADDRESS} for something that occurs in an
2500 @code{address_operand}). @var{index_code} is the code of the
2501 corresponding index expression if @var{outer_code} is @code{PLUS};
2502 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2503 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2506 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2507 A C expression which is nonzero if register number @var{num} is
2508 suitable for use as an index register in operand addresses. It may be
2509 either a suitable hard register or a pseudo register that has been
2510 allocated such a hard register.
2512 The difference between an index register and a base register is that
2513 the index register may be scaled. If an address involves the sum of
2514 two registers, neither one of them scaled, then either one may be
2515 labeled the ``base'' and the other the ``index''; but whichever
2516 labeling is used must fit the machine's constraints of which registers
2517 may serve in each capacity. The compiler will try both labelings,
2518 looking for one that is valid, and will reload one or both registers
2519 only if neither labeling works.
2522 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2523 A target hook that places additional preference on the register class to use when it is necessary to rename a register in class @var{rclass} to another class, or perhaps @var{NO_REGS}, if no preferred register class is found or hook @code{preferred_rename_class} is not implemented. Sometimes returning a more restrictive class makes better code. For example, on ARM, thumb-2 instructions using @code{LO_REGS} may be smaller than instructions using @code{GENERIC_REGS}. By returning @code{LO_REGS} from @code{preferred_rename_class}, code size can be reduced.
2526 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2527 A target hook that places additional restrictions on the register class
2528 to use when it is necessary to copy value @var{x} into a register in class
2529 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2530 another, smaller class.
2532 The default version of this hook always returns value of @code{rclass} argument.
2534 Sometimes returning a more restrictive class makes better code. For
2535 example, on the 68000, when @var{x} is an integer constant that is in range
2536 for a @samp{moveq} instruction, the value of this macro is always
2537 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2538 Requiring a data register guarantees that a @samp{moveq} will be used.
2540 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2541 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2542 loaded into some register class. By returning @code{NO_REGS} you can
2543 force @var{x} into a memory location. For example, rs6000 can load
2544 immediate values into general-purpose registers, but does not have an
2545 instruction for loading an immediate value into a floating-point
2546 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2547 @var{x} is a floating-point constant. If the constant can't be loaded
2548 into any kind of register, code generation will be better if
2549 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2550 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2552 If an insn has pseudos in it after register allocation, reload will go
2553 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2554 to find the best one. Returning @code{NO_REGS}, in this case, makes
2555 reload add a @code{!} in front of the constraint: the x86 back-end uses
2556 this feature to discourage usage of 387 registers when math is done in
2557 the SSE registers (and vice versa).
2560 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2561 A C expression that places additional restrictions on the register class
2562 to use when it is necessary to copy value @var{x} into a register in class
2563 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2564 another, smaller class. On many machines, the following definition is
2568 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2571 Sometimes returning a more restrictive class makes better code. For
2572 example, on the 68000, when @var{x} is an integer constant that is in range
2573 for a @samp{moveq} instruction, the value of this macro is always
2574 @code{DATA_REGS} as long as @var{class} includes the data registers.
2575 Requiring a data register guarantees that a @samp{moveq} will be used.
2577 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2578 @var{class} is if @var{x} is a legitimate constant which cannot be
2579 loaded into some register class. By returning @code{NO_REGS} you can
2580 force @var{x} into a memory location. For example, rs6000 can load
2581 immediate values into general-purpose registers, but does not have an
2582 instruction for loading an immediate value into a floating-point
2583 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2584 @var{x} is a floating-point constant. If the constant can't be loaded
2585 into any kind of register, code generation will be better if
2586 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2587 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2589 If an insn has pseudos in it after register allocation, reload will go
2590 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2591 to find the best one. Returning @code{NO_REGS}, in this case, makes
2592 reload add a @code{!} in front of the constraint: the x86 back-end uses
2593 this feature to discourage usage of 387 registers when math is done in
2594 the SSE registers (and vice versa).
2597 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2598 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2599 input reloads. If you don't define this macro, the default is to use
2600 @var{class}, unchanged.
2602 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2603 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2606 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2607 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2610 The default version of this hook always returns value of @code{rclass}
2613 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2614 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2617 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2618 A C expression that places additional restrictions on the register class
2619 to use when it is necessary to be able to hold a value of mode
2620 @var{mode} in a reload register for which class @var{class} would
2623 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2624 there are certain modes that simply can't go in certain reload classes.
2626 The value is a register class; perhaps @var{class}, or perhaps another,
2629 Don't define this macro unless the target machine has limitations which
2630 require the macro to do something nontrivial.
2633 @deftypefn {Target Hook} reg_class_t TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, reg_class_t @var{reload_class}, enum machine_mode @var{reload_mode}, secondary_reload_info *@var{sri})
2634 Many machines have some registers that cannot be copied directly to or
2635 from memory or even from other types of registers. An example is the
2636 @samp{MQ} register, which on most machines, can only be copied to or
2637 from general registers, but not memory. Below, we shall be using the
2638 term 'intermediate register' when a move operation cannot be performed
2639 directly, but has to be done by copying the source into the intermediate
2640 register first, and then copying the intermediate register to the
2641 destination. An intermediate register always has the same mode as
2642 source and destination. Since it holds the actual value being copied,
2643 reload might apply optimizations to re-use an intermediate register
2644 and eliding the copy from the source when it can determine that the
2645 intermediate register still holds the required value.
2647 Another kind of secondary reload is required on some machines which
2648 allow copying all registers to and from memory, but require a scratch
2649 register for stores to some memory locations (e.g., those with symbolic
2650 address on the RT, and those with certain symbolic address on the SPARC
2651 when compiling PIC)@. Scratch registers need not have the same mode
2652 as the value being copied, and usually hold a different value than
2653 that being copied. Special patterns in the md file are needed to
2654 describe how the copy is performed with the help of the scratch register;
2655 these patterns also describe the number, register class(es) and mode(s)
2656 of the scratch register(s).
2658 In some cases, both an intermediate and a scratch register are required.
2660 For input reloads, this target hook is called with nonzero @var{in_p},
2661 and @var{x} is an rtx that needs to be copied to a register of class
2662 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2663 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2664 needs to be copied to rtx @var{x} in @var{reload_mode}.
2666 If copying a register of @var{reload_class} from/to @var{x} requires
2667 an intermediate register, the hook @code{secondary_reload} should
2668 return the register class required for this intermediate register.
2669 If no intermediate register is required, it should return NO_REGS.
2670 If more than one intermediate register is required, describe the one
2671 that is closest in the copy chain to the reload register.
2673 If scratch registers are needed, you also have to describe how to
2674 perform the copy from/to the reload register to/from this
2675 closest intermediate register. Or if no intermediate register is
2676 required, but still a scratch register is needed, describe the
2677 copy from/to the reload register to/from the reload operand @var{x}.
2679 You do this by setting @code{sri->icode} to the instruction code of a pattern
2680 in the md file which performs the move. Operands 0 and 1 are the output
2681 and input of this copy, respectively. Operands from operand 2 onward are
2682 for scratch operands. These scratch operands must have a mode, and a
2683 single-register-class
2684 @c [later: or memory]
2687 When an intermediate register is used, the @code{secondary_reload}
2688 hook will be called again to determine how to copy the intermediate
2689 register to/from the reload operand @var{x}, so your hook must also
2690 have code to handle the register class of the intermediate operand.
2692 @c [For later: maybe we'll allow multi-alternative reload patterns -
2693 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2694 @c and match the constraints of input and output to determine the required
2695 @c alternative. A restriction would be that constraints used to match
2696 @c against reloads registers would have to be written as register class
2697 @c constraints, or we need a new target macro / hook that tells us if an
2698 @c arbitrary constraint can match an unknown register of a given class.
2699 @c Such a macro / hook would also be useful in other places.]
2702 @var{x} might be a pseudo-register or a @code{subreg} of a
2703 pseudo-register, which could either be in a hard register or in memory.
2704 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2705 in memory and the hard register number if it is in a register.
2707 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2708 currently not supported. For the time being, you will have to continue
2709 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2711 @code{copy_cost} also uses this target hook to find out how values are
2712 copied. If you want it to include some extra cost for the need to allocate
2713 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2714 Or if two dependent moves are supposed to have a lower cost than the sum
2715 of the individual moves due to expected fortuitous scheduling and/or special
2716 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2719 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2720 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2721 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2722 These macros are obsolete, new ports should use the target hook
2723 @code{TARGET_SECONDARY_RELOAD} instead.
2725 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2726 target hook. Older ports still define these macros to indicate to the
2727 reload phase that it may
2728 need to allocate at least one register for a reload in addition to the
2729 register to contain the data. Specifically, if copying @var{x} to a
2730 register @var{class} in @var{mode} requires an intermediate register,
2731 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2732 largest register class all of whose registers can be used as
2733 intermediate registers or scratch registers.
2735 If copying a register @var{class} in @var{mode} to @var{x} requires an
2736 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2737 was supposed to be defined be defined to return the largest register
2738 class required. If the
2739 requirements for input and output reloads were the same, the macro
2740 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2743 The values returned by these macros are often @code{GENERAL_REGS}.
2744 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2745 can be directly copied to or from a register of @var{class} in
2746 @var{mode} without requiring a scratch register. Do not define this
2747 macro if it would always return @code{NO_REGS}.
2749 If a scratch register is required (either with or without an
2750 intermediate register), you were supposed to define patterns for
2751 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2752 (@pxref{Standard Names}. These patterns, which were normally
2753 implemented with a @code{define_expand}, should be similar to the
2754 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2757 These patterns need constraints for the reload register and scratch
2759 contain a single register class. If the original reload register (whose
2760 class is @var{class}) can meet the constraint given in the pattern, the
2761 value returned by these macros is used for the class of the scratch
2762 register. Otherwise, two additional reload registers are required.
2763 Their classes are obtained from the constraints in the insn pattern.
2765 @var{x} might be a pseudo-register or a @code{subreg} of a
2766 pseudo-register, which could either be in a hard register or in memory.
2767 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2768 in memory and the hard register number if it is in a register.
2770 These macros should not be used in the case where a particular class of
2771 registers can only be copied to memory and not to another class of
2772 registers. In that case, secondary reload registers are not needed and
2773 would not be helpful. Instead, a stack location must be used to perform
2774 the copy and the @code{mov@var{m}} pattern should use memory as an
2775 intermediate storage. This case often occurs between floating-point and
2779 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2780 Certain machines have the property that some registers cannot be copied
2781 to some other registers without using memory. Define this macro on
2782 those machines to be a C expression that is nonzero if objects of mode
2783 @var{m} in registers of @var{class1} can only be copied to registers of
2784 class @var{class2} by storing a register of @var{class1} into memory
2785 and loading that memory location into a register of @var{class2}.
2787 Do not define this macro if its value would always be zero.
2790 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2791 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2792 allocates a stack slot for a memory location needed for register copies.
2793 If this macro is defined, the compiler instead uses the memory location
2794 defined by this macro.
2796 Do not define this macro if you do not define
2797 @code{SECONDARY_MEMORY_NEEDED}.
2800 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2801 When the compiler needs a secondary memory location to copy between two
2802 registers of mode @var{mode}, it normally allocates sufficient memory to
2803 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2804 load operations in a mode that many bits wide and whose class is the
2805 same as that of @var{mode}.
2807 This is right thing to do on most machines because it ensures that all
2808 bits of the register are copied and prevents accesses to the registers
2809 in a narrower mode, which some machines prohibit for floating-point
2812 However, this default behavior is not correct on some machines, such as
2813 the DEC Alpha, that store short integers in floating-point registers
2814 differently than in integer registers. On those machines, the default
2815 widening will not work correctly and you must define this macro to
2816 suppress that widening in some cases. See the file @file{alpha.h} for
2819 Do not define this macro if you do not define
2820 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2821 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2824 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2825 A target hook which returns @code{true} if pseudos that have been assigned
2826 to registers of class @var{rclass} would likely be spilled because
2827 registers of @var{rclass} are needed for spill registers.
2829 The default version of this target hook returns @code{true} if @var{rclass}
2830 has exactly one register and @code{false} otherwise. On most machines, this
2831 default should be used. Only use this target hook to some other expression
2832 if pseudos allocated by @file{local-alloc.c} end up in memory because their
2833 hard registers were needed for spill registers. If this target hook returns
2834 @code{false} for those classes, those pseudos will only be allocated by
2835 @file{global.c}, which knows how to reallocate the pseudo to another
2836 register. If there would not be another register available for reallocation,
2837 you should not change the implementation of this target hook since
2838 the only effect of such implementation would be to slow down register
2842 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2843 A C expression for the maximum number of consecutive registers
2844 of class @var{class} needed to hold a value of mode @var{mode}.
2846 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2847 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2848 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2849 @var{mode})} for all @var{regno} values in the class @var{class}.
2851 This macro helps control the handling of multiple-word values
2855 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2856 If defined, a C expression that returns nonzero for a @var{class} for which
2857 a change from mode @var{from} to mode @var{to} is invalid.
2859 For the example, loading 32-bit integer or floating-point objects into
2860 floating-point registers on the Alpha extends them to 64 bits.
2861 Therefore loading a 64-bit object and then storing it as a 32-bit object
2862 does not store the low-order 32 bits, as would be the case for a normal
2863 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2867 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2868 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2869 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2873 @node Old Constraints
2874 @section Obsolete Macros for Defining Constraints
2875 @cindex defining constraints, obsolete method
2876 @cindex constraints, defining, obsolete method
2878 Machine-specific constraints can be defined with these macros instead
2879 of the machine description constructs described in @ref{Define
2880 Constraints}. This mechanism is obsolete. New ports should not use
2881 it; old ports should convert to the new mechanism.
2883 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2884 For the constraint at the start of @var{str}, which starts with the letter
2885 @var{c}, return the length. This allows you to have register class /
2886 constant / extra constraints that are longer than a single letter;
2887 you don't need to define this macro if you can do with single-letter
2888 constraints only. The definition of this macro should use
2889 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2890 to handle specially.
2891 There are some sanity checks in genoutput.c that check the constraint lengths
2892 for the md file, so you can also use this macro to help you while you are
2893 transitioning from a byzantine single-letter-constraint scheme: when you
2894 return a negative length for a constraint you want to re-use, genoutput
2895 will complain about every instance where it is used in the md file.
2898 @defmac REG_CLASS_FROM_LETTER (@var{char})
2899 A C expression which defines the machine-dependent operand constraint
2900 letters for register classes. If @var{char} is such a letter, the
2901 value should be the register class corresponding to it. Otherwise,
2902 the value should be @code{NO_REGS}. The register letter @samp{r},
2903 corresponding to class @code{GENERAL_REGS}, will not be passed
2904 to this macro; you do not need to handle it.
2907 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2908 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2909 passed in @var{str}, so that you can use suffixes to distinguish between
2913 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2914 A C expression that defines the machine-dependent operand constraint
2915 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2916 particular ranges of integer values. If @var{c} is one of those
2917 letters, the expression should check that @var{value}, an integer, is in
2918 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2919 not one of those letters, the value should be 0 regardless of
2923 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2924 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2925 string passed in @var{str}, so that you can use suffixes to distinguish
2926 between different variants.
2929 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2930 A C expression that defines the machine-dependent operand constraint
2931 letters that specify particular ranges of @code{const_double} values
2932 (@samp{G} or @samp{H}).
2934 If @var{c} is one of those letters, the expression should check that
2935 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2936 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2937 letters, the value should be 0 regardless of @var{value}.
2939 @code{const_double} is used for all floating-point constants and for
2940 @code{DImode} fixed-point constants. A given letter can accept either
2941 or both kinds of values. It can use @code{GET_MODE} to distinguish
2942 between these kinds.
2945 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2946 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2947 string passed in @var{str}, so that you can use suffixes to distinguish
2948 between different variants.
2951 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2952 A C expression that defines the optional machine-dependent constraint
2953 letters that can be used to segregate specific types of operands, usually
2954 memory references, for the target machine. Any letter that is not
2955 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2956 @code{REG_CLASS_FROM_CONSTRAINT}
2957 may be used. Normally this macro will not be defined.
2959 If it is required for a particular target machine, it should return 1
2960 if @var{value} corresponds to the operand type represented by the
2961 constraint letter @var{c}. If @var{c} is not defined as an extra
2962 constraint, the value returned should be 0 regardless of @var{value}.
2964 For example, on the ROMP, load instructions cannot have their output
2965 in r0 if the memory reference contains a symbolic address. Constraint
2966 letter @samp{Q} is defined as representing a memory address that does
2967 @emph{not} contain a symbolic address. An alternative is specified with
2968 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2969 alternative specifies @samp{m} on the input and a register class that
2970 does not include r0 on the output.
2973 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2974 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2975 in @var{str}, so that you can use suffixes to distinguish between different
2979 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2980 A C expression that defines the optional machine-dependent constraint
2981 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2982 be treated like memory constraints by the reload pass.
2984 It should return 1 if the operand type represented by the constraint
2985 at the start of @var{str}, the first letter of which is the letter @var{c},
2986 comprises a subset of all memory references including
2987 all those whose address is simply a base register. This allows the reload
2988 pass to reload an operand, if it does not directly correspond to the operand
2989 type of @var{c}, by copying its address into a base register.
2991 For example, on the S/390, some instructions do not accept arbitrary
2992 memory references, but only those that do not make use of an index
2993 register. The constraint letter @samp{Q} is defined via
2994 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2995 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2996 a @samp{Q} constraint can handle any memory operand, because the
2997 reload pass knows it can be reloaded by copying the memory address
2998 into a base register if required. This is analogous to the way
2999 an @samp{o} constraint can handle any memory operand.
3002 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
3003 A C expression that defines the optional machine-dependent constraint
3004 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
3005 @code{EXTRA_CONSTRAINT_STR}, that should
3006 be treated like address constraints by the reload pass.
3008 It should return 1 if the operand type represented by the constraint
3009 at the start of @var{str}, which starts with the letter @var{c}, comprises
3010 a subset of all memory addresses including
3011 all those that consist of just a base register. This allows the reload
3012 pass to reload an operand, if it does not directly correspond to the operand
3013 type of @var{str}, by copying it into a base register.
3015 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3016 be used with the @code{address_operand} predicate. It is treated
3017 analogously to the @samp{p} constraint.
3020 @node Stack and Calling
3021 @section Stack Layout and Calling Conventions
3022 @cindex calling conventions
3024 @c prevent bad page break with this line
3025 This describes the stack layout and calling conventions.
3029 * Exception Handling::
3034 * Register Arguments::
3036 * Aggregate Return::
3041 * Stack Smashing Protection::
3045 @subsection Basic Stack Layout
3046 @cindex stack frame layout
3047 @cindex frame layout
3049 @c prevent bad page break with this line
3050 Here is the basic stack layout.
3052 @defmac STACK_GROWS_DOWNWARD
3053 Define this macro if pushing a word onto the stack moves the stack
3054 pointer to a smaller address.
3056 When we say, ``define this macro if @dots{}'', it means that the
3057 compiler checks this macro only with @code{#ifdef} so the precise
3058 definition used does not matter.
3061 @defmac STACK_PUSH_CODE
3062 This macro defines the operation used when something is pushed
3063 on the stack. In RTL, a push operation will be
3064 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3066 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3067 and @code{POST_INC}. Which of these is correct depends on
3068 the stack direction and on whether the stack pointer points
3069 to the last item on the stack or whether it points to the
3070 space for the next item on the stack.
3072 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3073 defined, which is almost always right, and @code{PRE_INC} otherwise,
3074 which is often wrong.
3077 @defmac FRAME_GROWS_DOWNWARD
3078 Define this macro to nonzero value if the addresses of local variable slots
3079 are at negative offsets from the frame pointer.
3082 @defmac ARGS_GROW_DOWNWARD
3083 Define this macro if successive arguments to a function occupy decreasing
3084 addresses on the stack.
3087 @defmac STARTING_FRAME_OFFSET
3088 Offset from the frame pointer to the first local variable slot to be allocated.
3090 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3091 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3092 Otherwise, it is found by adding the length of the first slot to the
3093 value @code{STARTING_FRAME_OFFSET}.
3094 @c i'm not sure if the above is still correct.. had to change it to get
3095 @c rid of an overfull. --mew 2feb93
3098 @defmac STACK_ALIGNMENT_NEEDED
3099 Define to zero to disable final alignment of the stack during reload.
3100 The nonzero default for this macro is suitable for most ports.
3102 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3103 is a register save block following the local block that doesn't require
3104 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3105 stack alignment and do it in the backend.
3108 @defmac STACK_POINTER_OFFSET
3109 Offset from the stack pointer register to the first location at which
3110 outgoing arguments are placed. If not specified, the default value of
3111 zero is used. This is the proper value for most machines.
3113 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3114 the first location at which outgoing arguments are placed.
3117 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3118 Offset from the argument pointer register to the first argument's
3119 address. On some machines it may depend on the data type of the
3122 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3123 the first argument's address.
3126 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3127 Offset from the stack pointer register to an item dynamically allocated
3128 on the stack, e.g., by @code{alloca}.
3130 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3131 length of the outgoing arguments. The default is correct for most
3132 machines. See @file{function.c} for details.
3135 @defmac INITIAL_FRAME_ADDRESS_RTX
3136 A C expression whose value is RTL representing the address of the initial
3137 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3138 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3139 default value will be used. Define this macro in order to make frame pointer
3140 elimination work in the presence of @code{__builtin_frame_address (count)} and
3141 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3144 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3145 A C expression whose value is RTL representing the address in a stack
3146 frame where the pointer to the caller's frame is stored. Assume that
3147 @var{frameaddr} is an RTL expression for the address of the stack frame
3150 If you don't define this macro, the default is to return the value
3151 of @var{frameaddr}---that is, the stack frame address is also the
3152 address of the stack word that points to the previous frame.
3155 @defmac SETUP_FRAME_ADDRESSES
3156 If defined, a C expression that produces the machine-specific code to
3157 setup the stack so that arbitrary frames can be accessed. For example,
3158 on the SPARC, we must flush all of the register windows to the stack
3159 before we can access arbitrary stack frames. You will seldom need to
3163 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3164 This target hook should return an rtx that is used to store
3165 the address of the current frame into the built in @code{setjmp} buffer.
3166 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3167 machines. One reason you may need to define this target hook is if
3168 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3171 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3172 A C expression whose value is RTL representing the value of the frame
3173 address for the current frame. @var{frameaddr} is the frame pointer
3174 of the current frame. This is used for __builtin_frame_address.
3175 You need only define this macro if the frame address is not the same
3176 as the frame pointer. Most machines do not need to define it.
3179 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3180 A C expression whose value is RTL representing the value of the return
3181 address for the frame @var{count} steps up from the current frame, after
3182 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3183 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3184 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3186 The value of the expression must always be the correct address when
3187 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3188 determine the return address of other frames.
3191 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3192 Define this if the return address of a particular stack frame is accessed
3193 from the frame pointer of the previous stack frame.
3196 @defmac INCOMING_RETURN_ADDR_RTX
3197 A C expression whose value is RTL representing the location of the
3198 incoming return address at the beginning of any function, before the
3199 prologue. This RTL is either a @code{REG}, indicating that the return
3200 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3203 You only need to define this macro if you want to support call frame
3204 debugging information like that provided by DWARF 2.
3206 If this RTL is a @code{REG}, you should also define
3207 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3210 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3211 A C expression whose value is an integer giving a DWARF 2 column
3212 number that may be used as an alternative return column. The column
3213 must not correspond to any gcc hard register (that is, it must not
3214 be in the range of @code{DWARF_FRAME_REGNUM}).
3216 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3217 general register, but an alternative column needs to be used for signal
3218 frames. Some targets have also used different frame return columns
3222 @defmac DWARF_ZERO_REG
3223 A C expression whose value is an integer giving a DWARF 2 register
3224 number that is considered to always have the value zero. This should
3225 only be defined if the target has an architected zero register, and
3226 someone decided it was a good idea to use that register number to
3227 terminate the stack backtrace. New ports should avoid this.
3230 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3231 This target hook allows the backend to emit frame-related insns that
3232 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3233 info engine will invoke it on insns of the form
3235 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3239 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3241 to let the backend emit the call frame instructions. @var{label} is
3242 the CFI label attached to the insn, @var{pattern} is the pattern of
3243 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3246 @defmac INCOMING_FRAME_SP_OFFSET
3247 A C expression whose value is an integer giving the offset, in bytes,
3248 from the value of the stack pointer register to the top of the stack
3249 frame at the beginning of any function, before the prologue. The top of
3250 the frame is defined to be the value of the stack pointer in the
3251 previous frame, just before the call instruction.
3253 You only need to define this macro if you want to support call frame
3254 debugging information like that provided by DWARF 2.
3257 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3258 A C expression whose value is an integer giving the offset, in bytes,
3259 from the argument pointer to the canonical frame address (cfa). The
3260 final value should coincide with that calculated by
3261 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3262 during virtual register instantiation.
3264 The default value for this macro is
3265 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3266 which is correct for most machines; in general, the arguments are found
3267 immediately before the stack frame. Note that this is not the case on
3268 some targets that save registers into the caller's frame, such as SPARC
3269 and rs6000, and so such targets need to define this macro.
3271 You only need to define this macro if the default is incorrect, and you
3272 want to support call frame debugging information like that provided by
3276 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3277 If defined, a C expression whose value is an integer giving the offset
3278 in bytes from the frame pointer to the canonical frame address (cfa).
3279 The final value should coincide with that calculated by
3280 @code{INCOMING_FRAME_SP_OFFSET}.
3282 Normally the CFA is calculated as an offset from the argument pointer,
3283 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3284 variable due to the ABI, this may not be possible. If this macro is
3285 defined, it implies that the virtual register instantiation should be
3286 based on the frame pointer instead of the argument pointer. Only one
3287 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3291 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3292 If defined, a C expression whose value is an integer giving the offset
3293 in bytes from the canonical frame address (cfa) to the frame base used
3294 in DWARF 2 debug information. The default is zero. A different value
3295 may reduce the size of debug information on some ports.
3298 @node Exception Handling
3299 @subsection Exception Handling Support
3300 @cindex exception handling
3302 @defmac EH_RETURN_DATA_REGNO (@var{N})
3303 A C expression whose value is the @var{N}th register number used for
3304 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3305 @var{N} registers are usable.
3307 The exception handling library routines communicate with the exception
3308 handlers via a set of agreed upon registers. Ideally these registers
3309 should be call-clobbered; it is possible to use call-saved registers,
3310 but may negatively impact code size. The target must support at least
3311 2 data registers, but should define 4 if there are enough free registers.
3313 You must define this macro if you want to support call frame exception
3314 handling like that provided by DWARF 2.
3317 @defmac EH_RETURN_STACKADJ_RTX
3318 A C expression whose value is RTL representing a location in which
3319 to store a stack adjustment to be applied before function return.
3320 This is used to unwind the stack to an exception handler's call frame.
3321 It will be assigned zero on code paths that return normally.
3323 Typically this is a call-clobbered hard register that is otherwise
3324 untouched by the epilogue, but could also be a stack slot.
3326 Do not define this macro if the stack pointer is saved and restored
3327 by the regular prolog and epilog code in the call frame itself; in
3328 this case, the exception handling library routines will update the
3329 stack location to be restored in place. Otherwise, you must define
3330 this macro if you want to support call frame exception handling like
3331 that provided by DWARF 2.
3334 @defmac EH_RETURN_HANDLER_RTX
3335 A C expression whose value is RTL representing a location in which
3336 to store the address of an exception handler to which we should
3337 return. It will not be assigned on code paths that return normally.
3339 Typically this is the location in the call frame at which the normal
3340 return address is stored. For targets that return by popping an
3341 address off the stack, this might be a memory address just below
3342 the @emph{target} call frame rather than inside the current call
3343 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3344 been assigned, so it may be used to calculate the location of the
3347 Some targets have more complex requirements than storing to an
3348 address calculable during initial code generation. In that case
3349 the @code{eh_return} instruction pattern should be used instead.
3351 If you want to support call frame exception handling, you must
3352 define either this macro or the @code{eh_return} instruction pattern.
3355 @defmac RETURN_ADDR_OFFSET
3356 If defined, an integer-valued C expression for which rtl will be generated
3357 to add it to the exception handler address before it is searched in the
3358 exception handling tables, and to subtract it again from the address before
3359 using it to return to the exception handler.
3362 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3363 This macro chooses the encoding of pointers embedded in the exception
3364 handling sections. If at all possible, this should be defined such
3365 that the exception handling section will not require dynamic relocations,
3366 and so may be read-only.
3368 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3369 @var{global} is true if the symbol may be affected by dynamic relocations.
3370 The macro should return a combination of the @code{DW_EH_PE_*} defines
3371 as found in @file{dwarf2.h}.
3373 If this macro is not defined, pointers will not be encoded but
3374 represented directly.
3377 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3378 This macro allows the target to emit whatever special magic is required
3379 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3380 Generic code takes care of pc-relative and indirect encodings; this must
3381 be defined if the target uses text-relative or data-relative encodings.
3383 This is a C statement that branches to @var{done} if the format was
3384 handled. @var{encoding} is the format chosen, @var{size} is the number
3385 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3389 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3390 This macro allows the target to add CPU and operating system specific
3391 code to the call-frame unwinder for use when there is no unwind data
3392 available. The most common reason to implement this macro is to unwind
3393 through signal frames.
3395 This macro is called from @code{uw_frame_state_for} in
3396 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3397 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3398 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3399 for the address of the code being executed and @code{context->cfa} for
3400 the stack pointer value. If the frame can be decoded, the register
3401 save addresses should be updated in @var{fs} and the macro should
3402 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3403 the macro should evaluate to @code{_URC_END_OF_STACK}.
3405 For proper signal handling in Java this macro is accompanied by
3406 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3409 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3410 This macro allows the target to add operating system specific code to the
3411 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3412 usually used for signal or interrupt frames.
3414 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3415 @var{context} is an @code{_Unwind_Context};
3416 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3417 for the abi and context in the @code{.unwabi} directive. If the
3418 @code{.unwabi} directive can be handled, the register save addresses should
3419 be updated in @var{fs}.
3422 @defmac TARGET_USES_WEAK_UNWIND_INFO
3423 A C expression that evaluates to true if the target requires unwind
3424 info to be given comdat linkage. Define it to be @code{1} if comdat
3425 linkage is necessary. The default is @code{0}.
3428 @node Stack Checking
3429 @subsection Specifying How Stack Checking is Done
3431 GCC will check that stack references are within the boundaries of the
3432 stack, if the option @option{-fstack-check} is specified, in one of
3437 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3438 will assume that you have arranged for full stack checking to be done
3439 at appropriate places in the configuration files. GCC will not do
3440 other special processing.
3443 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3444 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3445 that you have arranged for static stack checking (checking of the
3446 static stack frame of functions) to be done at appropriate places
3447 in the configuration files. GCC will only emit code to do dynamic
3448 stack checking (checking on dynamic stack allocations) using the third
3452 If neither of the above are true, GCC will generate code to periodically
3453 ``probe'' the stack pointer using the values of the macros defined below.
3456 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3457 GCC will change its allocation strategy for large objects if the option
3458 @option{-fstack-check} is specified: they will always be allocated
3459 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3461 @defmac STACK_CHECK_BUILTIN
3462 A nonzero value if stack checking is done by the configuration files in a
3463 machine-dependent manner. You should define this macro if stack checking
3464 is required by the ABI of your machine or if you would like to do stack
3465 checking in some more efficient way than the generic approach. The default
3466 value of this macro is zero.
3469 @defmac STACK_CHECK_STATIC_BUILTIN
3470 A nonzero value if static stack checking is done by the configuration files
3471 in a machine-dependent manner. You should define this macro if you would
3472 like to do static stack checking in some more efficient way than the generic
3473 approach. The default value of this macro is zero.
3476 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3477 An integer specifying the interval at which GCC must generate stack probe
3478 instructions, defined as 2 raised to this integer. You will normally
3479 define this macro so that the interval be no larger than the size of
3480 the ``guard pages'' at the end of a stack area. The default value
3481 of 12 (4096-byte interval) is suitable for most systems.
3484 @defmac STACK_CHECK_MOVING_SP
3485 An integer which is nonzero if GCC should move the stack pointer page by page
3486 when doing probes. This can be necessary on systems where the stack pointer
3487 contains the bottom address of the memory area accessible to the executing
3488 thread at any point in time. In this situation an alternate signal stack
3489 is required in order to be able to recover from a stack overflow. The
3490 default value of this macro is zero.
3493 @defmac STACK_CHECK_PROTECT
3494 The number of bytes of stack needed to recover from a stack overflow, for
3495 languages where such a recovery is supported. The default value of 75 words
3496 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3497 8192 bytes with other exception handling mechanisms should be adequate for
3501 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3502 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3503 in the opposite case.
3505 @defmac STACK_CHECK_MAX_FRAME_SIZE
3506 The maximum size of a stack frame, in bytes. GCC will generate probe
3507 instructions in non-leaf functions to ensure at least this many bytes of
3508 stack are available. If a stack frame is larger than this size, stack
3509 checking will not be reliable and GCC will issue a warning. The
3510 default is chosen so that GCC only generates one instruction on most
3511 systems. You should normally not change the default value of this macro.
3514 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3515 GCC uses this value to generate the above warning message. It
3516 represents the amount of fixed frame used by a function, not including
3517 space for any callee-saved registers, temporaries and user variables.
3518 You need only specify an upper bound for this amount and will normally
3519 use the default of four words.
3522 @defmac STACK_CHECK_MAX_VAR_SIZE
3523 The maximum size, in bytes, of an object that GCC will place in the
3524 fixed area of the stack frame when the user specifies
3525 @option{-fstack-check}.
3526 GCC computed the default from the values of the above macros and you will
3527 normally not need to override that default.
3531 @node Frame Registers
3532 @subsection Registers That Address the Stack Frame
3534 @c prevent bad page break with this line
3535 This discusses registers that address the stack frame.
3537 @defmac STACK_POINTER_REGNUM
3538 The register number of the stack pointer register, which must also be a
3539 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3540 the hardware determines which register this is.
3543 @defmac FRAME_POINTER_REGNUM
3544 The register number of the frame pointer register, which is used to
3545 access automatic variables in the stack frame. On some machines, the
3546 hardware determines which register this is. On other machines, you can
3547 choose any register you wish for this purpose.
3550 @defmac HARD_FRAME_POINTER_REGNUM
3551 On some machines the offset between the frame pointer and starting
3552 offset of the automatic variables is not known until after register
3553 allocation has been done (for example, because the saved registers are
3554 between these two locations). On those machines, define
3555 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3556 be used internally until the offset is known, and define
3557 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3558 used for the frame pointer.
3560 You should define this macro only in the very rare circumstances when it
3561 is not possible to calculate the offset between the frame pointer and
3562 the automatic variables until after register allocation has been
3563 completed. When this macro is defined, you must also indicate in your
3564 definition of @code{ELIMINABLE_REGS} how to eliminate
3565 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3566 or @code{STACK_POINTER_REGNUM}.
3568 Do not define this macro if it would be the same as
3569 @code{FRAME_POINTER_REGNUM}.
3572 @defmac ARG_POINTER_REGNUM
3573 The register number of the arg pointer register, which is used to access
3574 the function's argument list. On some machines, this is the same as the
3575 frame pointer register. On some machines, the hardware determines which
3576 register this is. On other machines, you can choose any register you
3577 wish for this purpose. If this is not the same register as the frame
3578 pointer register, then you must mark it as a fixed register according to
3579 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3580 (@pxref{Elimination}).
3583 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3584 Define this to a preprocessor constant that is nonzero if
3585 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3586 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3587 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3588 definition is not suitable for use in preprocessor conditionals.
3591 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3592 Define this to a preprocessor constant that is nonzero if
3593 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3594 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3595 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3596 definition is not suitable for use in preprocessor conditionals.
3599 @defmac RETURN_ADDRESS_POINTER_REGNUM
3600 The register number of the return address pointer register, which is used to
3601 access the current function's return address from the stack. On some
3602 machines, the return address is not at a fixed offset from the frame
3603 pointer or stack pointer or argument pointer. This register can be defined
3604 to point to the return address on the stack, and then be converted by
3605 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3607 Do not define this macro unless there is no other way to get the return
3608 address from the stack.
3611 @defmac STATIC_CHAIN_REGNUM
3612 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3613 Register numbers used for passing a function's static chain pointer. If
3614 register windows are used, the register number as seen by the called
3615 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3616 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3617 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3620 The static chain register need not be a fixed register.
3622 If the static chain is passed in memory, these macros should not be
3623 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3626 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl}, bool @var{incoming_p})
3627 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3628 targets that may use different static chain locations for different
3629 nested functions. This may be required if the target has function
3630 attributes that affect the calling conventions of the function and
3631 those calling conventions use different static chain locations.
3633 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3635 If the static chain is passed in memory, this hook should be used to
3636 provide rtx giving @code{mem} expressions that denote where they are stored.
3637 Often the @code{mem} expression as seen by the caller will be at an offset
3638 from the stack pointer and the @code{mem} expression as seen by the callee
3639 will be at an offset from the frame pointer.
3640 @findex stack_pointer_rtx
3641 @findex frame_pointer_rtx
3642 @findex arg_pointer_rtx
3643 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3644 @code{arg_pointer_rtx} will have been initialized and should be used
3645 to refer to those items.
3648 @defmac DWARF_FRAME_REGISTERS
3649 This macro specifies the maximum number of hard registers that can be
3650 saved in a call frame. This is used to size data structures used in
3651 DWARF2 exception handling.
3653 Prior to GCC 3.0, this macro was needed in order to establish a stable
3654 exception handling ABI in the face of adding new hard registers for ISA
3655 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3656 in the number of hard registers. Nevertheless, this macro can still be
3657 used to reduce the runtime memory requirements of the exception handling
3658 routines, which can be substantial if the ISA contains a lot of
3659 registers that are not call-saved.
3661 If this macro is not defined, it defaults to
3662 @code{FIRST_PSEUDO_REGISTER}.
3665 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3667 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3668 for backward compatibility in pre GCC 3.0 compiled code.
3670 If this macro is not defined, it defaults to
3671 @code{DWARF_FRAME_REGISTERS}.
3674 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3676 Define this macro if the target's representation for dwarf registers
3677 is different than the internal representation for unwind column.
3678 Given a dwarf register, this macro should return the internal unwind
3679 column number to use instead.
3681 See the PowerPC's SPE target for an example.
3684 @defmac DWARF_FRAME_REGNUM (@var{regno})
3686 Define this macro if the target's representation for dwarf registers
3687 used in .eh_frame or .debug_frame is different from that used in other
3688 debug info sections. Given a GCC hard register number, this macro
3689 should return the .eh_frame register number. The default is
3690 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3694 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3696 Define this macro to map register numbers held in the call frame info
3697 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3698 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3699 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3700 return @code{@var{regno}}.
3705 @subsection Eliminating Frame Pointer and Arg Pointer
3707 @c prevent bad page break with this line
3708 This is about eliminating the frame pointer and arg pointer.
3710 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3711 This target hook should return @code{true} if a function must have and use
3712 a frame pointer. This target hook is called in the reload pass. If its return
3713 value is @code{true} the function will have a frame pointer.
3715 This target hook can in principle examine the current function and decide
3716 according to the facts, but on most machines the constant @code{false} or the
3717 constant @code{true} suffices. Use @code{false} when the machine allows code
3718 to be generated with no frame pointer, and doing so saves some time or space.
3719 Use @code{true} when there is no possible advantage to avoiding a frame
3722 In certain cases, the compiler does not know how to produce valid code
3723 without a frame pointer. The compiler recognizes those cases and
3724 automatically gives the function a frame pointer regardless of what
3725 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3728 In a function that does not require a frame pointer, the frame pointer
3729 register can be allocated for ordinary usage, unless you mark it as a
3730 fixed register. See @code{FIXED_REGISTERS} for more information.
3732 Default return value is @code{false}.
3735 @findex get_frame_size
3736 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3737 A C statement to store in the variable @var{depth-var} the difference
3738 between the frame pointer and the stack pointer values immediately after
3739 the function prologue. The value would be computed from information
3740 such as the result of @code{get_frame_size ()} and the tables of
3741 registers @code{regs_ever_live} and @code{call_used_regs}.
3743 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3744 need not be defined. Otherwise, it must be defined even if
3745 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3746 case, you may set @var{depth-var} to anything.
3749 @defmac ELIMINABLE_REGS
3750 If defined, this macro specifies a table of register pairs used to
3751 eliminate unneeded registers that point into the stack frame. If it is not
3752 defined, the only elimination attempted by the compiler is to replace
3753 references to the frame pointer with references to the stack pointer.
3755 The definition of this macro is a list of structure initializations, each
3756 of which specifies an original and replacement register.
3758 On some machines, the position of the argument pointer is not known until
3759 the compilation is completed. In such a case, a separate hard register
3760 must be used for the argument pointer. This register can be eliminated by
3761 replacing it with either the frame pointer or the argument pointer,
3762 depending on whether or not the frame pointer has been eliminated.
3764 In this case, you might specify:
3766 #define ELIMINABLE_REGS \
3767 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3768 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3769 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3772 Note that the elimination of the argument pointer with the stack pointer is
3773 specified first since that is the preferred elimination.
3776 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3777 This target hook should returns @code{true} if the compiler is allowed to
3778 try to replace register number @var{from_reg} with register number
3779 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3780 is defined, and will usually be @code{true}, since most of the cases
3781 preventing register elimination are things that the compiler already
3784 Default return value is @code{true}.
3787 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3788 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3789 specifies the initial difference between the specified pair of
3790 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3794 @node Stack Arguments
3795 @subsection Passing Function Arguments on the Stack
3796 @cindex arguments on stack
3797 @cindex stack arguments
3799 The macros in this section control how arguments are passed
3800 on the stack. See the following section for other macros that
3801 control passing certain arguments in registers.
3803 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3804 This target hook returns @code{true} if an argument declared in a
3805 prototype as an integral type smaller than @code{int} should actually be
3806 passed as an @code{int}. In addition to avoiding errors in certain
3807 cases of mismatch, it also makes for better code on certain machines.
3808 The default is to not promote prototypes.
3812 A C expression. If nonzero, push insns will be used to pass
3814 If the target machine does not have a push instruction, set it to zero.
3815 That directs GCC to use an alternate strategy: to
3816 allocate the entire argument block and then store the arguments into
3817 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3820 @defmac PUSH_ARGS_REVERSED
3821 A C expression. If nonzero, function arguments will be evaluated from
3822 last to first, rather than from first to last. If this macro is not
3823 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3824 and args grow in opposite directions, and 0 otherwise.
3827 @defmac PUSH_ROUNDING (@var{npushed})
3828 A C expression that is the number of bytes actually pushed onto the
3829 stack when an instruction attempts to push @var{npushed} bytes.
3831 On some machines, the definition
3834 #define PUSH_ROUNDING(BYTES) (BYTES)
3838 will suffice. But on other machines, instructions that appear
3839 to push one byte actually push two bytes in an attempt to maintain
3840 alignment. Then the definition should be
3843 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3846 If the value of this macro has a type, it should be an unsigned type.
3849 @findex current_function_outgoing_args_size
3850 @defmac ACCUMULATE_OUTGOING_ARGS
3851 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3852 will be computed and placed into the variable
3853 @code{current_function_outgoing_args_size}. No space will be pushed
3854 onto the stack for each call; instead, the function prologue should
3855 increase the stack frame size by this amount.
3857 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3861 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3862 Define this macro if functions should assume that stack space has been
3863 allocated for arguments even when their values are passed in
3866 The value of this macro is the size, in bytes, of the area reserved for
3867 arguments passed in registers for the function represented by @var{fndecl},
3868 which can be zero if GCC is calling a library function.
3869 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3872 This space can be allocated by the caller, or be a part of the
3873 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3876 @c above is overfull. not sure what to do. --mew 5feb93 did
3877 @c something, not sure if it looks good. --mew 10feb93
3879 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3880 Define this to a nonzero value if it is the responsibility of the
3881 caller to allocate the area reserved for arguments passed in registers
3882 when calling a function of @var{fntype}. @var{fntype} may be NULL
3883 if the function called is a library function.
3885 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3886 whether the space for these arguments counts in the value of
3887 @code{current_function_outgoing_args_size}.
3890 @defmac STACK_PARMS_IN_REG_PARM_AREA
3891 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3892 stack parameters don't skip the area specified by it.
3893 @c i changed this, makes more sens and it should have taken care of the
3894 @c overfull.. not as specific, tho. --mew 5feb93
3896 Normally, when a parameter is not passed in registers, it is placed on the
3897 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3898 suppresses this behavior and causes the parameter to be passed on the
3899 stack in its natural location.
3902 @deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size})
3903 This target hook returns the number of bytes of its own arguments that
3904 a function pops on returning, or 0 if the function pops no arguments
3905 and the caller must therefore pop them all after the function returns.
3907 @var{fundecl} is a C variable whose value is a tree node that describes
3908 the function in question. Normally it is a node of type
3909 @code{FUNCTION_DECL} that describes the declaration of the function.
3910 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3912 @var{funtype} is a C variable whose value is a tree node that
3913 describes the function in question. Normally it is a node of type
3914 @code{FUNCTION_TYPE} that describes the data type of the function.
3915 From this it is possible to obtain the data types of the value and
3916 arguments (if known).
3918 When a call to a library function is being considered, @var{fundecl}
3919 will contain an identifier node for the library function. Thus, if
3920 you need to distinguish among various library functions, you can do so
3921 by their names. Note that ``library function'' in this context means
3922 a function used to perform arithmetic, whose name is known specially
3923 in the compiler and was not mentioned in the C code being compiled.
3925 @var{size} is the number of bytes of arguments passed on the
3926 stack. If a variable number of bytes is passed, it is zero, and
3927 argument popping will always be the responsibility of the calling function.
3929 On the VAX, all functions always pop their arguments, so the definition
3930 of this macro is @var{size}. On the 68000, using the standard
3931 calling convention, no functions pop their arguments, so the value of
3932 the macro is always 0 in this case. But an alternative calling
3933 convention is available in which functions that take a fixed number of
3934 arguments pop them but other functions (such as @code{printf}) pop
3935 nothing (the caller pops all). When this convention is in use,
3936 @var{funtype} is examined to determine whether a function takes a fixed
3937 number of arguments.
3940 @defmac CALL_POPS_ARGS (@var{cum})
3941 A C expression that should indicate the number of bytes a call sequence
3942 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3943 when compiling a function call.
3945 @var{cum} is the variable in which all arguments to the called function
3946 have been accumulated.
3948 On certain architectures, such as the SH5, a call trampoline is used
3949 that pops certain registers off the stack, depending on the arguments
3950 that have been passed to the function. Since this is a property of the
3951 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3955 @node Register Arguments
3956 @subsection Passing Arguments in Registers
3957 @cindex arguments in registers
3958 @cindex registers arguments
3960 This section describes the macros which let you control how various
3961 types of arguments are passed in registers or how they are arranged in
3964 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3965 Return an RTX indicating whether a function argument is passed in a
3966 register and if so, which register.
3968 The arguments are @var{ca}, which summarizes all the previous
3969 arguments; @var{mode}, the machine mode of the argument; @var{type},
3970 the data type of the argument as a tree node or 0 if that is not known
3971 (which happens for C support library functions); and @var{named},
3972 which is @code{true} for an ordinary argument and @code{false} for
3973 nameless arguments that correspond to @samp{@dots{}} in the called
3974 function's prototype. @var{type} can be an incomplete type if a
3975 syntax error has previously occurred.
3977 The return value is usually either a @code{reg} RTX for the hard
3978 register in which to pass the argument, or zero to pass the argument
3981 The value of the expression can also be a @code{parallel} RTX@. This is
3982 used when an argument is passed in multiple locations. The mode of the
3983 @code{parallel} should be the mode of the entire argument. The
3984 @code{parallel} holds any number of @code{expr_list} pairs; each one
3985 describes where part of the argument is passed. In each
3986 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3987 register in which to pass this part of the argument, and the mode of the
3988 register RTX indicates how large this part of the argument is. The
3989 second operand of the @code{expr_list} is a @code{const_int} which gives
3990 the offset in bytes into the entire argument of where this part starts.
3991 As a special exception the first @code{expr_list} in the @code{parallel}
3992 RTX may have a first operand of zero. This indicates that the entire
3993 argument is also stored on the stack.
3995 The last time this hook is called, it is called with @code{MODE ==
3996 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3997 pattern as operands 2 and 3 respectively.
3999 @cindex @file{stdarg.h} and register arguments
4000 The usual way to make the ISO library @file{stdarg.h} work on a
4001 machine where some arguments are usually passed in registers, is to
4002 cause nameless arguments to be passed on the stack instead. This is
4003 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
4004 @var{named} is @code{false}.
4006 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
4007 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
4008 You may use the hook @code{targetm.calls.must_pass_in_stack}
4009 in the definition of this macro to determine if this argument is of a
4010 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4011 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
4012 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4013 defined, the argument will be computed in the stack and then loaded into
4017 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, const_tree @var{type})
4018 This target hook should return @code{true} if we should not pass @var{type}
4019 solely in registers. The file @file{expr.h} defines a
4020 definition that is usually appropriate, refer to @file{expr.h} for additional
4024 @deftypefn {Target Hook} rtx TARGET_FUNCTION_INCOMING_ARG (cumulative_args_t @var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4025 Define this hook if the target machine has ``register windows'', so
4026 that the register in which a function sees an arguments is not
4027 necessarily the same as the one in which the caller passed the
4030 For such machines, @code{TARGET_FUNCTION_ARG} computes the register in
4031 which the caller passes the value, and
4032 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4033 fashion to tell the function being called where the arguments will
4036 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4037 @code{TARGET_FUNCTION_ARG} serves both purposes.
4040 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4041 This target hook returns the number of bytes at the beginning of an
4042 argument that must be put in registers. The value must be zero for
4043 arguments that are passed entirely in registers or that are entirely
4044 pushed on the stack.
4046 On some machines, certain arguments must be passed partially in
4047 registers and partially in memory. On these machines, typically the
4048 first few words of arguments are passed in registers, and the rest
4049 on the stack. If a multi-word argument (a @code{double} or a
4050 structure) crosses that boundary, its first few words must be passed
4051 in registers and the rest must be pushed. This macro tells the
4052 compiler when this occurs, and how many bytes should go in registers.
4054 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4055 register to be used by the caller for this argument; likewise
4056 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4059 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4060 This target hook should return @code{true} if an argument at the
4061 position indicated by @var{cum} should be passed by reference. This
4062 predicate is queried after target independent reasons for being
4063 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4065 If the hook returns true, a copy of that argument is made in memory and a
4066 pointer to the argument is passed instead of the argument itself.
4067 The pointer is passed in whatever way is appropriate for passing a pointer
4071 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4072 The function argument described by the parameters to this hook is
4073 known to be passed by reference. The hook should return true if the
4074 function argument should be copied by the callee instead of copied
4077 For any argument for which the hook returns true, if it can be
4078 determined that the argument is not modified, then a copy need
4081 The default version of this hook always returns false.
4084 @defmac CUMULATIVE_ARGS
4085 A C type for declaring a variable that is used as the first argument
4086 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4087 target machines, the type @code{int} suffices and can hold the number
4088 of bytes of argument so far.
4090 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4091 arguments that have been passed on the stack. The compiler has other
4092 variables to keep track of that. For target machines on which all
4093 arguments are passed on the stack, there is no need to store anything in
4094 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4095 should not be empty, so use @code{int}.
4098 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4099 If defined, this macro is called before generating any code for a
4100 function, but after the @var{cfun} descriptor for the function has been
4101 created. The back end may use this macro to update @var{cfun} to
4102 reflect an ABI other than that which would normally be used by default.
4103 If the compiler is generating code for a compiler-generated function,
4104 @var{fndecl} may be @code{NULL}.
4107 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4108 A C statement (sans semicolon) for initializing the variable
4109 @var{cum} for the state at the beginning of the argument list. The
4110 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4111 is the tree node for the data type of the function which will receive
4112 the args, or 0 if the args are to a compiler support library function.
4113 For direct calls that are not libcalls, @var{fndecl} contain the
4114 declaration node of the function. @var{fndecl} is also set when
4115 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4116 being compiled. @var{n_named_args} is set to the number of named
4117 arguments, including a structure return address if it is passed as a
4118 parameter, when making a call. When processing incoming arguments,
4119 @var{n_named_args} is set to @minus{}1.
4121 When processing a call to a compiler support library function,
4122 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4123 contains the name of the function, as a string. @var{libname} is 0 when
4124 an ordinary C function call is being processed. Thus, each time this
4125 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4126 never both of them at once.
4129 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4130 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4131 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4132 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4133 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4134 0)} is used instead.
4137 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4138 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4139 finding the arguments for the function being compiled. If this macro is
4140 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4142 The value passed for @var{libname} is always 0, since library routines
4143 with special calling conventions are never compiled with GCC@. The
4144 argument @var{libname} exists for symmetry with
4145 @code{INIT_CUMULATIVE_ARGS}.
4146 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4147 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4150 @deftypefn {Target Hook} void TARGET_FUNCTION_ARG_ADVANCE (cumulative_args_t @var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4151 This hook updates the summarizer variable pointed to by @var{ca} to
4152 advance past an argument in the argument list. The values @var{mode},
4153 @var{type} and @var{named} describe that argument. Once this is done,
4154 the variable @var{cum} is suitable for analyzing the @emph{following}
4155 argument with @code{TARGET_FUNCTION_ARG}, etc.
4157 This hook need not do anything if the argument in question was passed
4158 on the stack. The compiler knows how to track the amount of stack space
4159 used for arguments without any special help.
4162 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4163 If defined, a C expression that is the number of bytes to add to the
4164 offset of the argument passed in memory. This is needed for the SPU,
4165 which passes @code{char} and @code{short} arguments in the preferred
4166 slot that is in the middle of the quad word instead of starting at the
4170 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4171 If defined, a C expression which determines whether, and in which direction,
4172 to pad out an argument with extra space. The value should be of type
4173 @code{enum direction}: either @code{upward} to pad above the argument,
4174 @code{downward} to pad below, or @code{none} to inhibit padding.
4176 The @emph{amount} of padding is always just enough to reach the next
4177 multiple of @code{TARGET_FUNCTION_ARG_BOUNDARY}; this macro does not
4180 This macro has a default definition which is right for most systems.
4181 For little-endian machines, the default is to pad upward. For
4182 big-endian machines, the default is to pad downward for an argument of
4183 constant size shorter than an @code{int}, and upward otherwise.
4186 @defmac PAD_VARARGS_DOWN
4187 If defined, a C expression which determines whether the default
4188 implementation of va_arg will attempt to pad down before reading the
4189 next argument, if that argument is smaller than its aligned space as
4190 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4191 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4194 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4195 Specify padding for the last element of a block move between registers and
4196 memory. @var{first} is nonzero if this is the only element. Defining this
4197 macro allows better control of register function parameters on big-endian
4198 machines, without using @code{PARALLEL} rtl. In particular,
4199 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4200 registers, as there is no longer a "wrong" part of a register; For example,
4201 a three byte aggregate may be passed in the high part of a register if so
4205 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4206 This hook returns the alignment boundary, in bits, of an argument
4207 with the specified mode and type. The default hook returns
4208 @code{PARM_BOUNDARY} for all arguments.
4211 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4212 A C expression that is nonzero if @var{regno} is the number of a hard
4213 register in which function arguments are sometimes passed. This does
4214 @emph{not} include implicit arguments such as the static chain and
4215 the structure-value address. On many machines, no registers can be
4216 used for this purpose since all function arguments are pushed on the
4220 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4221 This hook should return true if parameter of type @var{type} are passed
4222 as two scalar parameters. By default, GCC will attempt to pack complex
4223 arguments into the target's word size. Some ABIs require complex arguments
4224 to be split and treated as their individual components. For example, on
4225 AIX64, complex floats should be passed in a pair of floating point
4226 registers, even though a complex float would fit in one 64-bit floating
4229 The default value of this hook is @code{NULL}, which is treated as always
4233 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4234 This hook returns a type node for @code{va_list} for the target.
4235 The default version of the hook returns @code{void*}.
4238 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4239 This target hook is used in function @code{c_common_nodes_and_builtins}
4240 to iterate through the target specific builtin types for va_list. The
4241 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4242 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4244 The arguments @var{pname} and @var{ptree} are used to store the result of
4245 this macro and are set to the name of the va_list builtin type and its
4247 If the return value of this macro is zero, then there is no more element.
4248 Otherwise the @var{IDX} should be increased for the next call of this
4249 macro to iterate through all types.
4252 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4253 This hook returns the va_list type of the calling convention specified by
4255 The default version of this hook returns @code{va_list_type_node}.
4258 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4259 This hook returns the va_list type of the calling convention specified by the
4260 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4264 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, gimple_seq *@var{pre_p}, gimple_seq *@var{post_p})
4265 This hook performs target-specific gimplification of
4266 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4267 arguments to @code{va_arg}; the latter two are as in
4268 @code{gimplify.c:gimplify_expr}.
4271 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4272 Define this to return nonzero if the port can handle pointers
4273 with machine mode @var{mode}. The default version of this
4274 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4277 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (struct ao_ref_s *@var{ref})
4278 Define this to return nonzero if the memory reference @var{ref} may alias with the system C library errno location. The default version of this hook assumes the system C library errno location is either a declaration of type int or accessed by dereferencing a pointer to int.
4281 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4282 Define this to return nonzero if the port is prepared to handle
4283 insns involving scalar mode @var{mode}. For a scalar mode to be
4284 considered supported, all the basic arithmetic and comparisons
4287 The default version of this hook returns true for any mode
4288 required to handle the basic C types (as defined by the port).
4289 Included here are the double-word arithmetic supported by the
4290 code in @file{optabs.c}.
4293 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4294 Define this to return nonzero if the port is prepared to handle
4295 insns involving vector mode @var{mode}. At the very least, it
4296 must have move patterns for this mode.
4299 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (enum machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4300 Return true if GCC should try to use a scalar mode to store an array
4301 of @var{nelems} elements, given that each element has mode @var{mode}.
4302 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4303 and allows GCC to use any defined integer mode.
4305 One use of this hook is to support vector load and store operations
4306 that operate on several homogeneous vectors. For example, ARM NEON
4307 has operations like:
4310 int8x8x3_t vld3_s8 (const int8_t *)
4313 where the return type is defined as:
4316 typedef struct int8x8x3_t
4322 If this hook allows @code{val} to have a scalar mode, then
4323 @code{int8x8x3_t} can have the same mode. GCC can then store
4324 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4327 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (enum machine_mode @var{mode})
4328 Define this to return nonzero for machine modes for which the port has
4329 small register classes. If this target hook returns nonzero for a given
4330 @var{mode}, the compiler will try to minimize the lifetime of registers
4331 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4332 In this case, the hook is expected to return nonzero if it returns nonzero
4335 On some machines, it is risky to let hard registers live across arbitrary
4336 insns. Typically, these machines have instructions that require values
4337 to be in specific registers (like an accumulator), and reload will fail
4338 if the required hard register is used for another purpose across such an
4341 Passes before reload do not know which hard registers will be used
4342 in an instruction, but the machine modes of the registers set or used in
4343 the instruction are already known. And for some machines, register
4344 classes are small for, say, integer registers but not for floating point
4345 registers. For example, the AMD x86-64 architecture requires specific
4346 registers for the legacy x86 integer instructions, but there are many
4347 SSE registers for floating point operations. On such targets, a good
4348 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4349 machine modes but zero for the SSE register classes.
4351 The default version of this hook returns false for any mode. It is always
4352 safe to redefine this hook to return with a nonzero value. But if you
4353 unnecessarily define it, you will reduce the amount of optimizations
4354 that can be performed in some cases. If you do not define this hook
4355 to return a nonzero value when it is required, the compiler will run out
4356 of spill registers and print a fatal error message.
4359 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
4360 If the target has a dedicated flags register, and it needs to use the post-reload comparison elimination pass, then this value should be set appropriately.
4364 @subsection How Scalar Function Values Are Returned
4365 @cindex return values in registers
4366 @cindex values, returned by functions
4367 @cindex scalars, returned as values
4369 This section discusses the macros that control returning scalars as
4370 values---values that can fit in registers.
4372 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4374 Define this to return an RTX representing the place where a function
4375 returns or receives a value of data type @var{ret_type}, a tree node
4376 representing a data type. @var{fn_decl_or_type} is a tree node
4377 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4378 function being called. If @var{outgoing} is false, the hook should
4379 compute the register in which the caller will see the return value.
4380 Otherwise, the hook should return an RTX representing the place where
4381 a function returns a value.
4383 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4384 (Actually, on most machines, scalar values are returned in the same
4385 place regardless of mode.) The value of the expression is usually a
4386 @code{reg} RTX for the hard register where the return value is stored.
4387 The value can also be a @code{parallel} RTX, if the return value is in
4388 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4389 @code{parallel} form. Note that the callee will populate every
4390 location specified in the @code{parallel}, but if the first element of
4391 the @code{parallel} contains the whole return value, callers will use
4392 that element as the canonical location and ignore the others. The m68k
4393 port uses this type of @code{parallel} to return pointers in both
4394 @samp{%a0} (the canonical location) and @samp{%d0}.
4396 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4397 the same promotion rules specified in @code{PROMOTE_MODE} if
4398 @var{valtype} is a scalar type.
4400 If the precise function being called is known, @var{func} is a tree
4401 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4402 pointer. This makes it possible to use a different value-returning
4403 convention for specific functions when all their calls are
4406 Some target machines have ``register windows'' so that the register in
4407 which a function returns its value is not the same as the one in which
4408 the caller sees the value. For such machines, you should return
4409 different RTX depending on @var{outgoing}.
4411 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4412 aggregate data types, because these are returned in another way. See
4413 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4416 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4417 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4418 a new target instead.
4421 @defmac LIBCALL_VALUE (@var{mode})
4422 A C expression to create an RTX representing the place where a library
4423 function returns a value of mode @var{mode}.
4425 Note that ``library function'' in this context means a compiler
4426 support routine, used to perform arithmetic, whose name is known
4427 specially by the compiler and was not mentioned in the C code being
4431 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (enum machine_mode @var{mode}, const_rtx @var{fun})
4432 Define this hook if the back-end needs to know the name of the libcall
4433 function in order to determine where the result should be returned.
4435 The mode of the result is given by @var{mode} and the name of the called
4436 library function is given by @var{fun}. The hook should return an RTX
4437 representing the place where the library function result will be returned.
4439 If this hook is not defined, then LIBCALL_VALUE will be used.
4442 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4443 A C expression that is nonzero if @var{regno} is the number of a hard
4444 register in which the values of called function may come back.
4446 A register whose use for returning values is limited to serving as the
4447 second of a pair (for a value of type @code{double}, say) need not be
4448 recognized by this macro. So for most machines, this definition
4452 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4455 If the machine has register windows, so that the caller and the called
4456 function use different registers for the return value, this macro
4457 should recognize only the caller's register numbers.
4459 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4460 for a new target instead.
4463 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4464 A target hook that return @code{true} if @var{regno} is the number of a hard
4465 register in which the values of called function may come back.
4467 A register whose use for returning values is limited to serving as the
4468 second of a pair (for a value of type @code{double}, say) need not be
4469 recognized by this target hook.
4471 If the machine has register windows, so that the caller and the called
4472 function use different registers for the return value, this target hook
4473 should recognize only the caller's register numbers.
4475 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4478 @defmac APPLY_RESULT_SIZE
4479 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4480 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4481 saving and restoring an arbitrary return value.
4484 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4485 This hook should return true if values of type @var{type} are returned
4486 at the most significant end of a register (in other words, if they are
4487 padded at the least significant end). You can assume that @var{type}
4488 is returned in a register; the caller is required to check this.
4490 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4491 be able to hold the complete return value. For example, if a 1-, 2-
4492 or 3-byte structure is returned at the most significant end of a
4493 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4497 @node Aggregate Return
4498 @subsection How Large Values Are Returned
4499 @cindex aggregates as return values
4500 @cindex large return values
4501 @cindex returning aggregate values
4502 @cindex structure value address
4504 When a function value's mode is @code{BLKmode} (and in some other
4505 cases), the value is not returned according to
4506 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4507 caller passes the address of a block of memory in which the value
4508 should be stored. This address is called the @dfn{structure value
4511 This section describes how to control returning structure values in
4514 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4515 This target hook should return a nonzero value to say to return the
4516 function value in memory, just as large structures are always returned.
4517 Here @var{type} will be the data type of the value, and @var{fntype}
4518 will be the type of the function doing the returning, or @code{NULL} for
4521 Note that values of mode @code{BLKmode} must be explicitly handled
4522 by this function. Also, the option @option{-fpcc-struct-return}
4523 takes effect regardless of this macro. On most systems, it is
4524 possible to leave the hook undefined; this causes a default
4525 definition to be used, whose value is the constant 1 for @code{BLKmode}
4526 values, and 0 otherwise.
4528 Do not use this hook to indicate that structures and unions should always
4529 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4533 @defmac DEFAULT_PCC_STRUCT_RETURN
4534 Define this macro to be 1 if all structure and union return values must be
4535 in memory. Since this results in slower code, this should be defined
4536 only if needed for compatibility with other compilers or with an ABI@.
4537 If you define this macro to be 0, then the conventions used for structure
4538 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4541 If not defined, this defaults to the value 1.
4544 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4545 This target hook should return the location of the structure value
4546 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4547 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4548 be @code{NULL}, for libcalls. You do not need to define this target
4549 hook if the address is always passed as an ``invisible'' first
4552 On some architectures the place where the structure value address
4553 is found by the called function is not the same place that the
4554 caller put it. This can be due to register windows, or it could
4555 be because the function prologue moves it to a different place.
4556 @var{incoming} is @code{1} or @code{2} when the location is needed in
4557 the context of the called function, and @code{0} in the context of
4560 If @var{incoming} is nonzero and the address is to be found on the
4561 stack, return a @code{mem} which refers to the frame pointer. If
4562 @var{incoming} is @code{2}, the result is being used to fetch the
4563 structure value address at the beginning of a function. If you need
4564 to emit adjusting code, you should do it at this point.
4567 @defmac PCC_STATIC_STRUCT_RETURN
4568 Define this macro if the usual system convention on the target machine
4569 for returning structures and unions is for the called function to return
4570 the address of a static variable containing the value.
4572 Do not define this if the usual system convention is for the caller to
4573 pass an address to the subroutine.
4575 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4576 nothing when you use @option{-freg-struct-return} mode.
4579 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4580 This target hook returns the mode to be used when accessing raw return registers in @code{__builtin_return}. Define this macro if the value in @var{reg_raw_mode} is not correct.
4583 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_ARG_MODE (int @var{regno})
4584 This target hook returns the mode to be used when accessing raw argument registers in @code{__builtin_apply_args}. Define this macro if the value in @var{reg_raw_mode} is not correct.
4588 @subsection Caller-Saves Register Allocation
4590 If you enable it, GCC can save registers around function calls. This
4591 makes it possible to use call-clobbered registers to hold variables that
4592 must live across calls.
4594 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4595 A C expression to determine whether it is worthwhile to consider placing
4596 a pseudo-register in a call-clobbered hard register and saving and
4597 restoring it around each function call. The expression should be 1 when
4598 this is worth doing, and 0 otherwise.
4600 If you don't define this macro, a default is used which is good on most
4601 machines: @code{4 * @var{calls} < @var{refs}}.
4604 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4605 A C expression specifying which mode is required for saving @var{nregs}
4606 of a pseudo-register in call-clobbered hard register @var{regno}. If
4607 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4608 returned. For most machines this macro need not be defined since GCC
4609 will select the smallest suitable mode.
4612 @node Function Entry
4613 @subsection Function Entry and Exit
4614 @cindex function entry and exit
4618 This section describes the macros that output function entry
4619 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4621 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4622 If defined, a function that outputs the assembler code for entry to a
4623 function. The prologue is responsible for setting up the stack frame,
4624 initializing the frame pointer register, saving registers that must be
4625 saved, and allocating @var{size} additional bytes of storage for the
4626 local variables. @var{size} is an integer. @var{file} is a stdio
4627 stream to which the assembler code should be output.
4629 The label for the beginning of the function need not be output by this
4630 macro. That has already been done when the macro is run.
4632 @findex regs_ever_live
4633 To determine which registers to save, the macro can refer to the array
4634 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4635 @var{r} is used anywhere within the function. This implies the function
4636 prologue should save register @var{r}, provided it is not one of the
4637 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4638 @code{regs_ever_live}.)
4640 On machines that have ``register windows'', the function entry code does
4641 not save on the stack the registers that are in the windows, even if
4642 they are supposed to be preserved by function calls; instead it takes
4643 appropriate steps to ``push'' the register stack, if any non-call-used
4644 registers are used in the function.
4646 @findex frame_pointer_needed
4647 On machines where functions may or may not have frame-pointers, the
4648 function entry code must vary accordingly; it must set up the frame
4649 pointer if one is wanted, and not otherwise. To determine whether a
4650 frame pointer is in wanted, the macro can refer to the variable
4651 @code{frame_pointer_needed}. The variable's value will be 1 at run
4652 time in a function that needs a frame pointer. @xref{Elimination}.
4654 The function entry code is responsible for allocating any stack space
4655 required for the function. This stack space consists of the regions
4656 listed below. In most cases, these regions are allocated in the
4657 order listed, with the last listed region closest to the top of the
4658 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4659 the highest address if it is not defined). You can use a different order
4660 for a machine if doing so is more convenient or required for
4661 compatibility reasons. Except in cases where required by standard
4662 or by a debugger, there is no reason why the stack layout used by GCC
4663 need agree with that used by other compilers for a machine.
4666 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4667 If defined, a function that outputs assembler code at the end of a
4668 prologue. This should be used when the function prologue is being
4669 emitted as RTL, and you have some extra assembler that needs to be
4670 emitted. @xref{prologue instruction pattern}.
4673 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4674 If defined, a function that outputs assembler code at the start of an
4675 epilogue. This should be used when the function epilogue is being
4676 emitted as RTL, and you have some extra assembler that needs to be
4677 emitted. @xref{epilogue instruction pattern}.
4680 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4681 If defined, a function that outputs the assembler code for exit from a
4682 function. The epilogue is responsible for restoring the saved
4683 registers and stack pointer to their values when the function was
4684 called, and returning control to the caller. This macro takes the
4685 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4686 registers to restore are determined from @code{regs_ever_live} and
4687 @code{CALL_USED_REGISTERS} in the same way.
4689 On some machines, there is a single instruction that does all the work
4690 of returning from the function. On these machines, give that
4691 instruction the name @samp{return} and do not define the macro
4692 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4694 Do not define a pattern named @samp{return} if you want the
4695 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4696 switches to control whether return instructions or epilogues are used,
4697 define a @samp{return} pattern with a validity condition that tests the
4698 target switches appropriately. If the @samp{return} pattern's validity
4699 condition is false, epilogues will be used.
4701 On machines where functions may or may not have frame-pointers, the
4702 function exit code must vary accordingly. Sometimes the code for these
4703 two cases is completely different. To determine whether a frame pointer
4704 is wanted, the macro can refer to the variable
4705 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4706 a function that needs a frame pointer.
4708 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4709 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4710 The C variable @code{current_function_is_leaf} is nonzero for such a
4711 function. @xref{Leaf Functions}.
4713 On some machines, some functions pop their arguments on exit while
4714 others leave that for the caller to do. For example, the 68020 when
4715 given @option{-mrtd} pops arguments in functions that take a fixed
4716 number of arguments.
4718 @findex current_function_pops_args
4719 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4720 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4721 needs to know what was decided. The number of bytes of the current
4722 function's arguments that this function should pop is available in
4723 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4728 @findex current_function_pretend_args_size
4729 A region of @code{current_function_pretend_args_size} bytes of
4730 uninitialized space just underneath the first argument arriving on the
4731 stack. (This may not be at the very start of the allocated stack region
4732 if the calling sequence has pushed anything else since pushing the stack
4733 arguments. But usually, on such machines, nothing else has been pushed
4734 yet, because the function prologue itself does all the pushing.) This
4735 region is used on machines where an argument may be passed partly in
4736 registers and partly in memory, and, in some cases to support the
4737 features in @code{<stdarg.h>}.
4740 An area of memory used to save certain registers used by the function.
4741 The size of this area, which may also include space for such things as
4742 the return address and pointers to previous stack frames, is
4743 machine-specific and usually depends on which registers have been used
4744 in the function. Machines with register windows often do not require
4748 A region of at least @var{size} bytes, possibly rounded up to an allocation
4749 boundary, to contain the local variables of the function. On some machines,
4750 this region and the save area may occur in the opposite order, with the
4751 save area closer to the top of the stack.
4754 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4755 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4756 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4757 argument lists of the function. @xref{Stack Arguments}.
4760 @defmac EXIT_IGNORE_STACK
4761 Define this macro as a C expression that is nonzero if the return
4762 instruction or the function epilogue ignores the value of the stack
4763 pointer; in other words, if it is safe to delete an instruction to
4764 adjust the stack pointer before a return from the function. The
4767 Note that this macro's value is relevant only for functions for which
4768 frame pointers are maintained. It is never safe to delete a final
4769 stack adjustment in a function that has no frame pointer, and the
4770 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4773 @defmac EPILOGUE_USES (@var{regno})
4774 Define this macro as a C expression that is nonzero for registers that are
4775 used by the epilogue or the @samp{return} pattern. The stack and frame
4776 pointer registers are already assumed to be used as needed.
4779 @defmac EH_USES (@var{regno})
4780 Define this macro as a C expression that is nonzero for registers that are
4781 used by the exception handling mechanism, and so should be considered live
4782 on entry to an exception edge.
4785 @defmac DELAY_SLOTS_FOR_EPILOGUE
4786 Define this macro if the function epilogue contains delay slots to which
4787 instructions from the rest of the function can be ``moved''. The
4788 definition should be a C expression whose value is an integer
4789 representing the number of delay slots there.
4792 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4793 A C expression that returns 1 if @var{insn} can be placed in delay
4794 slot number @var{n} of the epilogue.
4796 The argument @var{n} is an integer which identifies the delay slot now
4797 being considered (since different slots may have different rules of
4798 eligibility). It is never negative and is always less than the number
4799 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4800 If you reject a particular insn for a given delay slot, in principle, it
4801 may be reconsidered for a subsequent delay slot. Also, other insns may
4802 (at least in principle) be considered for the so far unfilled delay
4805 @findex current_function_epilogue_delay_list
4806 @findex final_scan_insn
4807 The insns accepted to fill the epilogue delay slots are put in an RTL
4808 list made with @code{insn_list} objects, stored in the variable
4809 @code{current_function_epilogue_delay_list}. The insn for the first
4810 delay slot comes first in the list. Your definition of the macro
4811 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4812 outputting the insns in this list, usually by calling
4813 @code{final_scan_insn}.
4815 You need not define this macro if you did not define
4816 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4819 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function})
4820 A function that outputs the assembler code for a thunk
4821 function, used to implement C++ virtual function calls with multiple
4822 inheritance. The thunk acts as a wrapper around a virtual function,
4823 adjusting the implicit object parameter before handing control off to
4826 First, emit code to add the integer @var{delta} to the location that
4827 contains the incoming first argument. Assume that this argument
4828 contains a pointer, and is the one used to pass the @code{this} pointer
4829 in C++. This is the incoming argument @emph{before} the function prologue,
4830 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4831 all other incoming arguments.
4833 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4834 made after adding @code{delta}. In particular, if @var{p} is the
4835 adjusted pointer, the following adjustment should be made:
4838 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4841 After the additions, emit code to jump to @var{function}, which is a
4842 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4843 not touch the return address. Hence returning from @var{FUNCTION} will
4844 return to whoever called the current @samp{thunk}.
4846 The effect must be as if @var{function} had been called directly with
4847 the adjusted first argument. This macro is responsible for emitting all
4848 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4849 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4851 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4852 have already been extracted from it.) It might possibly be useful on
4853 some targets, but probably not.
4855 If you do not define this macro, the target-independent code in the C++
4856 front end will generate a less efficient heavyweight thunk that calls
4857 @var{function} instead of jumping to it. The generic approach does
4858 not support varargs.
4861 @deftypefn {Target Hook} bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (const_tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, const_tree @var{function})
4862 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4863 to output the assembler code for the thunk function specified by the
4864 arguments it is passed, and false otherwise. In the latter case, the
4865 generic approach will be used by the C++ front end, with the limitations
4870 @subsection Generating Code for Profiling
4871 @cindex profiling, code generation
4873 These macros will help you generate code for profiling.
4875 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4876 A C statement or compound statement to output to @var{file} some
4877 assembler code to call the profiling subroutine @code{mcount}.
4880 The details of how @code{mcount} expects to be called are determined by
4881 your operating system environment, not by GCC@. To figure them out,
4882 compile a small program for profiling using the system's installed C
4883 compiler and look at the assembler code that results.
4885 Older implementations of @code{mcount} expect the address of a counter
4886 variable to be loaded into some register. The name of this variable is
4887 @samp{LP} followed by the number @var{labelno}, so you would generate
4888 the name using @samp{LP%d} in a @code{fprintf}.
4891 @defmac PROFILE_HOOK
4892 A C statement or compound statement to output to @var{file} some assembly
4893 code to call the profiling subroutine @code{mcount} even the target does
4894 not support profiling.
4897 @defmac NO_PROFILE_COUNTERS
4898 Define this macro to be an expression with a nonzero value if the
4899 @code{mcount} subroutine on your system does not need a counter variable
4900 allocated for each function. This is true for almost all modern
4901 implementations. If you define this macro, you must not use the
4902 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4905 @defmac PROFILE_BEFORE_PROLOGUE
4906 Define this macro if the code for function profiling should come before
4907 the function prologue. Normally, the profiling code comes after.
4911 @subsection Permitting tail calls
4914 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4915 True if it is ok to do sibling call optimization for the specified
4916 call expression @var{exp}. @var{decl} will be the called function,
4917 or @code{NULL} if this is an indirect call.
4919 It is not uncommon for limitations of calling conventions to prevent
4920 tail calls to functions outside the current unit of translation, or
4921 during PIC compilation. The hook is used to enforce these restrictions,
4922 as the @code{sibcall} md pattern can not fail, or fall over to a
4923 ``normal'' call. The criteria for successful sibling call optimization
4924 may vary greatly between different architectures.
4927 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4928 Add any hard registers to @var{regs} that are live on entry to the
4929 function. This hook only needs to be defined to provide registers that
4930 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4931 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4932 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4933 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4936 @node Stack Smashing Protection
4937 @subsection Stack smashing protection
4938 @cindex stack smashing protection
4940 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4941 This hook returns a @code{DECL} node for the external variable to use
4942 for the stack protection guard. This variable is initialized by the
4943 runtime to some random value and is used to initialize the guard value
4944 that is placed at the top of the local stack frame. The type of this
4945 variable must be @code{ptr_type_node}.
4947 The default version of this hook creates a variable called
4948 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4951 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4952 This hook returns a tree expression that alerts the runtime that the
4953 stack protect guard variable has been modified. This expression should
4954 involve a call to a @code{noreturn} function.
4956 The default version of this hook invokes a function called
4957 @samp{__stack_chk_fail}, taking no arguments. This function is
4958 normally defined in @file{libgcc2.c}.
4961 @deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
4962 Whether this target supports splitting the stack when the options described in @var{opts} have been passed. This is called after options have been parsed, so the target may reject splitting the stack in some configurations. The default version of this hook returns false. If @var{report} is true, this function may issue a warning or error; if @var{report} is false, it must simply return a value
4966 @section Implementing the Varargs Macros
4967 @cindex varargs implementation
4969 GCC comes with an implementation of @code{<varargs.h>} and
4970 @code{<stdarg.h>} that work without change on machines that pass arguments
4971 on the stack. Other machines require their own implementations of
4972 varargs, and the two machine independent header files must have
4973 conditionals to include it.
4975 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4976 the calling convention for @code{va_start}. The traditional
4977 implementation takes just one argument, which is the variable in which
4978 to store the argument pointer. The ISO implementation of
4979 @code{va_start} takes an additional second argument. The user is
4980 supposed to write the last named argument of the function here.
4982 However, @code{va_start} should not use this argument. The way to find
4983 the end of the named arguments is with the built-in functions described
4986 @defmac __builtin_saveregs ()
4987 Use this built-in function to save the argument registers in memory so
4988 that the varargs mechanism can access them. Both ISO and traditional
4989 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4990 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4992 On some machines, @code{__builtin_saveregs} is open-coded under the
4993 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4994 other machines, it calls a routine written in assembler language,
4995 found in @file{libgcc2.c}.
4997 Code generated for the call to @code{__builtin_saveregs} appears at the
4998 beginning of the function, as opposed to where the call to
4999 @code{__builtin_saveregs} is written, regardless of what the code is.
5000 This is because the registers must be saved before the function starts
5001 to use them for its own purposes.
5002 @c i rewrote the first sentence above to fix an overfull hbox. --mew
5006 @defmac __builtin_next_arg (@var{lastarg})
5007 This builtin returns the address of the first anonymous stack
5008 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
5009 returns the address of the location above the first anonymous stack
5010 argument. Use it in @code{va_start} to initialize the pointer for
5011 fetching arguments from the stack. Also use it in @code{va_start} to
5012 verify that the second parameter @var{lastarg} is the last named argument
5013 of the current function.
5016 @defmac __builtin_classify_type (@var{object})
5017 Since each machine has its own conventions for which data types are
5018 passed in which kind of register, your implementation of @code{va_arg}
5019 has to embody these conventions. The easiest way to categorize the
5020 specified data type is to use @code{__builtin_classify_type} together
5021 with @code{sizeof} and @code{__alignof__}.
5023 @code{__builtin_classify_type} ignores the value of @var{object},
5024 considering only its data type. It returns an integer describing what
5025 kind of type that is---integer, floating, pointer, structure, and so on.
5027 The file @file{typeclass.h} defines an enumeration that you can use to
5028 interpret the values of @code{__builtin_classify_type}.
5031 These machine description macros help implement varargs:
5033 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5034 If defined, this hook produces the machine-specific code for a call to
5035 @code{__builtin_saveregs}. This code will be moved to the very
5036 beginning of the function, before any parameter access are made. The
5037 return value of this function should be an RTX that contains the value
5038 to use as the return of @code{__builtin_saveregs}.
5041 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (cumulative_args_t @var{args_so_far}, enum machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
5042 This target hook offers an alternative to using
5043 @code{__builtin_saveregs} and defining the hook
5044 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5045 register arguments into the stack so that all the arguments appear to
5046 have been passed consecutively on the stack. Once this is done, you can
5047 use the standard implementation of varargs that works for machines that
5048 pass all their arguments on the stack.
5050 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5051 structure, containing the values that are obtained after processing the
5052 named arguments. The arguments @var{mode} and @var{type} describe the
5053 last named argument---its machine mode and its data type as a tree node.
5055 The target hook should do two things: first, push onto the stack all the
5056 argument registers @emph{not} used for the named arguments, and second,
5057 store the size of the data thus pushed into the @code{int}-valued
5058 variable pointed to by @var{pretend_args_size}. The value that you
5059 store here will serve as additional offset for setting up the stack
5062 Because you must generate code to push the anonymous arguments at
5063 compile time without knowing their data types,
5064 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5065 have just a single category of argument register and use it uniformly
5068 If the argument @var{second_time} is nonzero, it means that the
5069 arguments of the function are being analyzed for the second time. This
5070 happens for an inline function, which is not actually compiled until the
5071 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5072 not generate any instructions in this case.
5075 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca})
5076 Define this hook to return @code{true} if the location where a function
5077 argument is passed depends on whether or not it is a named argument.
5079 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5080 is set for varargs and stdarg functions. If this hook returns
5081 @code{true}, the @var{named} argument is always true for named
5082 arguments, and false for unnamed arguments. If it returns @code{false},
5083 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5084 then all arguments are treated as named. Otherwise, all named arguments
5085 except the last are treated as named.
5087 You need not define this hook if it always returns @code{false}.
5090 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca})
5091 If you need to conditionally change ABIs so that one works with
5092 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5093 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5094 defined, then define this hook to return @code{true} if
5095 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5096 Otherwise, you should not define this hook.
5100 @section Trampolines for Nested Functions
5101 @cindex trampolines for nested functions
5102 @cindex nested functions, trampolines for
5104 A @dfn{trampoline} is a small piece of code that is created at run time
5105 when the address of a nested function is taken. It normally resides on
5106 the stack, in the stack frame of the containing function. These macros
5107 tell GCC how to generate code to allocate and initialize a
5110 The instructions in the trampoline must do two things: load a constant
5111 address into the static chain register, and jump to the real address of
5112 the nested function. On CISC machines such as the m68k, this requires
5113 two instructions, a move immediate and a jump. Then the two addresses
5114 exist in the trampoline as word-long immediate operands. On RISC
5115 machines, it is often necessary to load each address into a register in
5116 two parts. Then pieces of each address form separate immediate
5119 The code generated to initialize the trampoline must store the variable
5120 parts---the static chain value and the function address---into the
5121 immediate operands of the instructions. On a CISC machine, this is
5122 simply a matter of copying each address to a memory reference at the
5123 proper offset from the start of the trampoline. On a RISC machine, it
5124 may be necessary to take out pieces of the address and store them
5127 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5128 This hook is called by @code{assemble_trampoline_template} to output,
5129 on the stream @var{f}, assembler code for a block of data that contains
5130 the constant parts of a trampoline. This code should not include a
5131 label---the label is taken care of automatically.
5133 If you do not define this hook, it means no template is needed
5134 for the target. Do not define this hook on systems where the block move
5135 code to copy the trampoline into place would be larger than the code
5136 to generate it on the spot.
5139 @defmac TRAMPOLINE_SECTION
5140 Return the section into which the trampoline template is to be placed
5141 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5144 @defmac TRAMPOLINE_SIZE
5145 A C expression for the size in bytes of the trampoline, as an integer.
5148 @defmac TRAMPOLINE_ALIGNMENT
5149 Alignment required for trampolines, in bits.
5151 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5152 is used for aligning trampolines.
5155 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5156 This hook is called to initialize a trampoline.
5157 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5158 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5159 RTX for the static chain value that should be passed to the function
5162 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5163 first thing this hook should do is emit a block move into @var{m_tramp}
5164 from the memory block returned by @code{assemble_trampoline_template}.
5165 Note that the block move need only cover the constant parts of the
5166 trampoline. If the target isolates the variable parts of the trampoline
5167 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5169 If the target requires any other actions, such as flushing caches or
5170 enabling stack execution, these actions should be performed after
5171 initializing the trampoline proper.
5174 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5175 This hook should perform any machine-specific adjustment in
5176 the address of the trampoline. Its argument contains the address of the
5177 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5178 the address to be used for a function call should be different from the
5179 address at which the template was stored, the different address should
5180 be returned; otherwise @var{addr} should be returned unchanged.
5181 If this hook is not defined, @var{addr} will be used for function calls.
5184 Implementing trampolines is difficult on many machines because they have
5185 separate instruction and data caches. Writing into a stack location
5186 fails to clear the memory in the instruction cache, so when the program
5187 jumps to that location, it executes the old contents.
5189 Here are two possible solutions. One is to clear the relevant parts of
5190 the instruction cache whenever a trampoline is set up. The other is to
5191 make all trampolines identical, by having them jump to a standard
5192 subroutine. The former technique makes trampoline execution faster; the
5193 latter makes initialization faster.
5195 To clear the instruction cache when a trampoline is initialized, define
5196 the following macro.
5198 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5199 If defined, expands to a C expression clearing the @emph{instruction
5200 cache} in the specified interval. The definition of this macro would
5201 typically be a series of @code{asm} statements. Both @var{beg} and
5202 @var{end} are both pointer expressions.
5205 To use a standard subroutine, define the following macro. In addition,
5206 you must make sure that the instructions in a trampoline fill an entire
5207 cache line with identical instructions, or else ensure that the
5208 beginning of the trampoline code is always aligned at the same point in
5209 its cache line. Look in @file{m68k.h} as a guide.
5211 @defmac TRANSFER_FROM_TRAMPOLINE
5212 Define this macro if trampolines need a special subroutine to do their
5213 work. The macro should expand to a series of @code{asm} statements
5214 which will be compiled with GCC@. They go in a library function named
5215 @code{__transfer_from_trampoline}.
5217 If you need to avoid executing the ordinary prologue code of a compiled
5218 C function when you jump to the subroutine, you can do so by placing a
5219 special label of your own in the assembler code. Use one @code{asm}
5220 statement to generate an assembler label, and another to make the label
5221 global. Then trampolines can use that label to jump directly to your
5222 special assembler code.
5226 @section Implicit Calls to Library Routines
5227 @cindex library subroutine names
5228 @cindex @file{libgcc.a}
5230 @c prevent bad page break with this line
5231 Here is an explanation of implicit calls to library routines.
5233 @defmac DECLARE_LIBRARY_RENAMES
5234 This macro, if defined, should expand to a piece of C code that will get
5235 expanded when compiling functions for libgcc.a. It can be used to
5236 provide alternate names for GCC's internal library functions if there
5237 are ABI-mandated names that the compiler should provide.
5240 @findex set_optab_libfunc
5241 @findex init_one_libfunc
5242 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5243 This hook should declare additional library routines or rename
5244 existing ones, using the functions @code{set_optab_libfunc} and
5245 @code{init_one_libfunc} defined in @file{optabs.c}.
5246 @code{init_optabs} calls this macro after initializing all the normal
5249 The default is to do nothing. Most ports don't need to define this hook.
5252 @deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX
5253 If false (the default), internal library routines start with two
5254 underscores. If set to true, these routines start with @code{__gnu_}
5255 instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This
5256 currently only affects functions defined in @file{libgcc2.c}. If this
5257 is set to true, the @file{tm.h} file must also
5258 @code{#define LIBGCC2_GNU_PREFIX}.
5261 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5262 This macro should return @code{true} if the library routine that
5263 implements the floating point comparison operator @var{comparison} in
5264 mode @var{mode} will return a boolean, and @var{false} if it will
5267 GCC's own floating point libraries return tristates from the
5268 comparison operators, so the default returns false always. Most ports
5269 don't need to define this macro.
5272 @defmac TARGET_LIB_INT_CMP_BIASED
5273 This macro should evaluate to @code{true} if the integer comparison
5274 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5275 operand is smaller than the second, 1 to indicate that they are equal,
5276 and 2 to indicate that the first operand is greater than the second.
5277 If this macro evaluates to @code{false} the comparison functions return
5278 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5279 in @file{libgcc.a}, you do not need to define this macro.
5282 @cindex @code{EDOM}, implicit usage
5285 The value of @code{EDOM} on the target machine, as a C integer constant
5286 expression. If you don't define this macro, GCC does not attempt to
5287 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5288 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5291 If you do not define @code{TARGET_EDOM}, then compiled code reports
5292 domain errors by calling the library function and letting it report the
5293 error. If mathematical functions on your system use @code{matherr} when
5294 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5295 that @code{matherr} is used normally.
5298 @cindex @code{errno}, implicit usage
5299 @defmac GEN_ERRNO_RTX
5300 Define this macro as a C expression to create an rtl expression that
5301 refers to the global ``variable'' @code{errno}. (On certain systems,
5302 @code{errno} may not actually be a variable.) If you don't define this
5303 macro, a reasonable default is used.
5306 @cindex C99 math functions, implicit usage
5307 @defmac TARGET_C99_FUNCTIONS
5308 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5309 @code{sinf} and similarly for other functions defined by C99 standard. The
5310 default is zero because a number of existing systems lack support for these
5311 functions in their runtime so this macro needs to be redefined to one on
5312 systems that do support the C99 runtime.
5315 @cindex sincos math function, implicit usage
5316 @defmac TARGET_HAS_SINCOS
5317 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5318 and @code{cos} with the same argument to a call to @code{sincos}. The
5319 default is zero. The target has to provide the following functions:
5321 void sincos(double x, double *sin, double *cos);
5322 void sincosf(float x, float *sin, float *cos);
5323 void sincosl(long double x, long double *sin, long double *cos);
5327 @defmac NEXT_OBJC_RUNTIME
5328 Define this macro to generate code for Objective-C message sending using
5329 the calling convention of the NeXT system. This calling convention
5330 involves passing the object, the selector and the method arguments all
5331 at once to the method-lookup library function.
5333 The default calling convention passes just the object and the selector
5334 to the lookup function, which returns a pointer to the method.
5337 @node Addressing Modes
5338 @section Addressing Modes
5339 @cindex addressing modes
5341 @c prevent bad page break with this line
5342 This is about addressing modes.
5344 @defmac HAVE_PRE_INCREMENT
5345 @defmacx HAVE_PRE_DECREMENT
5346 @defmacx HAVE_POST_INCREMENT
5347 @defmacx HAVE_POST_DECREMENT
5348 A C expression that is nonzero if the machine supports pre-increment,
5349 pre-decrement, post-increment, or post-decrement addressing respectively.
5352 @defmac HAVE_PRE_MODIFY_DISP
5353 @defmacx HAVE_POST_MODIFY_DISP
5354 A C expression that is nonzero if the machine supports pre- or
5355 post-address side-effect generation involving constants other than
5356 the size of the memory operand.
5359 @defmac HAVE_PRE_MODIFY_REG
5360 @defmacx HAVE_POST_MODIFY_REG
5361 A C expression that is nonzero if the machine supports pre- or
5362 post-address side-effect generation involving a register displacement.
5365 @defmac CONSTANT_ADDRESS_P (@var{x})
5366 A C expression that is 1 if the RTX @var{x} is a constant which
5367 is a valid address. On most machines the default definition of
5368 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5369 is acceptable, but a few machines are more restrictive as to which
5370 constant addresses are supported.
5373 @defmac CONSTANT_P (@var{x})
5374 @code{CONSTANT_P}, which is defined by target-independent code,
5375 accepts integer-values expressions whose values are not explicitly
5376 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5377 expressions and @code{const} arithmetic expressions, in addition to
5378 @code{const_int} and @code{const_double} expressions.
5381 @defmac MAX_REGS_PER_ADDRESS
5382 A number, the maximum number of registers that can appear in a valid
5383 memory address. Note that it is up to you to specify a value equal to
5384 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5388 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5389 A function that returns whether @var{x} (an RTX) is a legitimate memory
5390 address on the target machine for a memory operand of mode @var{mode}.
5392 Legitimate addresses are defined in two variants: a strict variant and a
5393 non-strict one. The @var{strict} parameter chooses which variant is
5394 desired by the caller.
5396 The strict variant is used in the reload pass. It must be defined so
5397 that any pseudo-register that has not been allocated a hard register is
5398 considered a memory reference. This is because in contexts where some
5399 kind of register is required, a pseudo-register with no hard register
5400 must be rejected. For non-hard registers, the strict variant should look
5401 up the @code{reg_renumber} array; it should then proceed using the hard
5402 register number in the array, or treat the pseudo as a memory reference
5403 if the array holds @code{-1}.
5405 The non-strict variant is used in other passes. It must be defined to
5406 accept all pseudo-registers in every context where some kind of
5407 register is required.
5409 Normally, constant addresses which are the sum of a @code{symbol_ref}
5410 and an integer are stored inside a @code{const} RTX to mark them as
5411 constant. Therefore, there is no need to recognize such sums
5412 specifically as legitimate addresses. Normally you would simply
5413 recognize any @code{const} as legitimate.
5415 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5416 sums that are not marked with @code{const}. It assumes that a naked
5417 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5418 naked constant sums as illegitimate addresses, so that none of them will
5419 be given to @code{PRINT_OPERAND_ADDRESS}.
5421 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5422 On some machines, whether a symbolic address is legitimate depends on
5423 the section that the address refers to. On these machines, define the
5424 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5425 into the @code{symbol_ref}, and then check for it here. When you see a
5426 @code{const}, you will have to look inside it to find the
5427 @code{symbol_ref} in order to determine the section. @xref{Assembler
5430 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5431 Some ports are still using a deprecated legacy substitute for
5432 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5436 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5440 and should @code{goto @var{label}} if the address @var{x} is a valid
5441 address on the target machine for a memory operand of mode @var{mode}.
5443 @findex REG_OK_STRICT
5444 Compiler source files that want to use the strict variant of this
5445 macro define the macro @code{REG_OK_STRICT}. You should use an
5446 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5447 that case and the non-strict variant otherwise.
5449 Using the hook is usually simpler because it limits the number of
5450 files that are recompiled when changes are made.
5453 @defmac TARGET_MEM_CONSTRAINT
5454 A single character to be used instead of the default @code{'m'}
5455 character for general memory addresses. This defines the constraint
5456 letter which matches the memory addresses accepted by
5457 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5458 support new address formats in your back end without changing the
5459 semantics of the @code{'m'} constraint. This is necessary in order to
5460 preserve functionality of inline assembly constructs using the
5461 @code{'m'} constraint.
5464 @defmac FIND_BASE_TERM (@var{x})
5465 A C expression to determine the base term of address @var{x},
5466 or to provide a simplified version of @var{x} from which @file{alias.c}
5467 can easily find the base term. This macro is used in only two places:
5468 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5470 It is always safe for this macro to not be defined. It exists so
5471 that alias analysis can understand machine-dependent addresses.
5473 The typical use of this macro is to handle addresses containing
5474 a label_ref or symbol_ref within an UNSPEC@.
5477 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode})
5478 This hook is given an invalid memory address @var{x} for an
5479 operand of mode @var{mode} and should try to return a valid memory
5482 @findex break_out_memory_refs
5483 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5484 and @var{oldx} will be the operand that was given to that function to produce
5487 The code of the hook should not alter the substructure of
5488 @var{x}. If it transforms @var{x} into a more legitimate form, it
5489 should return the new @var{x}.
5491 It is not necessary for this hook to come up with a legitimate address.
5492 The compiler has standard ways of doing so in all cases. In fact, it
5493 is safe to omit this hook or make it return @var{x} if it cannot find
5494 a valid way to legitimize the address. But often a machine-dependent
5495 strategy can generate better code.
5498 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5499 A C compound statement that attempts to replace @var{x}, which is an address
5500 that needs reloading, with a valid memory address for an operand of mode
5501 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5502 It is not necessary to define this macro, but it might be useful for
5503 performance reasons.
5505 For example, on the i386, it is sometimes possible to use a single
5506 reload register instead of two by reloading a sum of two pseudo
5507 registers into a register. On the other hand, for number of RISC
5508 processors offsets are limited so that often an intermediate address
5509 needs to be generated in order to address a stack slot. By defining
5510 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5511 generated for adjacent some stack slots can be made identical, and thus
5514 @emph{Note}: This macro should be used with caution. It is necessary
5515 to know something of how reload works in order to effectively use this,
5516 and it is quite easy to produce macros that build in too much knowledge
5517 of reload internals.
5519 @emph{Note}: This macro must be able to reload an address created by a
5520 previous invocation of this macro. If it fails to handle such addresses
5521 then the compiler may generate incorrect code or abort.
5524 The macro definition should use @code{push_reload} to indicate parts that
5525 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5526 suitable to be passed unaltered to @code{push_reload}.
5528 The code generated by this macro must not alter the substructure of
5529 @var{x}. If it transforms @var{x} into a more legitimate form, it
5530 should assign @var{x} (which will always be a C variable) a new value.
5531 This also applies to parts that you change indirectly by calling
5534 @findex strict_memory_address_p
5535 The macro definition may use @code{strict_memory_address_p} to test if
5536 the address has become legitimate.
5539 If you want to change only a part of @var{x}, one standard way of doing
5540 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5541 single level of rtl. Thus, if the part to be changed is not at the
5542 top level, you'll need to replace first the top level.
5543 It is not necessary for this macro to come up with a legitimate
5544 address; but often a machine-dependent strategy can generate better code.
5547 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr})
5548 This hook returns @code{true} if memory address @var{addr} can have
5549 different meanings depending on the machine mode of the memory
5550 reference it is used for or if the address is valid for some modes
5553 Autoincrement and autodecrement addresses typically have mode-dependent
5554 effects because the amount of the increment or decrement is the size
5555 of the operand being addressed. Some machines have other mode-dependent
5556 addresses. Many RISC machines have no mode-dependent addresses.
5558 You may assume that @var{addr} is a valid address for the machine.
5560 The default version of this hook returns @code{false}.
5563 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5564 A C statement or compound statement with a conditional @code{goto
5565 @var{label};} executed if memory address @var{x} (an RTX) can have
5566 different meanings depending on the machine mode of the memory
5567 reference it is used for or if the address is valid for some modes
5570 Autoincrement and autodecrement addresses typically have mode-dependent
5571 effects because the amount of the increment or decrement is the size
5572 of the operand being addressed. Some machines have other mode-dependent
5573 addresses. Many RISC machines have no mode-dependent addresses.
5575 You may assume that @var{addr} is a valid address for the machine.
5577 These are obsolete macros, replaced by the
5578 @code{TARGET_MODE_DEPENDENT_ADDRESS_P} target hook.
5581 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5582 This hook returns true if @var{x} is a legitimate constant for a
5583 @var{mode}-mode immediate operand on the target machine. You can assume that
5584 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5586 The default definition returns true.
5589 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5590 This hook is used to undo the possibly obfuscating effects of the
5591 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5592 macros. Some backend implementations of these macros wrap symbol
5593 references inside an @code{UNSPEC} rtx to represent PIC or similar
5594 addressing modes. This target hook allows GCC's optimizers to understand
5595 the semantics of these opaque @code{UNSPEC}s by converting them back
5596 into their original form.
5599 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (enum machine_mode @var{mode}, rtx @var{x})
5600 This hook should return true if @var{x} is of a form that cannot (or
5601 should not) be spilled to the constant pool. @var{mode} is the mode
5604 The default version of this hook returns false.
5606 The primary reason to define this hook is to prevent reload from
5607 deciding that a non-legitimate constant would be better reloaded
5608 from the constant pool instead of spilling and reloading a register
5609 holding the constant. This restriction is often true of addresses
5610 of TLS symbols for various targets.
5613 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, const_rtx @var{x})
5614 This hook should return true if pool entries for constant @var{x} can
5615 be placed in an @code{object_block} structure. @var{mode} is the mode
5618 The default version returns false for all constants.
5621 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (unsigned @var{fn}, bool @var{md_fn}, bool @var{sqrt})
5622 This hook should return the DECL of a function that implements reciprocal of
5623 the builtin function with builtin function code @var{fn}, or
5624 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5625 when @var{fn} is a code of a machine-dependent builtin function. When
5626 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5627 of a square root function are performed, and only reciprocals of @code{sqrt}
5631 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5632 This hook should return the DECL of a function @var{f} that given an
5633 address @var{addr} as an argument returns a mask @var{m} that can be
5634 used to extract from two vectors the relevant data that resides in
5635 @var{addr} in case @var{addr} is not properly aligned.
5637 The autovectorizer, when vectorizing a load operation from an address
5638 @var{addr} that may be unaligned, will generate two vector loads from
5639 the two aligned addresses around @var{addr}. It then generates a
5640 @code{REALIGN_LOAD} operation to extract the relevant data from the
5641 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5642 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5643 the third argument, @var{OFF}, defines how the data will be extracted
5644 from these two vectors: if @var{OFF} is 0, then the returned vector is
5645 @var{v2}; otherwise, the returned vector is composed from the last
5646 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5647 @var{OFF} elements of @var{v2}.
5649 If this hook is defined, the autovectorizer will generate a call
5650 to @var{f} (using the DECL tree that this hook returns) and will
5651 use the return value of @var{f} as the argument @var{OFF} to
5652 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5653 should comply with the semantics expected by @code{REALIGN_LOAD}
5655 If this hook is not defined, then @var{addr} will be used as
5656 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5657 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5660 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5661 This hook should return the DECL of a function @var{f} that implements
5662 widening multiplication of the even elements of two input vectors of type @var{x}.
5664 If this hook is defined, the autovectorizer will use it along with the
5665 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5666 widening multiplication in cases that the order of the results does not have to be
5667 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5668 @code{widen_mult_hi/lo} idioms will be used.
5671 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5672 This hook should return the DECL of a function @var{f} that implements
5673 widening multiplication of the odd elements of two input vectors of type @var{x}.
5675 If this hook is defined, the autovectorizer will use it along with the
5676 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5677 widening multiplication in cases that the order of the results does not have to be
5678 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5679 @code{widen_mult_hi/lo} idioms will be used.
5682 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5683 Returns cost of different scalar or vector statements for vectorization cost model.
5684 For vector memory operations the cost may depend on type (@var{vectype}) and
5685 misalignment value (@var{misalign}).
5688 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5689 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5692 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VEC_PERM (tree @var{type}, tree *@var{mask_element_type})
5693 Target builtin that implements vector permute.
5696 @deftypefn {Target Hook} bool TARGET_VECTORIZE_BUILTIN_VEC_PERM_OK (tree @var{vec_type}, tree @var{mask})
5697 Return true if a vector created for @code{builtin_vec_perm} is valid.
5700 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5701 This hook should return the DECL of a function that implements conversion of the
5702 input vector of type @var{src_type} to type @var{dest_type}.
5703 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5704 specifies how the conversion is to be applied
5705 (truncation, rounding, etc.).
5707 If this hook is defined, the autovectorizer will use the
5708 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5709 conversion. Otherwise, it will return @code{NULL_TREE}.
5712 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5713 This hook should return the decl of a function that implements the
5714 vectorized variant of the builtin function with builtin function code
5715 @var{code} or @code{NULL_TREE} if such a function is not available.
5716 The value of @var{fndecl} is the builtin function declaration. The
5717 return type of the vectorized function shall be of vector type
5718 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5721 @deftypefn {Target Hook} bool TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT (enum machine_mode @var{mode}, const_tree @var{type}, int @var{misalignment}, bool @var{is_packed})
5722 This hook should return true if the target supports misaligned vector
5723 store/load of a specific factor denoted in the @var{misalignment}
5724 parameter. The vector store/load should be of machine mode @var{mode} and
5725 the elements in the vectors should be of type @var{type}. @var{is_packed}
5726 parameter is true if the memory access is defined in a packed struct.
5729 @deftypefn {Target Hook} {enum machine_mode} TARGET_VECTORIZE_PREFERRED_SIMD_MODE (enum machine_mode @var{mode})
5730 This hook should return the preferred mode for vectorizing scalar
5731 mode @var{mode}. The default is
5732 equal to @code{word_mode}, because the vectorizer can do some
5733 transformations even in absence of specialized @acronym{SIMD} hardware.
5736 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void)
5737 This hook should return a mask of sizes that should be iterated over
5738 after trying to autovectorize using the vector size derived from the
5739 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5740 The default is zero which means to not iterate over other vector sizes.
5743 @node Anchored Addresses
5744 @section Anchored Addresses
5745 @cindex anchored addresses
5746 @cindex @option{-fsection-anchors}
5748 GCC usually addresses every static object as a separate entity.
5749 For example, if we have:
5753 int foo (void) @{ return a + b + c; @}
5756 the code for @code{foo} will usually calculate three separate symbolic
5757 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5758 it would be better to calculate just one symbolic address and access
5759 the three variables relative to it. The equivalent pseudocode would
5765 register int *xr = &x;
5766 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5770 (which isn't valid C). We refer to shared addresses like @code{x} as
5771 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5773 The hooks below describe the target properties that GCC needs to know
5774 in order to make effective use of section anchors. It won't use
5775 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5776 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5778 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5779 The minimum offset that should be applied to a section anchor.
5780 On most targets, it should be the smallest offset that can be
5781 applied to a base register while still giving a legitimate address
5782 for every mode. The default value is 0.
5785 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5786 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5787 offset that should be applied to section anchors. The default
5791 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5792 Write the assembly code to define section anchor @var{x}, which is a
5793 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5794 The hook is called with the assembly output position set to the beginning
5795 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5797 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5798 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5799 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5800 is @code{NULL}, which disables the use of section anchors altogether.
5803 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
5804 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5805 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5806 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5808 The default version is correct for most targets, but you might need to
5809 intercept this hook to handle things like target-specific attributes
5810 or target-specific sections.
5813 @node Condition Code
5814 @section Condition Code Status
5815 @cindex condition code status
5817 The macros in this section can be split in two families, according to the
5818 two ways of representing condition codes in GCC.
5820 The first representation is the so called @code{(cc0)} representation
5821 (@pxref{Jump Patterns}), where all instructions can have an implicit
5822 clobber of the condition codes. The second is the condition code
5823 register representation, which provides better schedulability for
5824 architectures that do have a condition code register, but on which
5825 most instructions do not affect it. The latter category includes
5828 The implicit clobbering poses a strong restriction on the placement of
5829 the definition and use of the condition code, which need to be in adjacent
5830 insns for machines using @code{(cc0)}. This can prevent important
5831 optimizations on some machines. For example, on the IBM RS/6000, there
5832 is a delay for taken branches unless the condition code register is set
5833 three instructions earlier than the conditional branch. The instruction
5834 scheduler cannot perform this optimization if it is not permitted to
5835 separate the definition and use of the condition code register.
5837 For this reason, it is possible and suggested to use a register to
5838 represent the condition code for new ports. If there is a specific
5839 condition code register in the machine, use a hard register. If the
5840 condition code or comparison result can be placed in any general register,
5841 or if there are multiple condition registers, use a pseudo register.
5842 Registers used to store the condition code value will usually have a mode
5843 that is in class @code{MODE_CC}.
5845 Alternatively, you can use @code{BImode} if the comparison operator is
5846 specified already in the compare instruction. In this case, you are not
5847 interested in most macros in this section.
5850 * CC0 Condition Codes:: Old style representation of condition codes.
5851 * MODE_CC Condition Codes:: Modern representation of condition codes.
5852 * Cond Exec Macros:: Macros to control conditional execution.
5855 @node CC0 Condition Codes
5856 @subsection Representation of condition codes using @code{(cc0)}
5860 The file @file{conditions.h} defines a variable @code{cc_status} to
5861 describe how the condition code was computed (in case the interpretation of
5862 the condition code depends on the instruction that it was set by). This
5863 variable contains the RTL expressions on which the condition code is
5864 currently based, and several standard flags.
5866 Sometimes additional machine-specific flags must be defined in the machine
5867 description header file. It can also add additional machine-specific
5868 information by defining @code{CC_STATUS_MDEP}.
5870 @defmac CC_STATUS_MDEP
5871 C code for a data type which is used for declaring the @code{mdep}
5872 component of @code{cc_status}. It defaults to @code{int}.
5874 This macro is not used on machines that do not use @code{cc0}.
5877 @defmac CC_STATUS_MDEP_INIT
5878 A C expression to initialize the @code{mdep} field to ``empty''.
5879 The default definition does nothing, since most machines don't use
5880 the field anyway. If you want to use the field, you should probably
5881 define this macro to initialize it.
5883 This macro is not used on machines that do not use @code{cc0}.
5886 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5887 A C compound statement to set the components of @code{cc_status}
5888 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5889 this macro's responsibility to recognize insns that set the condition
5890 code as a byproduct of other activity as well as those that explicitly
5893 This macro is not used on machines that do not use @code{cc0}.
5895 If there are insns that do not set the condition code but do alter
5896 other machine registers, this macro must check to see whether they
5897 invalidate the expressions that the condition code is recorded as
5898 reflecting. For example, on the 68000, insns that store in address
5899 registers do not set the condition code, which means that usually
5900 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5901 insns. But suppose that the previous insn set the condition code
5902 based on location @samp{a4@@(102)} and the current insn stores a new
5903 value in @samp{a4}. Although the condition code is not changed by
5904 this, it will no longer be true that it reflects the contents of
5905 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5906 @code{cc_status} in this case to say that nothing is known about the
5907 condition code value.
5909 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5910 with the results of peephole optimization: insns whose patterns are
5911 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5912 constants which are just the operands. The RTL structure of these
5913 insns is not sufficient to indicate what the insns actually do. What
5914 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5915 @code{CC_STATUS_INIT}.
5917 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5918 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5919 @samp{cc}. This avoids having detailed information about patterns in
5920 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5923 @node MODE_CC Condition Codes
5924 @subsection Representation of condition codes using registers
5928 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5929 On many machines, the condition code may be produced by other instructions
5930 than compares, for example the branch can use directly the condition
5931 code set by a subtract instruction. However, on some machines
5932 when the condition code is set this way some bits (such as the overflow
5933 bit) are not set in the same way as a test instruction, so that a different
5934 branch instruction must be used for some conditional branches. When
5935 this happens, use the machine mode of the condition code register to
5936 record different formats of the condition code register. Modes can
5937 also be used to record which compare instruction (e.g. a signed or an
5938 unsigned comparison) produced the condition codes.
5940 If other modes than @code{CCmode} are required, add them to
5941 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
5942 a mode given an operand of a compare. This is needed because the modes
5943 have to be chosen not only during RTL generation but also, for example,
5944 by instruction combination. The result of @code{SELECT_CC_MODE} should
5945 be consistent with the mode used in the patterns; for example to support
5946 the case of the add on the SPARC discussed above, we have the pattern
5950 [(set (reg:CC_NOOV 0)
5952 (plus:SI (match_operand:SI 0 "register_operand" "%r")
5953 (match_operand:SI 1 "arith_operand" "rI"))
5960 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
5961 for comparisons whose argument is a @code{plus}:
5964 #define SELECT_CC_MODE(OP,X,Y) \
5965 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5966 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5967 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5968 || GET_CODE (X) == NEG) \
5969 ? CC_NOOVmode : CCmode))
5972 Another reason to use modes is to retain information on which operands
5973 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
5976 You should define this macro if and only if you define extra CC modes
5977 in @file{@var{machine}-modes.def}.
5980 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5981 On some machines not all possible comparisons are defined, but you can
5982 convert an invalid comparison into a valid one. For example, the Alpha
5983 does not have a @code{GT} comparison, but you can use an @code{LT}
5984 comparison instead and swap the order of the operands.
5986 On such machines, define this macro to be a C statement to do any
5987 required conversions. @var{code} is the initial comparison code
5988 and @var{op0} and @var{op1} are the left and right operands of the
5989 comparison, respectively. You should modify @var{code}, @var{op0}, and
5990 @var{op1} as required.
5992 GCC will not assume that the comparison resulting from this macro is
5993 valid but will see if the resulting insn matches a pattern in the
5996 You need not define this macro if it would never change the comparison
6000 @defmac REVERSIBLE_CC_MODE (@var{mode})
6001 A C expression whose value is one if it is always safe to reverse a
6002 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6003 can ever return @var{mode} for a floating-point inequality comparison,
6004 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6006 You need not define this macro if it would always returns zero or if the
6007 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6008 For example, here is the definition used on the SPARC, where floating-point
6009 inequality comparisons are always given @code{CCFPEmode}:
6012 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
6016 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6017 A C expression whose value is reversed condition code of the @var{code} for
6018 comparison done in CC_MODE @var{mode}. The macro is used only in case
6019 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6020 machine has some non-standard way how to reverse certain conditionals. For
6021 instance in case all floating point conditions are non-trapping, compiler may
6022 freely convert unordered compares to ordered one. Then definition may look
6026 #define REVERSE_CONDITION(CODE, MODE) \
6027 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6028 : reverse_condition_maybe_unordered (CODE))
6032 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6033 On targets which do not use @code{(cc0)}, and which use a hard
6034 register rather than a pseudo-register to hold condition codes, the
6035 regular CSE passes are often not able to identify cases in which the
6036 hard register is set to a common value. Use this hook to enable a
6037 small pass which optimizes such cases. This hook should return true
6038 to enable this pass, and it should set the integers to which its
6039 arguments point to the hard register numbers used for condition codes.
6040 When there is only one such register, as is true on most systems, the
6041 integer pointed to by @var{p2} should be set to
6042 @code{INVALID_REGNUM}.
6044 The default version of this hook returns false.
6047 @deftypefn {Target Hook} {enum machine_mode} TARGET_CC_MODES_COMPATIBLE (enum machine_mode @var{m1}, enum machine_mode @var{m2})
6048 On targets which use multiple condition code modes in class
6049 @code{MODE_CC}, it is sometimes the case that a comparison can be
6050 validly done in more than one mode. On such a system, define this
6051 target hook to take two mode arguments and to return a mode in which
6052 both comparisons may be validly done. If there is no such mode,
6053 return @code{VOIDmode}.
6055 The default version of this hook checks whether the modes are the
6056 same. If they are, it returns that mode. If they are different, it
6057 returns @code{VOIDmode}.
6060 @node Cond Exec Macros
6061 @subsection Macros to control conditional execution
6062 @findex conditional execution
6065 There is one macro that may need to be defined for targets
6066 supporting conditional execution, independent of how they
6067 represent conditional branches.
6069 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6070 A C expression that returns true if the conditional execution predicate
6071 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6072 versa. Define this to return 0 if the target has conditional execution
6073 predicates that cannot be reversed safely. There is no need to validate
6074 that the arguments of op1 and op2 are the same, this is done separately.
6075 If no expansion is specified, this macro is defined as follows:
6078 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6079 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6084 @section Describing Relative Costs of Operations
6085 @cindex costs of instructions
6086 @cindex relative costs
6087 @cindex speed of instructions
6089 These macros let you describe the relative speed of various operations
6090 on the target machine.
6092 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6093 A C expression for the cost of moving data of mode @var{mode} from a
6094 register in class @var{from} to one in class @var{to}. The classes are
6095 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6096 value of 2 is the default; other values are interpreted relative to
6099 It is not required that the cost always equal 2 when @var{from} is the
6100 same as @var{to}; on some machines it is expensive to move between
6101 registers if they are not general registers.
6103 If reload sees an insn consisting of a single @code{set} between two
6104 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6105 classes returns a value of 2, reload does not check to ensure that the
6106 constraints of the insn are met. Setting a cost of other than 2 will
6107 allow reload to verify that the constraints are met. You should do this
6108 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6110 These macros are obsolete, new ports should use the target hook
6111 @code{TARGET_REGISTER_MOVE_COST} instead.
6114 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6115 This target hook should return the cost of moving data of mode @var{mode}
6116 from a register in class @var{from} to one in class @var{to}. The classes
6117 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6118 A value of 2 is the default; other values are interpreted relative to
6121 It is not required that the cost always equal 2 when @var{from} is the
6122 same as @var{to}; on some machines it is expensive to move between
6123 registers if they are not general registers.
6125 If reload sees an insn consisting of a single @code{set} between two
6126 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6127 classes returns a value of 2, reload does not check to ensure that the
6128 constraints of the insn are met. Setting a cost of other than 2 will
6129 allow reload to verify that the constraints are met. You should do this
6130 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6132 The default version of this function returns 2.
6135 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6136 A C expression for the cost of moving data of mode @var{mode} between a
6137 register of class @var{class} and memory; @var{in} is zero if the value
6138 is to be written to memory, nonzero if it is to be read in. This cost
6139 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6140 registers and memory is more expensive than between two registers, you
6141 should define this macro to express the relative cost.
6143 If you do not define this macro, GCC uses a default cost of 4 plus
6144 the cost of copying via a secondary reload register, if one is
6145 needed. If your machine requires a secondary reload register to copy
6146 between memory and a register of @var{class} but the reload mechanism is
6147 more complex than copying via an intermediate, define this macro to
6148 reflect the actual cost of the move.
6150 GCC defines the function @code{memory_move_secondary_cost} if
6151 secondary reloads are needed. It computes the costs due to copying via
6152 a secondary register. If your machine copies from memory using a
6153 secondary register in the conventional way but the default base value of
6154 4 is not correct for your machine, define this macro to add some other
6155 value to the result of that function. The arguments to that function
6156 are the same as to this macro.
6158 These macros are obsolete, new ports should use the target hook
6159 @code{TARGET_MEMORY_MOVE_COST} instead.
6162 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6163 This target hook should return the cost of moving data of mode @var{mode}
6164 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6165 if the value is to be written to memory, @code{true} if it is to be read in.
6166 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6167 If moving between registers and memory is more expensive than between two
6168 registers, you should add this target hook to express the relative cost.
6170 If you do not add this target hook, GCC uses a default cost of 4 plus
6171 the cost of copying via a secondary reload register, if one is
6172 needed. If your machine requires a secondary reload register to copy
6173 between memory and a register of @var{rclass} but the reload mechanism is
6174 more complex than copying via an intermediate, use this target hook to
6175 reflect the actual cost of the move.
6177 GCC defines the function @code{memory_move_secondary_cost} if
6178 secondary reloads are needed. It computes the costs due to copying via
6179 a secondary register. If your machine copies from memory using a
6180 secondary register in the conventional way but the default base value of
6181 4 is not correct for your machine, use this target hook to add some other
6182 value to the result of that function. The arguments to that function
6183 are the same as to this target hook.
6186 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6187 A C expression for the cost of a branch instruction. A value of 1 is
6188 the default; other values are interpreted relative to that. Parameter
6189 @var{speed_p} is true when the branch in question should be optimized
6190 for speed. When it is false, @code{BRANCH_COST} should return a value
6191 optimal for code size rather than performance. @var{predictable_p} is
6192 true for well-predicted branches. On many architectures the
6193 @code{BRANCH_COST} can be reduced then.
6196 Here are additional macros which do not specify precise relative costs,
6197 but only that certain actions are more expensive than GCC would
6200 @defmac SLOW_BYTE_ACCESS
6201 Define this macro as a C expression which is nonzero if accessing less
6202 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6203 faster than accessing a word of memory, i.e., if such access
6204 require more than one instruction or if there is no difference in cost
6205 between byte and (aligned) word loads.
6207 When this macro is not defined, the compiler will access a field by
6208 finding the smallest containing object; when it is defined, a fullword
6209 load will be used if alignment permits. Unless bytes accesses are
6210 faster than word accesses, using word accesses is preferable since it
6211 may eliminate subsequent memory access if subsequent accesses occur to
6212 other fields in the same word of the structure, but to different bytes.
6215 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6216 Define this macro to be the value 1 if memory accesses described by the
6217 @var{mode} and @var{alignment} parameters have a cost many times greater
6218 than aligned accesses, for example if they are emulated in a trap
6221 When this macro is nonzero, the compiler will act as if
6222 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6223 moves. This can cause significantly more instructions to be produced.
6224 Therefore, do not set this macro nonzero if unaligned accesses only add a
6225 cycle or two to the time for a memory access.
6227 If the value of this macro is always zero, it need not be defined. If
6228 this macro is defined, it should produce a nonzero value when
6229 @code{STRICT_ALIGNMENT} is nonzero.
6232 @defmac MOVE_RATIO (@var{speed})
6233 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6234 which a sequence of insns should be generated instead of a
6235 string move insn or a library call. Increasing the value will always
6236 make code faster, but eventually incurs high cost in increased code size.
6238 Note that on machines where the corresponding move insn is a
6239 @code{define_expand} that emits a sequence of insns, this macro counts
6240 the number of such sequences.
6242 The parameter @var{speed} is true if the code is currently being
6243 optimized for speed rather than size.
6245 If you don't define this, a reasonable default is used.
6248 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6249 A C expression used to determine whether @code{move_by_pieces} will be used to
6250 copy a chunk of memory, or whether some other block move mechanism
6251 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6252 than @code{MOVE_RATIO}.
6255 @defmac MOVE_MAX_PIECES
6256 A C expression used by @code{move_by_pieces} to determine the largest unit
6257 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6260 @defmac CLEAR_RATIO (@var{speed})
6261 The threshold of number of scalar move insns, @emph{below} which a sequence
6262 of insns should be generated to clear memory instead of a string clear insn
6263 or a library call. Increasing the value will always make code faster, but
6264 eventually incurs high cost in increased code size.
6266 The parameter @var{speed} is true if the code is currently being
6267 optimized for speed rather than size.
6269 If you don't define this, a reasonable default is used.
6272 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6273 A C expression used to determine whether @code{clear_by_pieces} will be used
6274 to clear a chunk of memory, or whether some other block clear mechanism
6275 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6276 than @code{CLEAR_RATIO}.
6279 @defmac SET_RATIO (@var{speed})
6280 The threshold of number of scalar move insns, @emph{below} which a sequence
6281 of insns should be generated to set memory to a constant value, instead of
6282 a block set insn or a library call.
6283 Increasing the value will always make code faster, but
6284 eventually incurs high cost in increased code size.
6286 The parameter @var{speed} is true if the code is currently being
6287 optimized for speed rather than size.
6289 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6292 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6293 A C expression used to determine whether @code{store_by_pieces} will be
6294 used to set a chunk of memory to a constant value, or whether some
6295 other mechanism will be used. Used by @code{__builtin_memset} when
6296 storing values other than constant zero.
6297 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6298 than @code{SET_RATIO}.
6301 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6302 A C expression used to determine whether @code{store_by_pieces} will be
6303 used to set a chunk of memory to a constant string value, or whether some
6304 other mechanism will be used. Used by @code{__builtin_strcpy} when
6305 called with a constant source string.
6306 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6307 than @code{MOVE_RATIO}.
6310 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6311 A C expression used to determine whether a load postincrement is a good
6312 thing to use for a given mode. Defaults to the value of
6313 @code{HAVE_POST_INCREMENT}.
6316 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6317 A C expression used to determine whether a load postdecrement is a good
6318 thing to use for a given mode. Defaults to the value of
6319 @code{HAVE_POST_DECREMENT}.
6322 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6323 A C expression used to determine whether a load preincrement is a good
6324 thing to use for a given mode. Defaults to the value of
6325 @code{HAVE_PRE_INCREMENT}.
6328 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6329 A C expression used to determine whether a load predecrement is a good
6330 thing to use for a given mode. Defaults to the value of
6331 @code{HAVE_PRE_DECREMENT}.
6334 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6335 A C expression used to determine whether a store postincrement is a good
6336 thing to use for a given mode. Defaults to the value of
6337 @code{HAVE_POST_INCREMENT}.
6340 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6341 A C expression used to determine whether a store postdecrement is a good
6342 thing to use for a given mode. Defaults to the value of
6343 @code{HAVE_POST_DECREMENT}.
6346 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6347 This macro is used to determine whether a store preincrement is a good
6348 thing to use for a given mode. Defaults to the value of
6349 @code{HAVE_PRE_INCREMENT}.
6352 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6353 This macro is used to determine whether a store predecrement is a good
6354 thing to use for a given mode. Defaults to the value of
6355 @code{HAVE_PRE_DECREMENT}.
6358 @defmac NO_FUNCTION_CSE
6359 Define this macro if it is as good or better to call a constant
6360 function address than to call an address kept in a register.
6363 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6364 Define this macro if a non-short-circuit operation produced by
6365 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6366 @code{BRANCH_COST} is greater than or equal to the value 2.
6369 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total}, bool @var{speed})
6370 This target hook describes the relative costs of RTL expressions.
6372 The cost may depend on the precise form of the expression, which is
6373 available for examination in @var{x}, and the rtx code of the expression
6374 in which it is contained, found in @var{outer_code}. @var{code} is the
6375 expression code---redundant, since it can be obtained with
6376 @code{GET_CODE (@var{x})}.
6378 In implementing this hook, you can use the construct
6379 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6382 On entry to the hook, @code{*@var{total}} contains a default estimate
6383 for the cost of the expression. The hook should modify this value as
6384 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6385 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6386 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6388 When optimizing for code size, i.e.@: when @code{speed} is
6389 false, this target hook should be used to estimate the relative
6390 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6392 The hook returns true when all subexpressions of @var{x} have been
6393 processed, and false when @code{rtx_cost} should recurse.
6396 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, bool @var{speed})
6397 This hook computes the cost of an addressing mode that contains
6398 @var{address}. If not defined, the cost is computed from
6399 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6401 For most CISC machines, the default cost is a good approximation of the
6402 true cost of the addressing mode. However, on RISC machines, all
6403 instructions normally have the same length and execution time. Hence
6404 all addresses will have equal costs.
6406 In cases where more than one form of an address is known, the form with
6407 the lowest cost will be used. If multiple forms have the same, lowest,
6408 cost, the one that is the most complex will be used.
6410 For example, suppose an address that is equal to the sum of a register
6411 and a constant is used twice in the same basic block. When this macro
6412 is not defined, the address will be computed in a register and memory
6413 references will be indirect through that register. On machines where
6414 the cost of the addressing mode containing the sum is no higher than
6415 that of a simple indirect reference, this will produce an additional
6416 instruction and possibly require an additional register. Proper
6417 specification of this macro eliminates this overhead for such machines.
6419 This hook is never called with an invalid address.
6421 On machines where an address involving more than one register is as
6422 cheap as an address computation involving only one register, defining
6423 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6424 be live over a region of code where only one would have been if
6425 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6426 should be considered in the definition of this macro. Equivalent costs
6427 should probably only be given to addresses with different numbers of
6428 registers on machines with lots of registers.
6432 @section Adjusting the Instruction Scheduler
6434 The instruction scheduler may need a fair amount of machine-specific
6435 adjustment in order to produce good code. GCC provides several target
6436 hooks for this purpose. It is usually enough to define just a few of
6437 them: try the first ones in this list first.
6439 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6440 This hook returns the maximum number of instructions that can ever
6441 issue at the same time on the target machine. The default is one.
6442 Although the insn scheduler can define itself the possibility of issue
6443 an insn on the same cycle, the value can serve as an additional
6444 constraint to issue insns on the same simulated processor cycle (see
6445 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6446 This value must be constant over the entire compilation. If you need
6447 it to vary depending on what the instructions are, you must use
6448 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6451 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6452 This hook is executed by the scheduler after it has scheduled an insn
6453 from the ready list. It should return the number of insns which can
6454 still be issued in the current cycle. The default is
6455 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6456 @code{USE}, which normally are not counted against the issue rate.
6457 You should define this hook if some insns take more machine resources
6458 than others, so that fewer insns can follow them in the same cycle.
6459 @var{file} is either a null pointer, or a stdio stream to write any
6460 debug output to. @var{verbose} is the verbose level provided by
6461 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6465 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6466 This function corrects the value of @var{cost} based on the
6467 relationship between @var{insn} and @var{dep_insn} through the
6468 dependence @var{link}. It should return the new value. The default
6469 is to make no adjustment to @var{cost}. This can be used for example
6470 to specify to the scheduler using the traditional pipeline description
6471 that an output- or anti-dependence does not incur the same cost as a
6472 data-dependence. If the scheduler using the automaton based pipeline
6473 description, the cost of anti-dependence is zero and the cost of
6474 output-dependence is maximum of one and the difference of latency
6475 times of the first and the second insns. If these values are not
6476 acceptable, you could use the hook to modify them too. See also
6477 @pxref{Processor pipeline description}.
6480 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6481 This hook adjusts the integer scheduling priority @var{priority} of
6482 @var{insn}. It should return the new priority. Increase the priority to
6483 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6484 later. Do not define this hook if you do not need to adjust the
6485 scheduling priorities of insns.
6488 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6489 This hook is executed by the scheduler after it has scheduled the ready
6490 list, to allow the machine description to reorder it (for example to
6491 combine two small instructions together on @samp{VLIW} machines).
6492 @var{file} is either a null pointer, or a stdio stream to write any
6493 debug output to. @var{verbose} is the verbose level provided by
6494 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6495 list of instructions that are ready to be scheduled. @var{n_readyp} is
6496 a pointer to the number of elements in the ready list. The scheduler
6497 reads the ready list in reverse order, starting with
6498 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6499 is the timer tick of the scheduler. You may modify the ready list and
6500 the number of ready insns. The return value is the number of insns that
6501 can issue this cycle; normally this is just @code{issue_rate}. See also
6502 @samp{TARGET_SCHED_REORDER2}.
6505 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6506 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6507 function is called whenever the scheduler starts a new cycle. This one
6508 is called once per iteration over a cycle, immediately after
6509 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6510 return the number of insns to be scheduled in the same cycle. Defining
6511 this hook can be useful if there are frequent situations where
6512 scheduling one insn causes other insns to become ready in the same
6513 cycle. These other insns can then be taken into account properly.
6516 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6517 This hook is called after evaluation forward dependencies of insns in
6518 chain given by two parameter values (@var{head} and @var{tail}
6519 correspondingly) but before insns scheduling of the insn chain. For
6520 example, it can be used for better insn classification if it requires
6521 analysis of dependencies. This hook can use backward and forward
6522 dependencies of the insn scheduler because they are already
6526 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6527 This hook is executed by the scheduler at the beginning of each block of
6528 instructions that are to be scheduled. @var{file} is either a null
6529 pointer, or a stdio stream to write any debug output to. @var{verbose}
6530 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6531 @var{max_ready} is the maximum number of insns in the current scheduling
6532 region that can be live at the same time. This can be used to allocate
6533 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6536 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6537 This hook is executed by the scheduler at the end of each block of
6538 instructions that are to be scheduled. It can be used to perform
6539 cleanup of any actions done by the other scheduling hooks. @var{file}
6540 is either a null pointer, or a stdio stream to write any debug output
6541 to. @var{verbose} is the verbose level provided by
6542 @option{-fsched-verbose-@var{n}}.
6545 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6546 This hook is executed by the scheduler after function level initializations.
6547 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6548 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6549 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6552 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6553 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6554 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6555 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6558 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6559 The hook returns an RTL insn. The automaton state used in the
6560 pipeline hazard recognizer is changed as if the insn were scheduled
6561 when the new simulated processor cycle starts. Usage of the hook may
6562 simplify the automaton pipeline description for some @acronym{VLIW}
6563 processors. If the hook is defined, it is used only for the automaton
6564 based pipeline description. The default is not to change the state
6565 when the new simulated processor cycle starts.
6568 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6569 The hook can be used to initialize data used by the previous hook.
6572 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6573 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6574 to changed the state as if the insn were scheduled when the new
6575 simulated processor cycle finishes.
6578 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6579 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6580 used to initialize data used by the previous hook.
6583 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
6584 The hook to notify target that the current simulated cycle is about to finish.
6585 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6586 to change the state in more complicated situations - e.g., when advancing
6587 state on a single insn is not enough.
6590 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
6591 The hook to notify target that new simulated cycle has just started.
6592 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6593 to change the state in more complicated situations - e.g., when advancing
6594 state on a single insn is not enough.
6597 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6598 This hook controls better choosing an insn from the ready insn queue
6599 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6600 chooses the first insn from the queue. If the hook returns a positive
6601 value, an additional scheduler code tries all permutations of
6602 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6603 subsequent ready insns to choose an insn whose issue will result in
6604 maximal number of issued insns on the same cycle. For the
6605 @acronym{VLIW} processor, the code could actually solve the problem of
6606 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6607 rules of @acronym{VLIW} packing are described in the automaton.
6609 This code also could be used for superscalar @acronym{RISC}
6610 processors. Let us consider a superscalar @acronym{RISC} processor
6611 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6612 @var{B}, some insns can be executed only in pipelines @var{B} or
6613 @var{C}, and one insn can be executed in pipeline @var{B}. The
6614 processor may issue the 1st insn into @var{A} and the 2nd one into
6615 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6616 until the next cycle. If the scheduler issues the 3rd insn the first,
6617 the processor could issue all 3 insns per cycle.
6619 Actually this code demonstrates advantages of the automaton based
6620 pipeline hazard recognizer. We try quickly and easy many insn
6621 schedules to choose the best one.
6623 The default is no multipass scheduling.
6626 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx @var{insn})
6628 This hook controls what insns from the ready insn queue will be
6629 considered for the multipass insn scheduling. If the hook returns
6630 zero for @var{insn}, the insn will be not chosen to
6633 The default is that any ready insns can be chosen to be issued.
6636 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN (void *@var{data}, char *@var{ready_try}, int @var{n_ready}, bool @var{first_cycle_insn_p})
6637 This hook prepares the target backend for a new round of multipass
6641 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE (void *@var{data}, char *@var{ready_try}, int @var{n_ready}, rtx @var{insn}, const void *@var{prev_data})
6642 This hook is called when multipass scheduling evaluates instruction INSN.
6645 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, char *@var{ready_try}, int @var{n_ready})
6646 This is called when multipass scheduling backtracks from evaluation of
6650 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
6651 This hook notifies the target about the result of the concluded current
6652 round of multipass scheduling.
6655 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
6656 This hook initializes target-specific data used in multipass scheduling.
6659 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
6660 This hook finalizes target-specific data used in multipass scheduling.
6663 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *@var{dump}, int @var{verbose}, rtx @var{insn}, int @var{last_clock}, int @var{clock}, int *@var{sort_p})
6664 This hook is called by the insn scheduler before issuing @var{insn}
6665 on cycle @var{clock}. If the hook returns nonzero,
6666 @var{insn} is not issued on this processor cycle. Instead,
6667 the processor cycle is advanced. If *@var{sort_p}
6668 is zero, the insn ready queue is not sorted on the new cycle
6669 start as usually. @var{dump} and @var{verbose} specify the file and
6670 verbosity level to use for debugging output.
6671 @var{last_clock} and @var{clock} are, respectively, the
6672 processor cycle on which the previous insn has been issued,
6673 and the current processor cycle.
6676 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
6677 This hook is used to define which dependences are considered costly by
6678 the target, so costly that it is not advisable to schedule the insns that
6679 are involved in the dependence too close to one another. The parameters
6680 to this hook are as follows: The first parameter @var{_dep} is the dependence
6681 being evaluated. The second parameter @var{cost} is the cost of the
6682 dependence as estimated by the scheduler, and the third
6683 parameter @var{distance} is the distance in cycles between the two insns.
6684 The hook returns @code{true} if considering the distance between the two
6685 insns the dependence between them is considered costly by the target,
6686 and @code{false} otherwise.
6688 Defining this hook can be useful in multiple-issue out-of-order machines,
6689 where (a) it's practically hopeless to predict the actual data/resource
6690 delays, however: (b) there's a better chance to predict the actual grouping
6691 that will be formed, and (c) correctly emulating the grouping can be very
6692 important. In such targets one may want to allow issuing dependent insns
6693 closer to one another---i.e., closer than the dependence distance; however,
6694 not in cases of ``costly dependences'', which this hooks allows to define.
6697 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6698 This hook is called by the insn scheduler after emitting a new instruction to
6699 the instruction stream. The hook notifies a target backend to extend its
6700 per instruction data structures.
6703 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6704 Return a pointer to a store large enough to hold target scheduling context.
6707 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6708 Initialize store pointed to by @var{tc} to hold target scheduling context.
6709 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6710 beginning of the block. Otherwise, copy the current context into @var{tc}.
6713 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6714 Copy target scheduling context pointed to by @var{tc} to the current context.
6717 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6718 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6721 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6722 Deallocate a store for target scheduling context pointed to by @var{tc}.
6725 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6726 This hook is called by the insn scheduler when @var{insn} has only
6727 speculative dependencies and therefore can be scheduled speculatively.
6728 The hook is used to check if the pattern of @var{insn} has a speculative
6729 version and, in case of successful check, to generate that speculative
6730 pattern. The hook should return 1, if the instruction has a speculative form,
6731 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6732 speculation. If the return value equals 1 then @var{new_pat} is assigned
6733 the generated speculative pattern.
6736 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (int @var{dep_status})
6737 This hook is called by the insn scheduler during generation of recovery code
6738 for @var{insn}. It should return @code{true}, if the corresponding check
6739 instruction should branch to recovery code, or @code{false} otherwise.
6742 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6743 This hook is called by the insn scheduler to generate a pattern for recovery
6744 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6745 speculative instruction for which the check should be generated.
6746 @var{label} is either a label of a basic block, where recovery code should
6747 be emitted, or a null pointer, when requested check doesn't branch to
6748 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6749 a pattern for a branchy check corresponding to a simple check denoted by
6750 @var{insn} should be generated. In this case @var{label} can't be null.
6753 @deftypefn {Target Hook} bool TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (const_rtx @var{insn})
6754 This hook is used as a workaround for
6755 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6756 called on the first instruction of the ready list. The hook is used to
6757 discard speculative instructions that stand first in the ready list from
6758 being scheduled on the current cycle. If the hook returns @code{false},
6759 @var{insn} will not be chosen to be issued.
6760 For non-speculative instructions,
6761 the hook should always return @code{true}. For example, in the ia64 backend
6762 the hook is used to cancel data speculative insns when the ALAT table
6766 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
6767 This hook is used by the insn scheduler to find out what features should be
6769 The structure *@var{spec_info} should be filled in by the target.
6770 The structure describes speculation types that can be used in the scheduler.
6773 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6774 This hook is called by the swing modulo scheduler to calculate a
6775 resource-based lower bound which is based on the resources available in
6776 the machine and the resources required by each instruction. The target
6777 backend can use @var{g} to calculate such bound. A very simple lower
6778 bound will be used in case this hook is not implemented: the total number
6779 of instructions divided by the issue rate.
6782 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx @var{insn}, int @var{x})
6783 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6784 is supported in hardware and the condition specified in the parameter is true.
6787 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx @var{insn}, int @var{x})
6788 This hook is called by Haifa Scheduler. It performs the operation specified
6789 in its second parameter.
6792 @deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE
6793 True if the processor has an exposed pipeline, which means that not just
6794 the order of instructions is important for correctness when scheduling, but
6795 also the latencies of operations.
6799 @section Dividing the Output into Sections (Texts, Data, @dots{})
6800 @c the above section title is WAY too long. maybe cut the part between
6801 @c the (...)? --mew 10feb93
6803 An object file is divided into sections containing different types of
6804 data. In the most common case, there are three sections: the @dfn{text
6805 section}, which holds instructions and read-only data; the @dfn{data
6806 section}, which holds initialized writable data; and the @dfn{bss
6807 section}, which holds uninitialized data. Some systems have other kinds
6810 @file{varasm.c} provides several well-known sections, such as
6811 @code{text_section}, @code{data_section} and @code{bss_section}.
6812 The normal way of controlling a @code{@var{foo}_section} variable
6813 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6814 as described below. The macros are only read once, when @file{varasm.c}
6815 initializes itself, so their values must be run-time constants.
6816 They may however depend on command-line flags.
6818 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6819 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6820 to be string literals.
6822 Some assemblers require a different string to be written every time a
6823 section is selected. If your assembler falls into this category, you
6824 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6825 @code{get_unnamed_section} to set up the sections.
6827 You must always create a @code{text_section}, either by defining
6828 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6829 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6830 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6831 create a distinct @code{readonly_data_section}, the default is to
6832 reuse @code{text_section}.
6834 All the other @file{varasm.c} sections are optional, and are null
6835 if the target does not provide them.
6837 @defmac TEXT_SECTION_ASM_OP
6838 A C expression whose value is a string, including spacing, containing the
6839 assembler operation that should precede instructions and read-only data.
6840 Normally @code{"\t.text"} is right.
6843 @defmac HOT_TEXT_SECTION_NAME
6844 If defined, a C string constant for the name of the section containing most
6845 frequently executed functions of the program. If not defined, GCC will provide
6846 a default definition if the target supports named sections.
6849 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6850 If defined, a C string constant for the name of the section containing unlikely
6851 executed functions in the program.
6854 @defmac DATA_SECTION_ASM_OP
6855 A C expression whose value is a string, including spacing, containing the
6856 assembler operation to identify the following data as writable initialized
6857 data. Normally @code{"\t.data"} is right.
6860 @defmac SDATA_SECTION_ASM_OP
6861 If defined, a C expression whose value is a string, including spacing,
6862 containing the assembler operation to identify the following data as
6863 initialized, writable small data.
6866 @defmac READONLY_DATA_SECTION_ASM_OP
6867 A C expression whose value is a string, including spacing, containing the
6868 assembler operation to identify the following data as read-only initialized
6872 @defmac BSS_SECTION_ASM_OP
6873 If defined, a C expression whose value is a string, including spacing,
6874 containing the assembler operation to identify the following data as
6875 uninitialized global data. If not defined, and
6876 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
6877 uninitialized global data will be output in the data section if
6878 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6882 @defmac SBSS_SECTION_ASM_OP
6883 If defined, a C expression whose value is a string, including spacing,
6884 containing the assembler operation to identify the following data as
6885 uninitialized, writable small data.
6888 @defmac TLS_COMMON_ASM_OP
6889 If defined, a C expression whose value is a string containing the
6890 assembler operation to identify the following data as thread-local
6891 common data. The default is @code{".tls_common"}.
6894 @defmac TLS_SECTION_ASM_FLAG
6895 If defined, a C expression whose value is a character constant
6896 containing the flag used to mark a section as a TLS section. The
6897 default is @code{'T'}.
6900 @defmac INIT_SECTION_ASM_OP
6901 If defined, a C expression whose value is a string, including spacing,
6902 containing the assembler operation to identify the following data as
6903 initialization code. If not defined, GCC will assume such a section does
6904 not exist. This section has no corresponding @code{init_section}
6905 variable; it is used entirely in runtime code.
6908 @defmac FINI_SECTION_ASM_OP
6909 If defined, a C expression whose value is a string, including spacing,
6910 containing the assembler operation to identify the following data as
6911 finalization code. If not defined, GCC will assume such a section does
6912 not exist. This section has no corresponding @code{fini_section}
6913 variable; it is used entirely in runtime code.
6916 @defmac INIT_ARRAY_SECTION_ASM_OP
6917 If defined, a C expression whose value is a string, including spacing,
6918 containing the assembler operation to identify the following data as
6919 part of the @code{.init_array} (or equivalent) section. If not
6920 defined, GCC will assume such a section does not exist. Do not define
6921 both this macro and @code{INIT_SECTION_ASM_OP}.
6924 @defmac FINI_ARRAY_SECTION_ASM_OP
6925 If defined, a C expression whose value is a string, including spacing,
6926 containing the assembler operation to identify the following data as
6927 part of the @code{.fini_array} (or equivalent) section. If not
6928 defined, GCC will assume such a section does not exist. Do not define
6929 both this macro and @code{FINI_SECTION_ASM_OP}.
6932 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6933 If defined, an ASM statement that switches to a different section
6934 via @var{section_op}, calls @var{function}, and switches back to
6935 the text section. This is used in @file{crtstuff.c} if
6936 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6937 to initialization and finalization functions from the init and fini
6938 sections. By default, this macro uses a simple function call. Some
6939 ports need hand-crafted assembly code to avoid dependencies on
6940 registers initialized in the function prologue or to ensure that
6941 constant pools don't end up too far way in the text section.
6944 @defmac TARGET_LIBGCC_SDATA_SECTION
6945 If defined, a string which names the section into which small
6946 variables defined in crtstuff and libgcc should go. This is useful
6947 when the target has options for optimizing access to small data, and
6948 you want the crtstuff and libgcc routines to be conservative in what
6949 they expect of your application yet liberal in what your application
6950 expects. For example, for targets with a @code{.sdata} section (like
6951 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6952 require small data support from your application, but use this macro
6953 to put small data into @code{.sdata} so that your application can
6954 access these variables whether it uses small data or not.
6957 @defmac FORCE_CODE_SECTION_ALIGN
6958 If defined, an ASM statement that aligns a code section to some
6959 arbitrary boundary. This is used to force all fragments of the
6960 @code{.init} and @code{.fini} sections to have to same alignment
6961 and thus prevent the linker from having to add any padding.
6964 @defmac JUMP_TABLES_IN_TEXT_SECTION
6965 Define this macro to be an expression with a nonzero value if jump
6966 tables (for @code{tablejump} insns) should be output in the text
6967 section, along with the assembler instructions. Otherwise, the
6968 readonly data section is used.
6970 This macro is irrelevant if there is no separate readonly data section.
6973 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6974 Define this hook if you need to do something special to set up the
6975 @file{varasm.c} sections, or if your target has some special sections
6976 of its own that you need to create.
6978 GCC calls this hook after processing the command line, but before writing
6979 any assembly code, and before calling any of the section-returning hooks
6983 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
6984 Return a mask describing how relocations should be treated when
6985 selecting sections. Bit 1 should be set if global relocations
6986 should be placed in a read-write section; bit 0 should be set if
6987 local relocations should be placed in a read-write section.
6989 The default version of this function returns 3 when @option{-fpic}
6990 is in effect, and 0 otherwise. The hook is typically redefined
6991 when the target cannot support (some kinds of) dynamic relocations
6992 in read-only sections even in executables.
6995 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6996 Return the section into which @var{exp} should be placed. You can
6997 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6998 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6999 requires link-time relocations. Bit 0 is set when variable contains
7000 local relocations only, while bit 1 is set for global relocations.
7001 @var{align} is the constant alignment in bits.
7003 The default version of this function takes care of putting read-only
7004 variables in @code{readonly_data_section}.
7006 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7009 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7010 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7011 for @code{FUNCTION_DECL}s as well as for variables and constants.
7013 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7014 function has been determined to be likely to be called, and nonzero if
7015 it is unlikely to be called.
7018 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7019 Build up a unique section name, expressed as a @code{STRING_CST} node,
7020 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7021 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7022 the initial value of @var{exp} requires link-time relocations.
7024 The default version of this function appends the symbol name to the
7025 ELF section name that would normally be used for the symbol. For
7026 example, the function @code{foo} would be placed in @code{.text.foo}.
7027 Whatever the actual target object format, this is often good enough.
7030 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7031 Return the readonly data section associated with
7032 @samp{DECL_SECTION_NAME (@var{decl})}.
7033 The default version of this function selects @code{.gnu.linkonce.r.name} if
7034 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7035 if function is in @code{.text.name}, and the normal readonly-data section
7039 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7040 Return the section into which a constant @var{x}, of mode @var{mode},
7041 should be placed. You can assume that @var{x} is some kind of
7042 constant in RTL@. The argument @var{mode} is redundant except in the
7043 case of a @code{const_int} rtx. @var{align} is the constant alignment
7046 The default version of this function takes care of putting symbolic
7047 constants in @code{flag_pic} mode in @code{data_section} and everything
7048 else in @code{readonly_data_section}.
7051 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7052 Define this hook if you need to postprocess the assembler name generated
7053 by target-independent code. The @var{id} provided to this hook will be
7054 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7055 or the mangled name of the @var{decl} in C++). The return value of the
7056 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7057 your target system. The default implementation of this hook just
7058 returns the @var{id} provided.
7061 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7062 Define this hook if references to a symbol or a constant must be
7063 treated differently depending on something about the variable or
7064 function named by the symbol (such as what section it is in).
7066 The hook is executed immediately after rtl has been created for
7067 @var{decl}, which may be a variable or function declaration or
7068 an entry in the constant pool. In either case, @var{rtl} is the
7069 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7070 in this hook; that field may not have been initialized yet.
7072 In the case of a constant, it is safe to assume that the rtl is
7073 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7074 will also have this form, but that is not guaranteed. Global
7075 register variables, for instance, will have a @code{reg} for their
7076 rtl. (Normally the right thing to do with such unusual rtl is
7079 The @var{new_decl_p} argument will be true if this is the first time
7080 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7081 be false for subsequent invocations, which will happen for duplicate
7082 declarations. Whether or not anything must be done for the duplicate
7083 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7084 @var{new_decl_p} is always true when the hook is called for a constant.
7086 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7087 The usual thing for this hook to do is to record flags in the
7088 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7089 Historically, the name string was modified if it was necessary to
7090 encode more than one bit of information, but this practice is now
7091 discouraged; use @code{SYMBOL_REF_FLAGS}.
7093 The default definition of this hook, @code{default_encode_section_info}
7094 in @file{varasm.c}, sets a number of commonly-useful bits in
7095 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7096 before overriding it.
7099 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7100 Decode @var{name} and return the real name part, sans
7101 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7105 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7106 Returns true if @var{exp} should be placed into a ``small data'' section.
7107 The default version of this hook always returns false.
7110 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7111 Contains the value true if the target places read-only
7112 ``small data'' into a separate section. The default value is false.
7115 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7116 It returns true if target wants profile code emitted before prologue.
7118 The default version of this hook use the target macro
7119 @code{PROFILE_BEFORE_PROLOGUE}.
7122 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7123 Returns true if @var{exp} names an object for which name resolution
7124 rules must resolve to the current ``module'' (dynamic shared library
7125 or executable image).
7127 The default version of this hook implements the name resolution rules
7128 for ELF, which has a looser model of global name binding than other
7129 currently supported object file formats.
7132 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7133 Contains the value true if the target supports thread-local storage.
7134 The default value is false.
7139 @section Position Independent Code
7140 @cindex position independent code
7143 This section describes macros that help implement generation of position
7144 independent code. Simply defining these macros is not enough to
7145 generate valid PIC; you must also add support to the hook
7146 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7147 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7148 must modify the definition of @samp{movsi} to do something appropriate
7149 when the source operand contains a symbolic address. You may also
7150 need to alter the handling of switch statements so that they use
7152 @c i rearranged the order of the macros above to try to force one of
7153 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7155 @defmac PIC_OFFSET_TABLE_REGNUM
7156 The register number of the register used to address a table of static
7157 data addresses in memory. In some cases this register is defined by a
7158 processor's ``application binary interface'' (ABI)@. When this macro
7159 is defined, RTL is generated for this register once, as with the stack
7160 pointer and frame pointer registers. If this macro is not defined, it
7161 is up to the machine-dependent files to allocate such a register (if
7162 necessary). Note that this register must be fixed when in use (e.g.@:
7163 when @code{flag_pic} is true).
7166 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7167 A C expression that is nonzero if the register defined by
7168 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7169 the default is zero. Do not define
7170 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7173 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7174 A C expression that is nonzero if @var{x} is a legitimate immediate
7175 operand on the target machine when generating position independent code.
7176 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7177 check this. You can also assume @var{flag_pic} is true, so you need not
7178 check it either. You need not define this macro if all constants
7179 (including @code{SYMBOL_REF}) can be immediate operands when generating
7180 position independent code.
7183 @node Assembler Format
7184 @section Defining the Output Assembler Language
7186 This section describes macros whose principal purpose is to describe how
7187 to write instructions in assembler language---rather than what the
7191 * File Framework:: Structural information for the assembler file.
7192 * Data Output:: Output of constants (numbers, strings, addresses).
7193 * Uninitialized Data:: Output of uninitialized variables.
7194 * Label Output:: Output and generation of labels.
7195 * Initialization:: General principles of initialization
7196 and termination routines.
7197 * Macros for Initialization::
7198 Specific macros that control the handling of
7199 initialization and termination routines.
7200 * Instruction Output:: Output of actual instructions.
7201 * Dispatch Tables:: Output of jump tables.
7202 * Exception Region Output:: Output of exception region code.
7203 * Alignment Output:: Pseudo ops for alignment and skipping data.
7206 @node File Framework
7207 @subsection The Overall Framework of an Assembler File
7208 @cindex assembler format
7209 @cindex output of assembler code
7211 @c prevent bad page break with this line
7212 This describes the overall framework of an assembly file.
7214 @findex default_file_start
7215 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7216 Output to @code{asm_out_file} any text which the assembler expects to
7217 find at the beginning of a file. The default behavior is controlled
7218 by two flags, documented below. Unless your target's assembler is
7219 quite unusual, if you override the default, you should call
7220 @code{default_file_start} at some point in your target hook. This
7221 lets other target files rely on these variables.
7224 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7225 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7226 printed as the very first line in the assembly file, unless
7227 @option{-fverbose-asm} is in effect. (If that macro has been defined
7228 to the empty string, this variable has no effect.) With the normal
7229 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7230 assembler that it need not bother stripping comments or extra
7231 whitespace from its input. This allows it to work a bit faster.
7233 The default is false. You should not set it to true unless you have
7234 verified that your port does not generate any extra whitespace or
7235 comments that will cause GAS to issue errors in NO_APP mode.
7238 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7239 If this flag is true, @code{output_file_directive} will be called
7240 for the primary source file, immediately after printing
7241 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7242 this to be done. The default is false.
7245 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7246 Output to @code{asm_out_file} any text which the assembler expects
7247 to find at the end of a file. The default is to output nothing.
7250 @deftypefun void file_end_indicate_exec_stack ()
7251 Some systems use a common convention, the @samp{.note.GNU-stack}
7252 special section, to indicate whether or not an object file relies on
7253 the stack being executable. If your system uses this convention, you
7254 should define @code{TARGET_ASM_FILE_END} to this function. If you
7255 need to do other things in that hook, have your hook function call
7259 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7260 Output to @code{asm_out_file} any text which the assembler expects
7261 to find at the start of an LTO section. The default is to output
7265 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7266 Output to @code{asm_out_file} any text which the assembler expects
7267 to find at the end of an LTO section. The default is to output
7271 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7272 Output to @code{asm_out_file} any text which is needed before emitting
7273 unwind info and debug info at the end of a file. Some targets emit
7274 here PIC setup thunks that cannot be emitted at the end of file,
7275 because they couldn't have unwind info then. The default is to output
7279 @defmac ASM_COMMENT_START
7280 A C string constant describing how to begin a comment in the target
7281 assembler language. The compiler assumes that the comment will end at
7282 the end of the line.
7286 A C string constant for text to be output before each @code{asm}
7287 statement or group of consecutive ones. Normally this is
7288 @code{"#APP"}, which is a comment that has no effect on most
7289 assemblers but tells the GNU assembler that it must check the lines
7290 that follow for all valid assembler constructs.
7294 A C string constant for text to be output after each @code{asm}
7295 statement or group of consecutive ones. Normally this is
7296 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7297 time-saving assumptions that are valid for ordinary compiler output.
7300 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7301 A C statement to output COFF information or DWARF debugging information
7302 which indicates that filename @var{name} is the current source file to
7303 the stdio stream @var{stream}.
7305 This macro need not be defined if the standard form of output
7306 for the file format in use is appropriate.
7309 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7310 Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7312 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7315 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7316 A C statement to output the string @var{string} to the stdio stream
7317 @var{stream}. If you do not call the function @code{output_quoted_string}
7318 in your config files, GCC will only call it to output filenames to
7319 the assembler source. So you can use it to canonicalize the format
7320 of the filename using this macro.
7323 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7324 A C statement to output something to the assembler file to handle a
7325 @samp{#ident} directive containing the text @var{string}. If this
7326 macro is not defined, nothing is output for a @samp{#ident} directive.
7329 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7330 Output assembly directives to switch to section @var{name}. The section
7331 should have attributes as specified by @var{flags}, which is a bit mask
7332 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7333 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7334 this section is associated.
7337 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
7338 Return preferred text (sub)section for function @var{decl}.
7339 Main purpose of this function is to separate cold, normal and hot
7340 functions. @var{startup} is true when function is known to be used only
7341 at startup (from static constructors or it is @code{main()}).
7342 @var{exit} is true when function is known to be used only at exit
7343 (from static destructors).
7344 Return NULL if function should go to default text section.
7347 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
7348 Used by the target to emit any assembler directives or additional labels needed when a function is partitioned between different sections. Output should be written to @var{file}. The function decl is available as @var{decl} and the new section is `cold' if @var{new_is_cold} is @code{true}.
7351 @deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7352 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7353 It must not be modified by command-line option processing.
7356 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7357 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7358 This flag is true if we can create zeroed data by switching to a BSS
7359 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7360 This is true on most ELF targets.
7363 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7364 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7365 based on a variable or function decl, a section name, and whether or not the
7366 declaration's initializer may contain runtime relocations. @var{decl} may be
7367 null, in which case read-write data should be assumed.
7369 The default version of this function handles choosing code vs data,
7370 read-only vs read-write data, and @code{flag_pic}. You should only
7371 need to override this if your target has special flags that might be
7372 set via @code{__attribute__}.
7375 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
7376 Provides the target with the ability to record the gcc command line
7377 switches that have been passed to the compiler, and options that are
7378 enabled. The @var{type} argument specifies what is being recorded.
7379 It can take the following values:
7382 @item SWITCH_TYPE_PASSED
7383 @var{text} is a command line switch that has been set by the user.
7385 @item SWITCH_TYPE_ENABLED
7386 @var{text} is an option which has been enabled. This might be as a
7387 direct result of a command line switch, or because it is enabled by
7388 default or because it has been enabled as a side effect of a different
7389 command line switch. For example, the @option{-O2} switch enables
7390 various different individual optimization passes.
7392 @item SWITCH_TYPE_DESCRIPTIVE
7393 @var{text} is either NULL or some descriptive text which should be
7394 ignored. If @var{text} is NULL then it is being used to warn the
7395 target hook that either recording is starting or ending. The first
7396 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7397 warning is for start up and the second time the warning is for
7398 wind down. This feature is to allow the target hook to make any
7399 necessary preparations before it starts to record switches and to
7400 perform any necessary tidying up after it has finished recording
7403 @item SWITCH_TYPE_LINE_START
7404 This option can be ignored by this target hook.
7406 @item SWITCH_TYPE_LINE_END
7407 This option can be ignored by this target hook.
7410 The hook's return value must be zero. Other return values may be
7411 supported in the future.
7413 By default this hook is set to NULL, but an example implementation is
7414 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7415 it records the switches as ASCII text inside a new, string mergeable
7416 section in the assembler output file. The name of the new section is
7417 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7421 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7422 This is the name of the section that will be created by the example
7423 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7429 @subsection Output of Data
7432 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7433 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7434 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7435 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7436 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7437 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7438 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7439 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7440 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7441 These hooks specify assembly directives for creating certain kinds
7442 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7443 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7444 aligned two-byte object, and so on. Any of the hooks may be
7445 @code{NULL}, indicating that no suitable directive is available.
7447 The compiler will print these strings at the start of a new line,
7448 followed immediately by the object's initial value. In most cases,
7449 the string should contain a tab, a pseudo-op, and then another tab.
7452 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7453 The @code{assemble_integer} function uses this hook to output an
7454 integer object. @var{x} is the object's value, @var{size} is its size
7455 in bytes and @var{aligned_p} indicates whether it is aligned. The
7456 function should return @code{true} if it was able to output the
7457 object. If it returns false, @code{assemble_integer} will try to
7458 split the object into smaller parts.
7460 The default implementation of this hook will use the
7461 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7462 when the relevant string is @code{NULL}.
7465 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
7466 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7467 can't deal with, and output assembly code to @var{file} corresponding to
7468 the pattern @var{x}. This may be used to allow machine-dependent
7469 @code{UNSPEC}s to appear within constants.
7471 If target hook fails to recognize a pattern, it must return @code{false},
7472 so that a standard error message is printed. If it prints an error message
7473 itself, by calling, for example, @code{output_operand_lossage}, it may just
7477 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
7478 A C statement to recognize @var{rtx} patterns that
7479 @code{output_addr_const} can't deal with, and output assembly code to
7480 @var{stream} corresponding to the pattern @var{x}. This may be used to
7481 allow machine-dependent @code{UNSPEC}s to appear within constants.
7483 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
7484 @code{goto fail}, so that a standard error message is printed. If it
7485 prints an error message itself, by calling, for example,
7486 @code{output_operand_lossage}, it may just complete normally.
7489 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7490 A C statement to output to the stdio stream @var{stream} an assembler
7491 instruction to assemble a string constant containing the @var{len}
7492 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7493 @code{char *} and @var{len} a C expression of type @code{int}.
7495 If the assembler has a @code{.ascii} pseudo-op as found in the
7496 Berkeley Unix assembler, do not define the macro
7497 @code{ASM_OUTPUT_ASCII}.
7500 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7501 A C statement to output word @var{n} of a function descriptor for
7502 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7503 is defined, and is otherwise unused.
7506 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7507 You may define this macro as a C expression. You should define the
7508 expression to have a nonzero value if GCC should output the constant
7509 pool for a function before the code for the function, or a zero value if
7510 GCC should output the constant pool after the function. If you do
7511 not define this macro, the usual case, GCC will output the constant
7512 pool before the function.
7515 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7516 A C statement to output assembler commands to define the start of the
7517 constant pool for a function. @var{funname} is a string giving
7518 the name of the function. Should the return type of the function
7519 be required, it can be obtained via @var{fundecl}. @var{size}
7520 is the size, in bytes, of the constant pool that will be written
7521 immediately after this call.
7523 If no constant-pool prefix is required, the usual case, this macro need
7527 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7528 A C statement (with or without semicolon) to output a constant in the
7529 constant pool, if it needs special treatment. (This macro need not do
7530 anything for RTL expressions that can be output normally.)
7532 The argument @var{file} is the standard I/O stream to output the
7533 assembler code on. @var{x} is the RTL expression for the constant to
7534 output, and @var{mode} is the machine mode (in case @var{x} is a
7535 @samp{const_int}). @var{align} is the required alignment for the value
7536 @var{x}; you should output an assembler directive to force this much
7539 The argument @var{labelno} is a number to use in an internal label for
7540 the address of this pool entry. The definition of this macro is
7541 responsible for outputting the label definition at the proper place.
7542 Here is how to do this:
7545 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7548 When you output a pool entry specially, you should end with a
7549 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7550 entry from being output a second time in the usual manner.
7552 You need not define this macro if it would do nothing.
7555 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7556 A C statement to output assembler commands to at the end of the constant
7557 pool for a function. @var{funname} is a string giving the name of the
7558 function. Should the return type of the function be required, you can
7559 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7560 constant pool that GCC wrote immediately before this call.
7562 If no constant-pool epilogue is required, the usual case, you need not
7566 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7567 Define this macro as a C expression which is nonzero if @var{C} is
7568 used as a logical line separator by the assembler. @var{STR} points
7569 to the position in the string where @var{C} was found; this can be used if
7570 a line separator uses multiple characters.
7572 If you do not define this macro, the default is that only
7573 the character @samp{;} is treated as a logical line separator.
7576 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7577 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7578 These target hooks are C string constants, describing the syntax in the
7579 assembler for grouping arithmetic expressions. If not overridden, they
7580 default to normal parentheses, which is correct for most assemblers.
7583 These macros are provided by @file{real.h} for writing the definitions
7584 of @code{ASM_OUTPUT_DOUBLE} and the like:
7586 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7587 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7588 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7589 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7590 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7591 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7592 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7593 target's floating point representation, and store its bit pattern in
7594 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7595 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7596 simple @code{long int}. For the others, it should be an array of
7597 @code{long int}. The number of elements in this array is determined
7598 by the size of the desired target floating point data type: 32 bits of
7599 it go in each @code{long int} array element. Each array element holds
7600 32 bits of the result, even if @code{long int} is wider than 32 bits
7601 on the host machine.
7603 The array element values are designed so that you can print them out
7604 using @code{fprintf} in the order they should appear in the target
7608 @node Uninitialized Data
7609 @subsection Output of Uninitialized Variables
7611 Each of the macros in this section is used to do the whole job of
7612 outputting a single uninitialized variable.
7614 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7615 A C statement (sans semicolon) to output to the stdio stream
7616 @var{stream} the assembler definition of a common-label named
7617 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7618 is the size rounded up to whatever alignment the caller wants. It is
7619 possible that @var{size} may be zero, for instance if a struct with no
7620 other member than a zero-length array is defined. In this case, the
7621 backend must output a symbol definition that allocates at least one
7622 byte, both so that the address of the resulting object does not compare
7623 equal to any other, and because some object formats cannot even express
7624 the concept of a zero-sized common symbol, as that is how they represent
7625 an ordinary undefined external.
7627 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7628 output the name itself; before and after that, output the additional
7629 assembler syntax for defining the name, and a newline.
7631 This macro controls how the assembler definitions of uninitialized
7632 common global variables are output.
7635 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7636 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7637 separate, explicit argument. If you define this macro, it is used in
7638 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7639 handling the required alignment of the variable. The alignment is specified
7640 as the number of bits.
7643 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7644 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7645 variable to be output, if there is one, or @code{NULL_TREE} if there
7646 is no corresponding variable. If you define this macro, GCC will use it
7647 in place of both @code{ASM_OUTPUT_COMMON} and
7648 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7649 the variable's decl in order to chose what to output.
7652 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7653 A C statement (sans semicolon) to output to the stdio stream
7654 @var{stream} the assembler definition of uninitialized global @var{decl} named
7655 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
7656 is the alignment specified as the number of bits.
7658 Try to use function @code{asm_output_aligned_bss} defined in file
7659 @file{varasm.c} when defining this macro. If unable, use the expression
7660 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7661 before and after that, output the additional assembler syntax for defining
7662 the name, and a newline.
7664 There are two ways of handling global BSS@. One is to define this macro.
7665 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7666 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7667 You do not need to do both.
7669 Some languages do not have @code{common} data, and require a
7670 non-common form of global BSS in order to handle uninitialized globals
7671 efficiently. C++ is one example of this. However, if the target does
7672 not support global BSS, the front end may choose to make globals
7673 common in order to save space in the object file.
7676 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7677 A C statement (sans semicolon) to output to the stdio stream
7678 @var{stream} the assembler definition of a local-common-label named
7679 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7680 is the size rounded up to whatever alignment the caller wants.
7682 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7683 output the name itself; before and after that, output the additional
7684 assembler syntax for defining the name, and a newline.
7686 This macro controls how the assembler definitions of uninitialized
7687 static variables are output.
7690 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7691 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7692 separate, explicit argument. If you define this macro, it is used in
7693 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7694 handling the required alignment of the variable. The alignment is specified
7695 as the number of bits.
7698 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7699 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7700 variable to be output, if there is one, or @code{NULL_TREE} if there
7701 is no corresponding variable. If you define this macro, GCC will use it
7702 in place of both @code{ASM_OUTPUT_DECL} and
7703 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7704 the variable's decl in order to chose what to output.
7708 @subsection Output and Generation of Labels
7710 @c prevent bad page break with this line
7711 This is about outputting labels.
7713 @findex assemble_name
7714 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7715 A C statement (sans semicolon) to output to the stdio stream
7716 @var{stream} the assembler definition of a label named @var{name}.
7717 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7718 output the name itself; before and after that, output the additional
7719 assembler syntax for defining the name, and a newline. A default
7720 definition of this macro is provided which is correct for most systems.
7723 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7724 A C statement (sans semicolon) to output to the stdio stream
7725 @var{stream} the assembler definition of a label named @var{name} of
7727 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7728 output the name itself; before and after that, output the additional
7729 assembler syntax for defining the name, and a newline. A default
7730 definition of this macro is provided which is correct for most systems.
7732 If this macro is not defined, then the function name is defined in the
7733 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7736 @findex assemble_name_raw
7737 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7738 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7739 to refer to a compiler-generated label. The default definition uses
7740 @code{assemble_name_raw}, which is like @code{assemble_name} except
7741 that it is more efficient.
7745 A C string containing the appropriate assembler directive to specify the
7746 size of a symbol, without any arguments. On systems that use ELF, the
7747 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7748 systems, the default is not to define this macro.
7750 Define this macro only if it is correct to use the default definitions
7751 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7752 for your system. If you need your own custom definitions of those
7753 macros, or if you do not need explicit symbol sizes at all, do not
7757 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7758 A C statement (sans semicolon) to output to the stdio stream
7759 @var{stream} a directive telling the assembler that the size of the
7760 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7761 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7765 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7766 A C statement (sans semicolon) to output to the stdio stream
7767 @var{stream} a directive telling the assembler to calculate the size of
7768 the symbol @var{name} by subtracting its address from the current
7771 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7772 provided. The default assumes that the assembler recognizes a special
7773 @samp{.} symbol as referring to the current address, and can calculate
7774 the difference between this and another symbol. If your assembler does
7775 not recognize @samp{.} or cannot do calculations with it, you will need
7776 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7780 A C string containing the appropriate assembler directive to specify the
7781 type of a symbol, without any arguments. On systems that use ELF, the
7782 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7783 systems, the default is not to define this macro.
7785 Define this macro only if it is correct to use the default definition of
7786 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7787 custom definition of this macro, or if you do not need explicit symbol
7788 types at all, do not define this macro.
7791 @defmac TYPE_OPERAND_FMT
7792 A C string which specifies (using @code{printf} syntax) the format of
7793 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7794 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7795 the default is not to define this macro.
7797 Define this macro only if it is correct to use the default definition of
7798 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7799 custom definition of this macro, or if you do not need explicit symbol
7800 types at all, do not define this macro.
7803 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7804 A C statement (sans semicolon) to output to the stdio stream
7805 @var{stream} a directive telling the assembler that the type of the
7806 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7807 that string is always either @samp{"function"} or @samp{"object"}, but
7808 you should not count on this.
7810 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7811 definition of this macro is provided.
7814 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7815 A C statement (sans semicolon) to output to the stdio stream
7816 @var{stream} any text necessary for declaring the name @var{name} of a
7817 function which is being defined. This macro is responsible for
7818 outputting the label definition (perhaps using
7819 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7820 @code{FUNCTION_DECL} tree node representing the function.
7822 If this macro is not defined, then the function name is defined in the
7823 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7825 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7829 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7830 A C statement (sans semicolon) to output to the stdio stream
7831 @var{stream} any text necessary for declaring the size of a function
7832 which is being defined. The argument @var{name} is the name of the
7833 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7834 representing the function.
7836 If this macro is not defined, then the function size is not defined.
7838 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7842 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7843 A C statement (sans semicolon) to output to the stdio stream
7844 @var{stream} any text necessary for declaring the name @var{name} of an
7845 initialized variable which is being defined. This macro must output the
7846 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7847 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7849 If this macro is not defined, then the variable name is defined in the
7850 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7852 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7853 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7856 @deftypefn {Target Hook} void TARGET_ASM_DECLARE_CONSTANT_NAME (FILE *@var{file}, const char *@var{name}, const_tree @var{expr}, HOST_WIDE_INT @var{size})
7857 A target hook to output to the stdio stream @var{file} any text necessary
7858 for declaring the name @var{name} of a constant which is being defined. This
7859 target hook is responsible for outputting the label definition (perhaps using
7860 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7861 and @var{size} is the size of the constant in bytes. The @var{name}
7862 will be an internal label.
7864 The default version of this target hook, define the @var{name} in the
7865 usual manner as a label (by means of @code{assemble_label}).
7867 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7870 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7871 A C statement (sans semicolon) to output to the stdio stream
7872 @var{stream} any text necessary for claiming a register @var{regno}
7873 for a global variable @var{decl} with name @var{name}.
7875 If you don't define this macro, that is equivalent to defining it to do
7879 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7880 A C statement (sans semicolon) to finish up declaring a variable name
7881 once the compiler has processed its initializer fully and thus has had a
7882 chance to determine the size of an array when controlled by an
7883 initializer. This is used on systems where it's necessary to declare
7884 something about the size of the object.
7886 If you don't define this macro, that is equivalent to defining it to do
7889 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7890 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7893 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7894 This target hook is a function to output to the stdio stream
7895 @var{stream} some commands that will make the label @var{name} global;
7896 that is, available for reference from other files.
7898 The default implementation relies on a proper definition of
7899 @code{GLOBAL_ASM_OP}.
7902 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7903 This target hook is a function to output to the stdio stream
7904 @var{stream} some commands that will make the name associated with @var{decl}
7905 global; that is, available for reference from other files.
7907 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7910 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7911 A C statement (sans semicolon) to output to the stdio stream
7912 @var{stream} some commands that will make the label @var{name} weak;
7913 that is, available for reference from other files but only used if
7914 no other definition is available. Use the expression
7915 @code{assemble_name (@var{stream}, @var{name})} to output the name
7916 itself; before and after that, output the additional assembler syntax
7917 for making that name weak, and a newline.
7919 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7920 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7924 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7925 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7926 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7927 or variable decl. If @var{value} is not @code{NULL}, this C statement
7928 should output to the stdio stream @var{stream} assembler code which
7929 defines (equates) the weak symbol @var{name} to have the value
7930 @var{value}. If @var{value} is @code{NULL}, it should output commands
7931 to make @var{name} weak.
7934 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7935 Outputs a directive that enables @var{name} to be used to refer to
7936 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7937 declaration of @code{name}.
7940 @defmac SUPPORTS_WEAK
7941 A preprocessor constant expression which evaluates to true if the target
7942 supports weak symbols.
7944 If you don't define this macro, @file{defaults.h} provides a default
7945 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7946 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
7949 @defmac TARGET_SUPPORTS_WEAK
7950 A C expression which evaluates to true if the target supports weak symbols.
7952 If you don't define this macro, @file{defaults.h} provides a default
7953 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
7954 this macro if you want to control weak symbol support with a compiler
7955 flag such as @option{-melf}.
7958 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7959 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7960 public symbol such that extra copies in multiple translation units will
7961 be discarded by the linker. Define this macro if your object file
7962 format provides support for this concept, such as the @samp{COMDAT}
7963 section flags in the Microsoft Windows PE/COFF format, and this support
7964 requires changes to @var{decl}, such as putting it in a separate section.
7967 @defmac SUPPORTS_ONE_ONLY
7968 A C expression which evaluates to true if the target supports one-only
7971 If you don't define this macro, @file{varasm.c} provides a default
7972 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7973 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7974 you want to control one-only symbol support with a compiler flag, or if
7975 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7976 be emitted as one-only.
7979 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
7980 This target hook is a function to output to @var{asm_out_file} some
7981 commands that will make the symbol(s) associated with @var{decl} have
7982 hidden, protected or internal visibility as specified by @var{visibility}.
7985 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7986 A C expression that evaluates to true if the target's linker expects
7987 that weak symbols do not appear in a static archive's table of contents.
7988 The default is @code{0}.
7990 Leaving weak symbols out of an archive's table of contents means that,
7991 if a symbol will only have a definition in one translation unit and
7992 will have undefined references from other translation units, that
7993 symbol should not be weak. Defining this macro to be nonzero will
7994 thus have the effect that certain symbols that would normally be weak
7995 (explicit template instantiations, and vtables for polymorphic classes
7996 with noninline key methods) will instead be nonweak.
7998 The C++ ABI requires this macro to be zero. Define this macro for
7999 targets where full C++ ABI compliance is impossible and where linker
8000 restrictions require weak symbols to be left out of a static archive's
8004 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8005 A C statement (sans semicolon) to output to the stdio stream
8006 @var{stream} any text necessary for declaring the name of an external
8007 symbol named @var{name} which is referenced in this compilation but
8008 not defined. The value of @var{decl} is the tree node for the
8011 This macro need not be defined if it does not need to output anything.
8012 The GNU assembler and most Unix assemblers don't require anything.
8015 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8016 This target hook is a function to output to @var{asm_out_file} an assembler
8017 pseudo-op to declare a library function name external. The name of the
8018 library function is given by @var{symref}, which is a @code{symbol_ref}.
8021 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8022 This target hook is a function to output to @var{asm_out_file} an assembler
8023 directive to annotate @var{symbol} as used. The Darwin target uses the
8024 .no_dead_code_strip directive.
8027 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8028 A C statement (sans semicolon) to output to the stdio stream
8029 @var{stream} a reference in assembler syntax to a label named
8030 @var{name}. This should add @samp{_} to the front of the name, if that
8031 is customary on your operating system, as it is in most Berkeley Unix
8032 systems. This macro is used in @code{assemble_name}.
8035 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8036 Given a symbol @var{name}, perform same mangling as @code{varasm.c}'s @code{assemble_name}, but in memory rather than to a file stream, returning result as an @code{IDENTIFIER_NODE}. Required for correct LTO symtabs. The default implementation calls the @code{TARGET_STRIP_NAME_ENCODING} hook and then prepends the @code{USER_LABEL_PREFIX}, if any.
8039 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8040 A C statement (sans semicolon) to output a reference to
8041 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8042 will be used to output the name of the symbol. This macro may be used
8043 to modify the way a symbol is referenced depending on information
8044 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8047 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8048 A C statement (sans semicolon) to output a reference to @var{buf}, the
8049 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8050 @code{assemble_name} will be used to output the name of the symbol.
8051 This macro is not used by @code{output_asm_label}, or the @code{%l}
8052 specifier that calls it; the intention is that this macro should be set
8053 when it is necessary to output a label differently when its address is
8057 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8058 A function to output to the stdio stream @var{stream} a label whose
8059 name is made from the string @var{prefix} and the number @var{labelno}.
8061 It is absolutely essential that these labels be distinct from the labels
8062 used for user-level functions and variables. Otherwise, certain programs
8063 will have name conflicts with internal labels.
8065 It is desirable to exclude internal labels from the symbol table of the
8066 object file. Most assemblers have a naming convention for labels that
8067 should be excluded; on many systems, the letter @samp{L} at the
8068 beginning of a label has this effect. You should find out what
8069 convention your system uses, and follow it.
8071 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8074 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8075 A C statement to output to the stdio stream @var{stream} a debug info
8076 label whose name is made from the string @var{prefix} and the number
8077 @var{num}. This is useful for VLIW targets, where debug info labels
8078 may need to be treated differently than branch target labels. On some
8079 systems, branch target labels must be at the beginning of instruction
8080 bundles, but debug info labels can occur in the middle of instruction
8083 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8087 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8088 A C statement to store into the string @var{string} a label whose name
8089 is made from the string @var{prefix} and the number @var{num}.
8091 This string, when output subsequently by @code{assemble_name}, should
8092 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8093 with the same @var{prefix} and @var{num}.
8095 If the string begins with @samp{*}, then @code{assemble_name} will
8096 output the rest of the string unchanged. It is often convenient for
8097 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8098 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8099 to output the string, and may change it. (Of course,
8100 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8101 you should know what it does on your machine.)
8104 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8105 A C expression to assign to @var{outvar} (which is a variable of type
8106 @code{char *}) a newly allocated string made from the string
8107 @var{name} and the number @var{number}, with some suitable punctuation
8108 added. Use @code{alloca} to get space for the string.
8110 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8111 produce an assembler label for an internal static variable whose name is
8112 @var{name}. Therefore, the string must be such as to result in valid
8113 assembler code. The argument @var{number} is different each time this
8114 macro is executed; it prevents conflicts between similarly-named
8115 internal static variables in different scopes.
8117 Ideally this string should not be a valid C identifier, to prevent any
8118 conflict with the user's own symbols. Most assemblers allow periods
8119 or percent signs in assembler symbols; putting at least one of these
8120 between the name and the number will suffice.
8122 If this macro is not defined, a default definition will be provided
8123 which is correct for most systems.
8126 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8127 A C statement to output to the stdio stream @var{stream} assembler code
8128 which defines (equates) the symbol @var{name} to have the value @var{value}.
8131 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8132 correct for most systems.
8135 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8136 A C statement to output to the stdio stream @var{stream} assembler code
8137 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8138 to have the value of the tree node @var{decl_of_value}. This macro will
8139 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8140 the tree nodes are available.
8143 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8144 correct for most systems.
8147 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8148 A C statement that evaluates to true if the assembler code which defines
8149 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8150 of the tree node @var{decl_of_value} should be emitted near the end of the
8151 current compilation unit. The default is to not defer output of defines.
8152 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8153 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8156 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8157 A C statement to output to the stdio stream @var{stream} assembler code
8158 which defines (equates) the weak symbol @var{name} to have the value
8159 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8160 an undefined weak symbol.
8162 Define this macro if the target only supports weak aliases; define
8163 @code{ASM_OUTPUT_DEF} instead if possible.
8166 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8167 Define this macro to override the default assembler names used for
8168 Objective-C methods.
8170 The default name is a unique method number followed by the name of the
8171 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8172 the category is also included in the assembler name (e.g.@:
8175 These names are safe on most systems, but make debugging difficult since
8176 the method's selector is not present in the name. Therefore, particular
8177 systems define other ways of computing names.
8179 @var{buf} is an expression of type @code{char *} which gives you a
8180 buffer in which to store the name; its length is as long as
8181 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8182 50 characters extra.
8184 The argument @var{is_inst} specifies whether the method is an instance
8185 method or a class method; @var{class_name} is the name of the class;
8186 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8187 in a category); and @var{sel_name} is the name of the selector.
8189 On systems where the assembler can handle quoted names, you can use this
8190 macro to provide more human-readable names.
8193 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
8194 A C statement (sans semicolon) to output to the stdio stream
8195 @var{stream} commands to declare that the label @var{name} is an
8196 Objective-C class reference. This is only needed for targets whose
8197 linkers have special support for NeXT-style runtimes.
8200 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
8201 A C statement (sans semicolon) to output to the stdio stream
8202 @var{stream} commands to declare that the label @var{name} is an
8203 unresolved Objective-C class reference. This is only needed for targets
8204 whose linkers have special support for NeXT-style runtimes.
8207 @node Initialization
8208 @subsection How Initialization Functions Are Handled
8209 @cindex initialization routines
8210 @cindex termination routines
8211 @cindex constructors, output of
8212 @cindex destructors, output of
8214 The compiled code for certain languages includes @dfn{constructors}
8215 (also called @dfn{initialization routines})---functions to initialize
8216 data in the program when the program is started. These functions need
8217 to be called before the program is ``started''---that is to say, before
8218 @code{main} is called.
8220 Compiling some languages generates @dfn{destructors} (also called
8221 @dfn{termination routines}) that should be called when the program
8224 To make the initialization and termination functions work, the compiler
8225 must output something in the assembler code to cause those functions to
8226 be called at the appropriate time. When you port the compiler to a new
8227 system, you need to specify how to do this.
8229 There are two major ways that GCC currently supports the execution of
8230 initialization and termination functions. Each way has two variants.
8231 Much of the structure is common to all four variations.
8233 @findex __CTOR_LIST__
8234 @findex __DTOR_LIST__
8235 The linker must build two lists of these functions---a list of
8236 initialization functions, called @code{__CTOR_LIST__}, and a list of
8237 termination functions, called @code{__DTOR_LIST__}.
8239 Each list always begins with an ignored function pointer (which may hold
8240 0, @minus{}1, or a count of the function pointers after it, depending on
8241 the environment). This is followed by a series of zero or more function
8242 pointers to constructors (or destructors), followed by a function
8243 pointer containing zero.
8245 Depending on the operating system and its executable file format, either
8246 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8247 time and exit time. Constructors are called in reverse order of the
8248 list; destructors in forward order.
8250 The best way to handle static constructors works only for object file
8251 formats which provide arbitrarily-named sections. A section is set
8252 aside for a list of constructors, and another for a list of destructors.
8253 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8254 object file that defines an initialization function also puts a word in
8255 the constructor section to point to that function. The linker
8256 accumulates all these words into one contiguous @samp{.ctors} section.
8257 Termination functions are handled similarly.
8259 This method will be chosen as the default by @file{target-def.h} if
8260 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8261 support arbitrary sections, but does support special designated
8262 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8263 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8265 When arbitrary sections are available, there are two variants, depending
8266 upon how the code in @file{crtstuff.c} is called. On systems that
8267 support a @dfn{.init} section which is executed at program startup,
8268 parts of @file{crtstuff.c} are compiled into that section. The
8269 program is linked by the @command{gcc} driver like this:
8272 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8275 The prologue of a function (@code{__init}) appears in the @code{.init}
8276 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8277 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8278 files are provided by the operating system or by the GNU C library, but
8279 are provided by GCC for a few targets.
8281 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8282 compiled from @file{crtstuff.c}. They contain, among other things, code
8283 fragments within the @code{.init} and @code{.fini} sections that branch
8284 to routines in the @code{.text} section. The linker will pull all parts
8285 of a section together, which results in a complete @code{__init} function
8286 that invokes the routines we need at startup.
8288 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8291 If no init section is available, when GCC compiles any function called
8292 @code{main} (or more accurately, any function designated as a program
8293 entry point by the language front end calling @code{expand_main_function}),
8294 it inserts a procedure call to @code{__main} as the first executable code
8295 after the function prologue. The @code{__main} function is defined
8296 in @file{libgcc2.c} and runs the global constructors.
8298 In file formats that don't support arbitrary sections, there are again
8299 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8300 and an `a.out' format must be used. In this case,
8301 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8302 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8303 and with the address of the void function containing the initialization
8304 code as its value. The GNU linker recognizes this as a request to add
8305 the value to a @dfn{set}; the values are accumulated, and are eventually
8306 placed in the executable as a vector in the format described above, with
8307 a leading (ignored) count and a trailing zero element.
8308 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8309 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8310 the compilation of @code{main} to call @code{__main} as above, starting
8311 the initialization process.
8313 The last variant uses neither arbitrary sections nor the GNU linker.
8314 This is preferable when you want to do dynamic linking and when using
8315 file formats which the GNU linker does not support, such as `ECOFF'@. In
8316 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8317 termination functions are recognized simply by their names. This requires
8318 an extra program in the linkage step, called @command{collect2}. This program
8319 pretends to be the linker, for use with GCC; it does its job by running
8320 the ordinary linker, but also arranges to include the vectors of
8321 initialization and termination functions. These functions are called
8322 via @code{__main} as described above. In order to use this method,
8323 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8326 The following section describes the specific macros that control and
8327 customize the handling of initialization and termination functions.
8330 @node Macros for Initialization
8331 @subsection Macros Controlling Initialization Routines
8333 Here are the macros that control how the compiler handles initialization
8334 and termination functions:
8336 @defmac INIT_SECTION_ASM_OP
8337 If defined, a C string constant, including spacing, for the assembler
8338 operation to identify the following data as initialization code. If not
8339 defined, GCC will assume such a section does not exist. When you are
8340 using special sections for initialization and termination functions, this
8341 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8342 run the initialization functions.
8345 @defmac HAS_INIT_SECTION
8346 If defined, @code{main} will not call @code{__main} as described above.
8347 This macro should be defined for systems that control start-up code
8348 on a symbol-by-symbol basis, such as OSF/1, and should not
8349 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8352 @defmac LD_INIT_SWITCH
8353 If defined, a C string constant for a switch that tells the linker that
8354 the following symbol is an initialization routine.
8357 @defmac LD_FINI_SWITCH
8358 If defined, a C string constant for a switch that tells the linker that
8359 the following symbol is a finalization routine.
8362 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8363 If defined, a C statement that will write a function that can be
8364 automatically called when a shared library is loaded. The function
8365 should call @var{func}, which takes no arguments. If not defined, and
8366 the object format requires an explicit initialization function, then a
8367 function called @code{_GLOBAL__DI} will be generated.
8369 This function and the following one are used by collect2 when linking a
8370 shared library that needs constructors or destructors, or has DWARF2
8371 exception tables embedded in the code.
8374 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8375 If defined, a C statement that will write a function that can be
8376 automatically called when a shared library is unloaded. The function
8377 should call @var{func}, which takes no arguments. If not defined, and
8378 the object format requires an explicit finalization function, then a
8379 function called @code{_GLOBAL__DD} will be generated.
8382 @defmac INVOKE__main
8383 If defined, @code{main} will call @code{__main} despite the presence of
8384 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8385 where the init section is not actually run automatically, but is still
8386 useful for collecting the lists of constructors and destructors.
8389 @defmac SUPPORTS_INIT_PRIORITY
8390 If nonzero, the C++ @code{init_priority} attribute is supported and the
8391 compiler should emit instructions to control the order of initialization
8392 of objects. If zero, the compiler will issue an error message upon
8393 encountering an @code{init_priority} attribute.
8396 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8397 This value is true if the target supports some ``native'' method of
8398 collecting constructors and destructors to be run at startup and exit.
8399 It is false if we must use @command{collect2}.
8402 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8403 If defined, a function that outputs assembler code to arrange to call
8404 the function referenced by @var{symbol} at initialization time.
8406 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8407 no arguments and with no return value. If the target supports initialization
8408 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8409 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8411 If this macro is not defined by the target, a suitable default will
8412 be chosen if (1) the target supports arbitrary section names, (2) the
8413 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8417 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8418 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8419 functions rather than initialization functions.
8422 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8423 generated for the generated object file will have static linkage.
8425 If your system uses @command{collect2} as the means of processing
8426 constructors, then that program normally uses @command{nm} to scan
8427 an object file for constructor functions to be called.
8429 On certain kinds of systems, you can define this macro to make
8430 @command{collect2} work faster (and, in some cases, make it work at all):
8432 @defmac OBJECT_FORMAT_COFF
8433 Define this macro if the system uses COFF (Common Object File Format)
8434 object files, so that @command{collect2} can assume this format and scan
8435 object files directly for dynamic constructor/destructor functions.
8437 This macro is effective only in a native compiler; @command{collect2} as
8438 part of a cross compiler always uses @command{nm} for the target machine.
8441 @defmac REAL_NM_FILE_NAME
8442 Define this macro as a C string constant containing the file name to use
8443 to execute @command{nm}. The default is to search the path normally for
8448 @command{collect2} calls @command{nm} to scan object files for static
8449 constructors and destructors and LTO info. By default, @option{-n} is
8450 passed. Define @code{NM_FLAGS} to a C string constant if other options
8451 are needed to get the same output format as GNU @command{nm -n}
8455 If your system supports shared libraries and has a program to list the
8456 dynamic dependencies of a given library or executable, you can define
8457 these macros to enable support for running initialization and
8458 termination functions in shared libraries:
8461 Define this macro to a C string constant containing the name of the program
8462 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8465 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8466 Define this macro to be C code that extracts filenames from the output
8467 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8468 of type @code{char *} that points to the beginning of a line of output
8469 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8470 code must advance @var{ptr} to the beginning of the filename on that
8471 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8474 @defmac SHLIB_SUFFIX
8475 Define this macro to a C string constant containing the default shared
8476 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8477 strips version information after this suffix when generating global
8478 constructor and destructor names. This define is only needed on targets
8479 that use @command{collect2} to process constructors and destructors.
8482 @node Instruction Output
8483 @subsection Output of Assembler Instructions
8485 @c prevent bad page break with this line
8486 This describes assembler instruction output.
8488 @defmac REGISTER_NAMES
8489 A C initializer containing the assembler's names for the machine
8490 registers, each one as a C string constant. This is what translates
8491 register numbers in the compiler into assembler language.
8494 @defmac ADDITIONAL_REGISTER_NAMES
8495 If defined, a C initializer for an array of structures containing a name
8496 and a register number. This macro defines additional names for hard
8497 registers, thus allowing the @code{asm} option in declarations to refer
8498 to registers using alternate names.
8501 @defmac OVERLAPPING_REGISTER_NAMES
8502 If defined, a C initializer for an array of structures containing a
8503 name, a register number and a count of the number of consecutive
8504 machine registers the name overlaps. This macro defines additional
8505 names for hard registers, thus allowing the @code{asm} option in
8506 declarations to refer to registers using alternate names. Unlike
8507 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8508 register name implies multiple underlying registers.
8510 This macro should be used when it is important that a clobber in an
8511 @code{asm} statement clobbers all the underlying values implied by the
8512 register name. For example, on ARM, clobbering the double-precision
8513 VFP register ``d0'' implies clobbering both single-precision registers
8517 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8518 Define this macro if you are using an unusual assembler that
8519 requires different names for the machine instructions.
8521 The definition is a C statement or statements which output an
8522 assembler instruction opcode to the stdio stream @var{stream}. The
8523 macro-operand @var{ptr} is a variable of type @code{char *} which
8524 points to the opcode name in its ``internal'' form---the form that is
8525 written in the machine description. The definition should output the
8526 opcode name to @var{stream}, performing any translation you desire, and
8527 increment the variable @var{ptr} to point at the end of the opcode
8528 so that it will not be output twice.
8530 In fact, your macro definition may process less than the entire opcode
8531 name, or more than the opcode name; but if you want to process text
8532 that includes @samp{%}-sequences to substitute operands, you must take
8533 care of the substitution yourself. Just be sure to increment
8534 @var{ptr} over whatever text should not be output normally.
8536 @findex recog_data.operand
8537 If you need to look at the operand values, they can be found as the
8538 elements of @code{recog_data.operand}.
8540 If the macro definition does nothing, the instruction is output
8544 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8545 If defined, a C statement to be executed just prior to the output of
8546 assembler code for @var{insn}, to modify the extracted operands so
8547 they will be output differently.
8549 Here the argument @var{opvec} is the vector containing the operands
8550 extracted from @var{insn}, and @var{noperands} is the number of
8551 elements of the vector which contain meaningful data for this insn.
8552 The contents of this vector are what will be used to convert the insn
8553 template into assembler code, so you can change the assembler output
8554 by changing the contents of the vector.
8556 This macro is useful when various assembler syntaxes share a single
8557 file of instruction patterns; by defining this macro differently, you
8558 can cause a large class of instructions to be output differently (such
8559 as with rearranged operands). Naturally, variations in assembler
8560 syntax affecting individual insn patterns ought to be handled by
8561 writing conditional output routines in those patterns.
8563 If this macro is not defined, it is equivalent to a null statement.
8566 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx @var{insn}, rtx *@var{opvec}, int @var{noperands})
8567 If defined, this target hook is a function which is executed just after the
8568 output of assembler code for @var{insn}, to change the mode of the assembler
8571 Here the argument @var{opvec} is the vector containing the operands
8572 extracted from @var{insn}, and @var{noperands} is the number of
8573 elements of the vector which contain meaningful data for this insn.
8574 The contents of this vector are what was used to convert the insn
8575 template into assembler code, so you can change the assembler mode
8576 by checking the contents of the vector.
8579 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8580 A C compound statement to output to stdio stream @var{stream} the
8581 assembler syntax for an instruction operand @var{x}. @var{x} is an
8584 @var{code} is a value that can be used to specify one of several ways
8585 of printing the operand. It is used when identical operands must be
8586 printed differently depending on the context. @var{code} comes from
8587 the @samp{%} specification that was used to request printing of the
8588 operand. If the specification was just @samp{%@var{digit}} then
8589 @var{code} is 0; if the specification was @samp{%@var{ltr}
8590 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8593 If @var{x} is a register, this macro should print the register's name.
8594 The names can be found in an array @code{reg_names} whose type is
8595 @code{char *[]}. @code{reg_names} is initialized from
8596 @code{REGISTER_NAMES}.
8598 When the machine description has a specification @samp{%@var{punct}}
8599 (a @samp{%} followed by a punctuation character), this macro is called
8600 with a null pointer for @var{x} and the punctuation character for
8604 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8605 A C expression which evaluates to true if @var{code} is a valid
8606 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8607 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8608 punctuation characters (except for the standard one, @samp{%}) are used
8612 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8613 A C compound statement to output to stdio stream @var{stream} the
8614 assembler syntax for an instruction operand that is a memory reference
8615 whose address is @var{x}. @var{x} is an RTL expression.
8617 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8618 On some machines, the syntax for a symbolic address depends on the
8619 section that the address refers to. On these machines, define the hook
8620 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8621 @code{symbol_ref}, and then check for it here. @xref{Assembler
8625 @findex dbr_sequence_length
8626 @defmac DBR_OUTPUT_SEQEND (@var{file})
8627 A C statement, to be executed after all slot-filler instructions have
8628 been output. If necessary, call @code{dbr_sequence_length} to
8629 determine the number of slots filled in a sequence (zero if not
8630 currently outputting a sequence), to decide how many no-ops to output,
8633 Don't define this macro if it has nothing to do, but it is helpful in
8634 reading assembly output if the extent of the delay sequence is made
8635 explicit (e.g.@: with white space).
8638 @findex final_sequence
8639 Note that output routines for instructions with delay slots must be
8640 prepared to deal with not being output as part of a sequence
8641 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8642 found.) The variable @code{final_sequence} is null when not
8643 processing a sequence, otherwise it contains the @code{sequence} rtx
8647 @defmac REGISTER_PREFIX
8648 @defmacx LOCAL_LABEL_PREFIX
8649 @defmacx USER_LABEL_PREFIX
8650 @defmacx IMMEDIATE_PREFIX
8651 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8652 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8653 @file{final.c}). These are useful when a single @file{md} file must
8654 support multiple assembler formats. In that case, the various @file{tm.h}
8655 files can define these macros differently.
8658 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8659 If defined this macro should expand to a series of @code{case}
8660 statements which will be parsed inside the @code{switch} statement of
8661 the @code{asm_fprintf} function. This allows targets to define extra
8662 printf formats which may useful when generating their assembler
8663 statements. Note that uppercase letters are reserved for future
8664 generic extensions to asm_fprintf, and so are not available to target
8665 specific code. The output file is given by the parameter @var{file}.
8666 The varargs input pointer is @var{argptr} and the rest of the format
8667 string, starting the character after the one that is being switched
8668 upon, is pointed to by @var{format}.
8671 @defmac ASSEMBLER_DIALECT
8672 If your target supports multiple dialects of assembler language (such as
8673 different opcodes), define this macro as a C expression that gives the
8674 numeric index of the assembler language dialect to use, with zero as the
8677 If this macro is defined, you may use constructs of the form
8679 @samp{@{option0|option1|option2@dots{}@}}
8682 in the output templates of patterns (@pxref{Output Template}) or in the
8683 first argument of @code{asm_fprintf}. This construct outputs
8684 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8685 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8686 within these strings retain their usual meaning. If there are fewer
8687 alternatives within the braces than the value of
8688 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8690 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8691 @samp{@}} do not have any special meaning when used in templates or
8692 operands to @code{asm_fprintf}.
8694 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8695 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8696 the variations in assembler language syntax with that mechanism. Define
8697 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8698 if the syntax variant are larger and involve such things as different
8699 opcodes or operand order.
8702 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8703 A C expression to output to @var{stream} some assembler code
8704 which will push hard register number @var{regno} onto the stack.
8705 The code need not be optimal, since this macro is used only when
8709 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8710 A C expression to output to @var{stream} some assembler code
8711 which will pop hard register number @var{regno} off of the stack.
8712 The code need not be optimal, since this macro is used only when
8716 @node Dispatch Tables
8717 @subsection Output of Dispatch Tables
8719 @c prevent bad page break with this line
8720 This concerns dispatch tables.
8722 @cindex dispatch table
8723 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8724 A C statement to output to the stdio stream @var{stream} an assembler
8725 pseudo-instruction to generate a difference between two labels.
8726 @var{value} and @var{rel} are the numbers of two internal labels. The
8727 definitions of these labels are output using
8728 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8729 way here. For example,
8732 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8733 @var{value}, @var{rel})
8736 You must provide this macro on machines where the addresses in a
8737 dispatch table are relative to the table's own address. If defined, GCC
8738 will also use this macro on all machines when producing PIC@.
8739 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8740 mode and flags can be read.
8743 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8744 This macro should be provided on machines where the addresses
8745 in a dispatch table are absolute.
8747 The definition should be a C statement to output to the stdio stream
8748 @var{stream} an assembler pseudo-instruction to generate a reference to
8749 a label. @var{value} is the number of an internal label whose
8750 definition is output using @code{(*targetm.asm_out.internal_label)}.
8754 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8758 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8759 Define this if the label before a jump-table needs to be output
8760 specially. The first three arguments are the same as for
8761 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8762 jump-table which follows (a @code{jump_insn} containing an
8763 @code{addr_vec} or @code{addr_diff_vec}).
8765 This feature is used on system V to output a @code{swbeg} statement
8768 If this macro is not defined, these labels are output with
8769 @code{(*targetm.asm_out.internal_label)}.
8772 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8773 Define this if something special must be output at the end of a
8774 jump-table. The definition should be a C statement to be executed
8775 after the assembler code for the table is written. It should write
8776 the appropriate code to stdio stream @var{stream}. The argument
8777 @var{table} is the jump-table insn, and @var{num} is the label-number
8778 of the preceding label.
8780 If this macro is not defined, nothing special is output at the end of
8784 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
8785 This target hook emits a label at the beginning of each FDE@. It
8786 should be defined on targets where FDEs need special labels, and it
8787 should write the appropriate label, for the FDE associated with the
8788 function declaration @var{decl}, to the stdio stream @var{stream}.
8789 The third argument, @var{for_eh}, is a boolean: true if this is for an
8790 exception table. The fourth argument, @var{empty}, is a boolean:
8791 true if this is a placeholder label for an omitted FDE@.
8793 The default is that FDEs are not given nonlocal labels.
8796 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
8797 This target hook emits a label at the beginning of the exception table.
8798 It should be defined on targets where it is desirable for the table
8799 to be broken up according to function.
8801 The default is that no label is emitted.
8804 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
8805 If the target implements @code{TARGET_ASM_UNWIND_EMIT}, this hook may be used to emit a directive to install a personality hook into the unwind info. This hook should not be used if dwarf2 unwind info is used.
8808 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx @var{insn})
8809 This target hook emits assembly directives required to unwind the
8810 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8811 returns @code{UI_TARGET}.
8814 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8815 True if the @code{TARGET_ASM_UNWIND_EMIT} hook should be called before the assembly for @var{insn} has been emitted, false if the hook should be called afterward.
8818 @node Exception Region Output
8819 @subsection Assembler Commands for Exception Regions
8821 @c prevent bad page break with this line
8823 This describes commands marking the start and the end of an exception
8826 @defmac EH_FRAME_SECTION_NAME
8827 If defined, a C string constant for the name of the section containing
8828 exception handling frame unwind information. If not defined, GCC will
8829 provide a default definition if the target supports named sections.
8830 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8832 You should define this symbol if your target supports DWARF 2 frame
8833 unwind information and the default definition does not work.
8836 @defmac EH_FRAME_IN_DATA_SECTION
8837 If defined, DWARF 2 frame unwind information will be placed in the
8838 data section even though the target supports named sections. This
8839 might be necessary, for instance, if the system linker does garbage
8840 collection and sections cannot be marked as not to be collected.
8842 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8846 @defmac EH_TABLES_CAN_BE_READ_ONLY
8847 Define this macro to 1 if your target is such that no frame unwind
8848 information encoding used with non-PIC code will ever require a
8849 runtime relocation, but the linker may not support merging read-only
8850 and read-write sections into a single read-write section.
8853 @defmac MASK_RETURN_ADDR
8854 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8855 that it does not contain any extraneous set bits in it.
8858 @defmac DWARF2_UNWIND_INFO
8859 Define this macro to 0 if your target supports DWARF 2 frame unwind
8860 information, but it does not yet work with exception handling.
8861 Otherwise, if your target supports this information (if it defines
8862 @code{INCOMING_RETURN_ADDR_RTX} and either @code{UNALIGNED_INT_ASM_OP}
8863 or @code{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8866 @deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
8867 This hook defines the mechanism that will be used for exception handling
8868 by the target. If the target has ABI specified unwind tables, the hook
8869 should return @code{UI_TARGET}. If the target is to use the
8870 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8871 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8872 information, the hook should return @code{UI_DWARF2}.
8874 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8875 This may end up simplifying other parts of target-specific code. The
8876 default implementation of this hook never returns @code{UI_NONE}.
8878 Note that the value returned by this hook should be constant. It should
8879 not depend on anything except the command-line switches described by
8880 @var{opts}. In particular, the
8881 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8882 macros and builtin functions related to exception handling are set up
8883 depending on this setting.
8885 The default implementation of the hook first honors the
8886 @option{--enable-sjlj-exceptions} configure option, then
8887 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
8888 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
8889 must define this hook so that @var{opts} is used correctly.
8892 @deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8893 This variable should be set to @code{true} if the target ABI requires unwinding
8894 tables even when exceptions are not used. It must not be modified by
8895 command-line option processing.
8898 @defmac DONT_USE_BUILTIN_SETJMP
8899 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8900 should use the @code{setjmp}/@code{longjmp} functions from the C library
8901 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8904 @defmac DWARF_CIE_DATA_ALIGNMENT
8905 This macro need only be defined if the target might save registers in the
8906 function prologue at an offset to the stack pointer that is not aligned to
8907 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8908 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8909 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8910 the target supports DWARF 2 frame unwind information.
8913 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8914 Contains the value true if the target should add a zero word onto the
8915 end of a Dwarf-2 frame info section when used for exception handling.
8916 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8920 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8921 Given a register, this hook should return a parallel of registers to
8922 represent where to find the register pieces. Define this hook if the
8923 register and its mode are represented in Dwarf in non-contiguous
8924 locations, or if the register should be represented in more than one
8925 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8926 If not defined, the default is to return @code{NULL_RTX}.
8929 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8930 If some registers are represented in Dwarf-2 unwind information in
8931 multiple pieces, define this hook to fill in information about the
8932 sizes of those pieces in the table used by the unwinder at runtime.
8933 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8934 filling in a single size corresponding to each hard register;
8935 @var{address} is the address of the table.
8938 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8939 This hook is used to output a reference from a frame unwinding table to
8940 the type_info object identified by @var{sym}. It should return @code{true}
8941 if the reference was output. Returning @code{false} will cause the
8942 reference to be output using the normal Dwarf2 routines.
8945 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8946 This flag should be set to @code{true} on targets that use an ARM EABI
8947 based unwinding library, and @code{false} on other targets. This effects
8948 the format of unwinding tables, and how the unwinder in entered after
8949 running a cleanup. The default is @code{false}.
8952 @node Alignment Output
8953 @subsection Assembler Commands for Alignment
8955 @c prevent bad page break with this line
8956 This describes commands for alignment.
8958 @defmac JUMP_ALIGN (@var{label})
8959 The alignment (log base 2) to put in front of @var{label}, which is
8960 a common destination of jumps and has no fallthru incoming edge.
8962 This macro need not be defined if you don't want any special alignment
8963 to be done at such a time. Most machine descriptions do not currently
8966 Unless it's necessary to inspect the @var{label} parameter, it is better
8967 to set the variable @var{align_jumps} in the target's
8968 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8969 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8972 @deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx @var{label})
8973 The maximum number of bytes to skip before @var{label} when applying
8974 @code{JUMP_ALIGN}. This works only if
8975 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8978 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8979 The alignment (log base 2) to put in front of @var{label}, which follows
8982 This macro need not be defined if you don't want any special alignment
8983 to be done at such a time. Most machine descriptions do not currently
8987 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx @var{label})
8988 The maximum number of bytes to skip before @var{label} when applying
8989 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8990 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8993 @defmac LOOP_ALIGN (@var{label})
8994 The alignment (log base 2) to put in front of @var{label}, which follows
8995 a @code{NOTE_INSN_LOOP_BEG} note.
8997 This macro need not be defined if you don't want any special alignment
8998 to be done at such a time. Most machine descriptions do not currently
9001 Unless it's necessary to inspect the @var{label} parameter, it is better
9002 to set the variable @code{align_loops} in the target's
9003 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9004 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
9007 @deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx @var{label})
9008 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
9009 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
9013 @defmac LABEL_ALIGN (@var{label})
9014 The alignment (log base 2) to put in front of @var{label}.
9015 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
9016 the maximum of the specified values is used.
9018 Unless it's necessary to inspect the @var{label} parameter, it is better
9019 to set the variable @code{align_labels} in the target's
9020 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9021 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9024 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx @var{label})
9025 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
9026 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
9030 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9031 A C statement to output to the stdio stream @var{stream} an assembler
9032 instruction to advance the location counter by @var{nbytes} bytes.
9033 Those bytes should be zero when loaded. @var{nbytes} will be a C
9034 expression of type @code{unsigned HOST_WIDE_INT}.
9037 @defmac ASM_NO_SKIP_IN_TEXT
9038 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9039 text section because it fails to put zeros in the bytes that are skipped.
9040 This is true on many Unix systems, where the pseudo--op to skip bytes
9041 produces no-op instructions rather than zeros when used in the text
9045 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9046 A C statement to output to the stdio stream @var{stream} an assembler
9047 command to advance the location counter to a multiple of 2 to the
9048 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9051 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9052 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9053 for padding, if necessary.
9056 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9057 A C statement to output to the stdio stream @var{stream} an assembler
9058 command to advance the location counter to a multiple of 2 to the
9059 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9060 satisfy the alignment request. @var{power} and @var{max_skip} will be
9061 a C expression of type @code{int}.
9065 @node Debugging Info
9066 @section Controlling Debugging Information Format
9068 @c prevent bad page break with this line
9069 This describes how to specify debugging information.
9072 * All Debuggers:: Macros that affect all debugging formats uniformly.
9073 * DBX Options:: Macros enabling specific options in DBX format.
9074 * DBX Hooks:: Hook macros for varying DBX format.
9075 * File Names and DBX:: Macros controlling output of file names in DBX format.
9076 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9077 * VMS Debug:: Macros for VMS debug format.
9081 @subsection Macros Affecting All Debugging Formats
9083 @c prevent bad page break with this line
9084 These macros affect all debugging formats.
9086 @defmac DBX_REGISTER_NUMBER (@var{regno})
9087 A C expression that returns the DBX register number for the compiler
9088 register number @var{regno}. In the default macro provided, the value
9089 of this expression will be @var{regno} itself. But sometimes there are
9090 some registers that the compiler knows about and DBX does not, or vice
9091 versa. In such cases, some register may need to have one number in the
9092 compiler and another for DBX@.
9094 If two registers have consecutive numbers inside GCC, and they can be
9095 used as a pair to hold a multiword value, then they @emph{must} have
9096 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9097 Otherwise, debuggers will be unable to access such a pair, because they
9098 expect register pairs to be consecutive in their own numbering scheme.
9100 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9101 does not preserve register pairs, then what you must do instead is
9102 redefine the actual register numbering scheme.
9105 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9106 A C expression that returns the integer offset value for an automatic
9107 variable having address @var{x} (an RTL expression). The default
9108 computation assumes that @var{x} is based on the frame-pointer and
9109 gives the offset from the frame-pointer. This is required for targets
9110 that produce debugging output for DBX or COFF-style debugging output
9111 for SDB and allow the frame-pointer to be eliminated when the
9112 @option{-g} options is used.
9115 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9116 A C expression that returns the integer offset value for an argument
9117 having address @var{x} (an RTL expression). The nominal offset is
9121 @defmac PREFERRED_DEBUGGING_TYPE
9122 A C expression that returns the type of debugging output GCC should
9123 produce when the user specifies just @option{-g}. Define
9124 this if you have arranged for GCC to support more than one format of
9125 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9126 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9127 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9129 When the user specifies @option{-ggdb}, GCC normally also uses the
9130 value of this macro to select the debugging output format, but with two
9131 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9132 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9133 defined, GCC uses @code{DBX_DEBUG}.
9135 The value of this macro only affects the default debugging output; the
9136 user can always get a specific type of output by using @option{-gstabs},
9137 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9141 @subsection Specific Options for DBX Output
9143 @c prevent bad page break with this line
9144 These are specific options for DBX output.
9146 @defmac DBX_DEBUGGING_INFO
9147 Define this macro if GCC should produce debugging output for DBX
9148 in response to the @option{-g} option.
9151 @defmac XCOFF_DEBUGGING_INFO
9152 Define this macro if GCC should produce XCOFF format debugging output
9153 in response to the @option{-g} option. This is a variant of DBX format.
9156 @defmac DEFAULT_GDB_EXTENSIONS
9157 Define this macro to control whether GCC should by default generate
9158 GDB's extended version of DBX debugging information (assuming DBX-format
9159 debugging information is enabled at all). If you don't define the
9160 macro, the default is 1: always generate the extended information
9161 if there is any occasion to.
9164 @defmac DEBUG_SYMS_TEXT
9165 Define this macro if all @code{.stabs} commands should be output while
9166 in the text section.
9169 @defmac ASM_STABS_OP
9170 A C string constant, including spacing, naming the assembler pseudo op to
9171 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9172 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9173 applies only to DBX debugging information format.
9176 @defmac ASM_STABD_OP
9177 A C string constant, including spacing, naming the assembler pseudo op to
9178 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9179 value is the current location. If you don't define this macro,
9180 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9184 @defmac ASM_STABN_OP
9185 A C string constant, including spacing, naming the assembler pseudo op to
9186 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9187 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9188 macro applies only to DBX debugging information format.
9191 @defmac DBX_NO_XREFS
9192 Define this macro if DBX on your system does not support the construct
9193 @samp{xs@var{tagname}}. On some systems, this construct is used to
9194 describe a forward reference to a structure named @var{tagname}.
9195 On other systems, this construct is not supported at all.
9198 @defmac DBX_CONTIN_LENGTH
9199 A symbol name in DBX-format debugging information is normally
9200 continued (split into two separate @code{.stabs} directives) when it
9201 exceeds a certain length (by default, 80 characters). On some
9202 operating systems, DBX requires this splitting; on others, splitting
9203 must not be done. You can inhibit splitting by defining this macro
9204 with the value zero. You can override the default splitting-length by
9205 defining this macro as an expression for the length you desire.
9208 @defmac DBX_CONTIN_CHAR
9209 Normally continuation is indicated by adding a @samp{\} character to
9210 the end of a @code{.stabs} string when a continuation follows. To use
9211 a different character instead, define this macro as a character
9212 constant for the character you want to use. Do not define this macro
9213 if backslash is correct for your system.
9216 @defmac DBX_STATIC_STAB_DATA_SECTION
9217 Define this macro if it is necessary to go to the data section before
9218 outputting the @samp{.stabs} pseudo-op for a non-global static
9222 @defmac DBX_TYPE_DECL_STABS_CODE
9223 The value to use in the ``code'' field of the @code{.stabs} directive
9224 for a typedef. The default is @code{N_LSYM}.
9227 @defmac DBX_STATIC_CONST_VAR_CODE
9228 The value to use in the ``code'' field of the @code{.stabs} directive
9229 for a static variable located in the text section. DBX format does not
9230 provide any ``right'' way to do this. The default is @code{N_FUN}.
9233 @defmac DBX_REGPARM_STABS_CODE
9234 The value to use in the ``code'' field of the @code{.stabs} directive
9235 for a parameter passed in registers. DBX format does not provide any
9236 ``right'' way to do this. The default is @code{N_RSYM}.
9239 @defmac DBX_REGPARM_STABS_LETTER
9240 The letter to use in DBX symbol data to identify a symbol as a parameter
9241 passed in registers. DBX format does not customarily provide any way to
9242 do this. The default is @code{'P'}.
9245 @defmac DBX_FUNCTION_FIRST
9246 Define this macro if the DBX information for a function and its
9247 arguments should precede the assembler code for the function. Normally,
9248 in DBX format, the debugging information entirely follows the assembler
9252 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9253 Define this macro, with value 1, if the value of a symbol describing
9254 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9255 relative to the start of the enclosing function. Normally, GCC uses
9256 an absolute address.
9259 @defmac DBX_LINES_FUNCTION_RELATIVE
9260 Define this macro, with value 1, if the value of a symbol indicating
9261 the current line number (@code{N_SLINE}) should be relative to the
9262 start of the enclosing function. Normally, GCC uses an absolute address.
9265 @defmac DBX_USE_BINCL
9266 Define this macro if GCC should generate @code{N_BINCL} and
9267 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9268 macro also directs GCC to output a type number as a pair of a file
9269 number and a type number within the file. Normally, GCC does not
9270 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9271 number for a type number.
9275 @subsection Open-Ended Hooks for DBX Format
9277 @c prevent bad page break with this line
9278 These are hooks for DBX format.
9280 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9281 Define this macro to say how to output to @var{stream} the debugging
9282 information for the start of a scope level for variable names. The
9283 argument @var{name} is the name of an assembler symbol (for use with
9284 @code{assemble_name}) whose value is the address where the scope begins.
9287 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9288 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9291 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9292 Define this macro if the target machine requires special handling to
9293 output an @code{N_FUN} entry for the function @var{decl}.
9296 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9297 A C statement to output DBX debugging information before code for line
9298 number @var{line} of the current source file to the stdio stream
9299 @var{stream}. @var{counter} is the number of time the macro was
9300 invoked, including the current invocation; it is intended to generate
9301 unique labels in the assembly output.
9303 This macro should not be defined if the default output is correct, or
9304 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9307 @defmac NO_DBX_FUNCTION_END
9308 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9309 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9310 On those machines, define this macro to turn this feature off without
9311 disturbing the rest of the gdb extensions.
9314 @defmac NO_DBX_BNSYM_ENSYM
9315 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9316 extension construct. On those machines, define this macro to turn this
9317 feature off without disturbing the rest of the gdb extensions.
9320 @node File Names and DBX
9321 @subsection File Names in DBX Format
9323 @c prevent bad page break with this line
9324 This describes file names in DBX format.
9326 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9327 A C statement to output DBX debugging information to the stdio stream
9328 @var{stream}, which indicates that file @var{name} is the main source
9329 file---the file specified as the input file for compilation.
9330 This macro is called only once, at the beginning of compilation.
9332 This macro need not be defined if the standard form of output
9333 for DBX debugging information is appropriate.
9335 It may be necessary to refer to a label equal to the beginning of the
9336 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9337 to do so. If you do this, you must also set the variable
9338 @var{used_ltext_label_name} to @code{true}.
9341 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9342 Define this macro, with value 1, if GCC should not emit an indication
9343 of the current directory for compilation and current source language at
9344 the beginning of the file.
9347 @defmac NO_DBX_GCC_MARKER
9348 Define this macro, with value 1, if GCC should not emit an indication
9349 that this object file was compiled by GCC@. The default is to emit
9350 an @code{N_OPT} stab at the beginning of every source file, with
9351 @samp{gcc2_compiled.} for the string and value 0.
9354 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9355 A C statement to output DBX debugging information at the end of
9356 compilation of the main source file @var{name}. Output should be
9357 written to the stdio stream @var{stream}.
9359 If you don't define this macro, nothing special is output at the end
9360 of compilation, which is correct for most machines.
9363 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9364 Define this macro @emph{instead of} defining
9365 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9366 the end of compilation is an @code{N_SO} stab with an empty string,
9367 whose value is the highest absolute text address in the file.
9372 @subsection Macros for SDB and DWARF Output
9374 @c prevent bad page break with this line
9375 Here are macros for SDB and DWARF output.
9377 @defmac SDB_DEBUGGING_INFO
9378 Define this macro if GCC should produce COFF-style debugging output
9379 for SDB in response to the @option{-g} option.
9382 @defmac DWARF2_DEBUGGING_INFO
9383 Define this macro if GCC should produce dwarf version 2 format
9384 debugging output in response to the @option{-g} option.
9386 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
9387 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9388 be emitted for each function. Instead of an integer return the enum
9389 value for the @code{DW_CC_} tag.
9392 To support optional call frame debugging information, you must also
9393 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9394 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9395 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9396 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9399 @defmac DWARF2_FRAME_INFO
9400 Define this macro to a nonzero value if GCC should always output
9401 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9402 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9403 exceptions are enabled, GCC will output this information not matter
9404 how you define @code{DWARF2_FRAME_INFO}.
9407 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
9408 This hook defines the mechanism that will be used for describing frame
9409 unwind information to the debugger. Normally the hook will return
9410 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9411 return @code{UI_NONE} otherwise.
9413 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9414 is disabled in order to always output DWARF 2 frame information.
9416 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9417 This will suppress generation of the normal debug frame unwind information.
9420 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9421 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9422 line debug info sections. This will result in much more compact line number
9423 tables, and hence is desirable if it works.
9426 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
9427 True if the @code{.debug_pubtypes} and @code{.debug_pubnames} sections should be emitted. These sections are not used on most platforms, and in particular GDB does not use them.
9430 @deftypevr {Target Hook} bool TARGET_DELAY_SCHED2
9431 True if sched2 is not to be run at its normal place. This usually means it will be run as part of machine-specific reorg.
9434 @deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK
9435 True if vartrack is not to be run at its normal place. This usually means it will be run as part of machine-specific reorg.
9438 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9439 A C statement to issue assembly directives that create a difference
9440 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9443 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9444 A C statement to issue assembly directives that create a difference
9445 between the two given labels in system defined units, e.g. instruction
9446 slots on IA64 VMS, using an integer of the given size.
9449 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9450 A C statement to issue assembly directives that create a
9451 section-relative reference to the given @var{label}, using an integer of the
9452 given @var{size}. The label is known to be defined in the given @var{section}.
9455 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9456 A C statement to issue assembly directives that create a self-relative
9457 reference to the given @var{label}, using an integer of the given @var{size}.
9460 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9461 A C statement to issue assembly directives that create a reference to
9462 the DWARF table identifier @var{label} from the current section. This
9463 is used on some systems to avoid garbage collecting a DWARF table which
9464 is referenced by a function.
9467 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
9468 If defined, this target hook is a function which outputs a DTP-relative
9469 reference to the given TLS symbol of the specified size.
9472 @defmac PUT_SDB_@dots{}
9473 Define these macros to override the assembler syntax for the special
9474 SDB assembler directives. See @file{sdbout.c} for a list of these
9475 macros and their arguments. If the standard syntax is used, you need
9476 not define them yourself.
9480 Some assemblers do not support a semicolon as a delimiter, even between
9481 SDB assembler directives. In that case, define this macro to be the
9482 delimiter to use (usually @samp{\n}). It is not necessary to define
9483 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9487 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9488 Define this macro to allow references to unknown structure,
9489 union, or enumeration tags to be emitted. Standard COFF does not
9490 allow handling of unknown references, MIPS ECOFF has support for
9494 @defmac SDB_ALLOW_FORWARD_REFERENCES
9495 Define this macro to allow references to structure, union, or
9496 enumeration tags that have not yet been seen to be handled. Some
9497 assemblers choke if forward tags are used, while some require it.
9500 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9501 A C statement to output SDB debugging information before code for line
9502 number @var{line} of the current source file to the stdio stream
9503 @var{stream}. The default is to emit an @code{.ln} directive.
9508 @subsection Macros for VMS Debug Format
9510 @c prevent bad page break with this line
9511 Here are macros for VMS debug format.
9513 @defmac VMS_DEBUGGING_INFO
9514 Define this macro if GCC should produce debugging output for VMS
9515 in response to the @option{-g} option. The default behavior for VMS
9516 is to generate minimal debug info for a traceback in the absence of
9517 @option{-g} unless explicitly overridden with @option{-g0}. This
9518 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9519 @code{TARGET_OPTION_OVERRIDE}.
9522 @node Floating Point
9523 @section Cross Compilation and Floating Point
9524 @cindex cross compilation and floating point
9525 @cindex floating point and cross compilation
9527 While all modern machines use twos-complement representation for integers,
9528 there are a variety of representations for floating point numbers. This
9529 means that in a cross-compiler the representation of floating point numbers
9530 in the compiled program may be different from that used in the machine
9531 doing the compilation.
9533 Because different representation systems may offer different amounts of
9534 range and precision, all floating point constants must be represented in
9535 the target machine's format. Therefore, the cross compiler cannot
9536 safely use the host machine's floating point arithmetic; it must emulate
9537 the target's arithmetic. To ensure consistency, GCC always uses
9538 emulation to work with floating point values, even when the host and
9539 target floating point formats are identical.
9541 The following macros are provided by @file{real.h} for the compiler to
9542 use. All parts of the compiler which generate or optimize
9543 floating-point calculations must use these macros. They may evaluate
9544 their operands more than once, so operands must not have side effects.
9546 @defmac REAL_VALUE_TYPE
9547 The C data type to be used to hold a floating point value in the target
9548 machine's format. Typically this is a @code{struct} containing an
9549 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9553 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9554 Compares for equality the two values, @var{x} and @var{y}. If the target
9555 floating point format supports negative zeroes and/or NaNs,
9556 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9557 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9560 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9561 Tests whether @var{x} is less than @var{y}.
9564 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9565 Truncates @var{x} to a signed integer, rounding toward zero.
9568 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9569 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9570 @var{x} is negative, returns zero.
9573 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9574 Converts @var{string} into a floating point number in the target machine's
9575 representation for mode @var{mode}. This routine can handle both
9576 decimal and hexadecimal floating point constants, using the syntax
9577 defined by the C language for both.
9580 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9581 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9584 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9585 Determines whether @var{x} represents infinity (positive or negative).
9588 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9589 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9592 @deftypefn Macro void REAL_ARITHMETIC (REAL_VALUE_TYPE @var{output}, enum tree_code @var{code}, REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9593 Calculates an arithmetic operation on the two floating point values
9594 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9597 The operation to be performed is specified by @var{code}. Only the
9598 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9599 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9601 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9602 target's floating point format cannot represent infinity, it will call
9603 @code{abort}. Callers should check for this situation first, using
9604 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9607 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9608 Returns the negative of the floating point value @var{x}.
9611 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9612 Returns the absolute value of @var{x}.
9615 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9616 Truncates the floating point value @var{x} to fit in @var{mode}. The
9617 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9618 appropriate bit pattern to be output as a floating constant whose
9619 precision accords with mode @var{mode}.
9622 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9623 Converts a floating point value @var{x} into a double-precision integer
9624 which is then stored into @var{low} and @var{high}. If the value is not
9625 integral, it is truncated.
9628 @deftypefn Macro void REAL_VALUE_FROM_INT (REAL_VALUE_TYPE @var{x}, HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, enum machine_mode @var{mode})
9629 Converts a double-precision integer found in @var{low} and @var{high},
9630 into a floating point value which is then stored into @var{x}. The
9631 value is truncated to fit in mode @var{mode}.
9634 @node Mode Switching
9635 @section Mode Switching Instructions
9636 @cindex mode switching
9637 The following macros control mode switching optimizations:
9639 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9640 Define this macro if the port needs extra instructions inserted for mode
9641 switching in an optimizing compilation.
9643 For an example, the SH4 can perform both single and double precision
9644 floating point operations, but to perform a single precision operation,
9645 the FPSCR PR bit has to be cleared, while for a double precision
9646 operation, this bit has to be set. Changing the PR bit requires a general
9647 purpose register as a scratch register, hence these FPSCR sets have to
9648 be inserted before reload, i.e.@: you can't put this into instruction emitting
9649 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9651 You can have multiple entities that are mode-switched, and select at run time
9652 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9653 return nonzero for any @var{entity} that needs mode-switching.
9654 If you define this macro, you also have to define
9655 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9656 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9657 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9661 @defmac NUM_MODES_FOR_MODE_SWITCHING
9662 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9663 initializer for an array of integers. Each initializer element
9664 N refers to an entity that needs mode switching, and specifies the number
9665 of different modes that might need to be set for this entity.
9666 The position of the initializer in the initializer---starting counting at
9667 zero---determines the integer that is used to refer to the mode-switched
9669 In macros that take mode arguments / yield a mode result, modes are
9670 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9671 switch is needed / supplied.
9674 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9675 @var{entity} is an integer specifying a mode-switched entity. If
9676 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9677 return an integer value not larger than the corresponding element in
9678 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9679 be switched into prior to the execution of @var{insn}.
9682 @defmac MODE_AFTER (@var{mode}, @var{insn})
9683 If this macro is defined, it is evaluated for every @var{insn} during
9684 mode switching. It determines the mode that an insn results in (if
9685 different from the incoming mode).
9688 @defmac MODE_ENTRY (@var{entity})
9689 If this macro is defined, it is evaluated for every @var{entity} that needs
9690 mode switching. It should evaluate to an integer, which is a mode that
9691 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9692 is defined then @code{MODE_EXIT} must be defined.
9695 @defmac MODE_EXIT (@var{entity})
9696 If this macro is defined, it is evaluated for every @var{entity} that needs
9697 mode switching. It should evaluate to an integer, which is a mode that
9698 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9699 is defined then @code{MODE_ENTRY} must be defined.
9702 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9703 This macro specifies the order in which modes for @var{entity} are processed.
9704 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9705 lowest. The value of the macro should be an integer designating a mode
9706 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9707 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9708 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9711 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9712 Generate one or more insns to set @var{entity} to @var{mode}.
9713 @var{hard_reg_live} is the set of hard registers live at the point where
9714 the insn(s) are to be inserted.
9717 @node Target Attributes
9718 @section Defining target-specific uses of @code{__attribute__}
9719 @cindex target attributes
9720 @cindex machine attributes
9721 @cindex attributes, target-specific
9723 Target-specific attributes may be defined for functions, data and types.
9724 These are described using the following target hooks; they also need to
9725 be documented in @file{extend.texi}.
9727 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9728 If defined, this target hook points to an array of @samp{struct
9729 attribute_spec} (defined in @file{tree.h}) specifying the machine
9730 specific attributes for this target and some of the restrictions on the
9731 entities to which these attributes are applied and the arguments they
9735 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
9736 If defined, this target hook is a function which returns true if the
9737 machine-specific attribute named @var{name} expects an identifier
9738 given as its first argument to be passed on as a plain identifier, not
9739 subjected to name lookup. If this is not defined, the default is
9740 false for all machine-specific attributes.
9743 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
9744 If defined, this target hook is a function which returns zero if the attributes on
9745 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9746 and two if they are nearly compatible (which causes a warning to be
9747 generated). If this is not defined, machine-specific attributes are
9748 supposed always to be compatible.
9751 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9752 If defined, this target hook is a function which assigns default attributes to
9753 the newly defined @var{type}.
9756 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9757 Define this target hook if the merging of type attributes needs special
9758 handling. If defined, the result is a list of the combined
9759 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9760 that @code{comptypes} has already been called and returned 1. This
9761 function may call @code{merge_attributes} to handle machine-independent
9765 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9766 Define this target hook if the merging of decl attributes needs special
9767 handling. If defined, the result is a list of the combined
9768 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9769 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9770 when this is needed are when one attribute overrides another, or when an
9771 attribute is nullified by a subsequent definition. This function may
9772 call @code{merge_attributes} to handle machine-independent merging.
9774 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9775 If the only target-specific handling you require is @samp{dllimport}
9776 for Microsoft Windows targets, you should define the macro
9777 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9778 will then define a function called
9779 @code{merge_dllimport_decl_attributes} which can then be defined as
9780 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9781 add @code{handle_dll_attribute} in the attribute table for your port
9782 to perform initial processing of the @samp{dllimport} and
9783 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9784 @file{i386/i386.c}, for example.
9787 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
9788 @var{decl} is a variable or function with @code{__attribute__((dllimport))} specified. Use this hook if the target needs to add extra validation checks to @code{handle_dll_attribute}.
9791 @defmac TARGET_DECLSPEC
9792 Define this macro to a nonzero value if you want to treat
9793 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9794 default, this behavior is enabled only for targets that define
9795 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9796 of @code{__declspec} is via a built-in macro, but you should not rely
9797 on this implementation detail.
9800 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9801 Define this target hook if you want to be able to add attributes to a decl
9802 when it is being created. This is normally useful for back ends which
9803 wish to implement a pragma by using the attributes which correspond to
9804 the pragma's effect. The @var{node} argument is the decl which is being
9805 created. The @var{attr_ptr} argument is a pointer to the attribute list
9806 for this decl. The list itself should not be modified, since it may be
9807 shared with other decls, but attributes may be chained on the head of
9808 the list and @code{*@var{attr_ptr}} modified to point to the new
9809 attributes, or a copy of the list may be made if further changes are
9813 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
9815 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9816 into the current function, despite its having target-specific
9817 attributes, @code{false} otherwise. By default, if a function has a
9818 target specific attribute attached to it, it will not be inlined.
9821 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9822 This hook is called to parse the @code{attribute(option("..."))}, and
9823 it allows the function to set different target machine compile time
9824 options for the current function that might be different than the
9825 options specified on the command line. The hook should return
9826 @code{true} if the options are valid.
9828 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9829 the function declaration to hold a pointer to a target specific
9830 @var{struct cl_target_option} structure.
9833 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr})
9834 This hook is called to save any additional target specific information
9835 in the @var{struct cl_target_option} structure for function specific
9837 @xref{Option file format}.
9840 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct cl_target_option *@var{ptr})
9841 This hook is called to restore any additional target specific
9842 information in the @var{struct cl_target_option} structure for
9843 function specific options.
9846 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
9847 This hook is called to print any additional target specific
9848 information in the @var{struct cl_target_option} structure for
9849 function specific options.
9852 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
9853 This target hook parses the options for @code{#pragma GCC option} to
9854 set the machine specific options for functions that occur later in the
9855 input stream. The options should be the same as handled by the
9856 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9859 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
9860 Sometimes certain combinations of command options do not make sense on
9861 a particular target machine. You can override the hook
9862 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9863 once just after all the command options have been parsed.
9865 Don't use this hook to turn on various extra optimizations for
9866 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9868 If you need to do something whenever the optimization level is
9869 changed via the optimize attribute or pragma, see
9870 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9873 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
9874 This target hook returns @code{false} if the @var{caller} function
9875 cannot inline @var{callee}, based on target specific information. By
9876 default, inlining is not allowed if the callee function has function
9877 specific target options and the caller does not use the same options.
9881 @section Emulating TLS
9882 @cindex Emulated TLS
9884 For targets whose psABI does not provide Thread Local Storage via
9885 specific relocations and instruction sequences, an emulation layer is
9886 used. A set of target hooks allows this emulation layer to be
9887 configured for the requirements of a particular target. For instance
9888 the psABI may in fact specify TLS support in terms of an emulation
9891 The emulation layer works by creating a control object for every TLS
9892 object. To access the TLS object, a lookup function is provided
9893 which, when given the address of the control object, will return the
9894 address of the current thread's instance of the TLS object.
9896 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9897 Contains the name of the helper function that uses a TLS control
9898 object to locate a TLS instance. The default causes libgcc's
9899 emulated TLS helper function to be used.
9902 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9903 Contains the name of the helper function that should be used at
9904 program startup to register TLS objects that are implicitly
9905 initialized to zero. If this is @code{NULL}, all TLS objects will
9906 have explicit initializers. The default causes libgcc's emulated TLS
9907 registration function to be used.
9910 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9911 Contains the name of the section in which TLS control variables should
9912 be placed. The default of @code{NULL} allows these to be placed in
9916 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9917 Contains the name of the section in which TLS initializers should be
9918 placed. The default of @code{NULL} allows these to be placed in any
9922 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9923 Contains the prefix to be prepended to TLS control variable names.
9924 The default of @code{NULL} uses a target-specific prefix.
9927 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9928 Contains the prefix to be prepended to TLS initializer objects. The
9929 default of @code{NULL} uses a target-specific prefix.
9932 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
9933 Specifies a function that generates the FIELD_DECLs for a TLS control
9934 object type. @var{type} is the RECORD_TYPE the fields are for and
9935 @var{name} should be filled with the structure tag, if the default of
9936 @code{__emutls_object} is unsuitable. The default creates a type suitable
9937 for libgcc's emulated TLS function.
9940 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
9941 Specifies a function that generates the CONSTRUCTOR to initialize a
9942 TLS control object. @var{var} is the TLS control object, @var{decl}
9943 is the TLS object and @var{tmpl_addr} is the address of the
9944 initializer. The default initializes libgcc's emulated TLS control object.
9947 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
9948 Specifies whether the alignment of TLS control variable objects is
9949 fixed and should not be increased as some backends may do to optimize
9950 single objects. The default is false.
9953 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9954 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9955 may be used to describe emulated TLS control objects.
9958 @node MIPS Coprocessors
9959 @section Defining coprocessor specifics for MIPS targets.
9960 @cindex MIPS coprocessor-definition macros
9962 The MIPS specification allows MIPS implementations to have as many as 4
9963 coprocessors, each with as many as 32 private registers. GCC supports
9964 accessing these registers and transferring values between the registers
9965 and memory using asm-ized variables. For example:
9968 register unsigned int cp0count asm ("c0r1");
9974 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9975 names may be added as described below, or the default names may be
9976 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9978 Coprocessor registers are assumed to be epilogue-used; sets to them will
9979 be preserved even if it does not appear that the register is used again
9980 later in the function.
9982 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9983 the FPU@. One accesses COP1 registers through standard mips
9984 floating-point support; they are not included in this mechanism.
9986 There is one macro used in defining the MIPS coprocessor interface which
9987 you may want to override in subtargets; it is described below.
9989 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9990 A comma-separated list (with leading comma) of pairs describing the
9991 alternate names of coprocessor registers. The format of each entry should be
9993 @{ @var{alternatename}, @var{register_number}@}
9999 @section Parameters for Precompiled Header Validity Checking
10000 @cindex parameters, precompiled headers
10002 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
10003 This hook returns a pointer to the data needed by
10004 @code{TARGET_PCH_VALID_P} and sets
10005 @samp{*@var{sz}} to the size of the data in bytes.
10008 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
10009 This hook checks whether the options used to create a PCH file are
10010 compatible with the current settings. It returns @code{NULL}
10011 if so and a suitable error message if not. Error messages will
10012 be presented to the user and must be localized using @samp{_(@var{msg})}.
10014 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
10015 when the PCH file was created and @var{sz} is the size of that data in bytes.
10016 It's safe to assume that the data was created by the same version of the
10017 compiler, so no format checking is needed.
10019 The default definition of @code{default_pch_valid_p} should be
10020 suitable for most targets.
10023 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
10024 If this hook is nonnull, the default implementation of
10025 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
10026 of @code{target_flags}. @var{pch_flags} specifies the value that
10027 @code{target_flags} had when the PCH file was created. The return
10028 value is the same as for @code{TARGET_PCH_VALID_P}.
10032 @section C++ ABI parameters
10033 @cindex parameters, c++ abi
10035 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10036 Define this hook to override the integer type used for guard variables.
10037 These are used to implement one-time construction of static objects. The
10038 default is long_long_integer_type_node.
10041 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10042 This hook determines how guard variables are used. It should return
10043 @code{false} (the default) if the first byte should be used. A return value of
10044 @code{true} indicates that only the least significant bit should be used.
10047 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10048 This hook returns the size of the cookie to use when allocating an array
10049 whose elements have the indicated @var{type}. Assumes that it is already
10050 known that a cookie is needed. The default is
10051 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10052 IA64/Generic C++ ABI@.
10055 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10056 This hook should return @code{true} if the element size should be stored in
10057 array cookies. The default is to return @code{false}.
10060 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10061 If defined by a backend this hook allows the decision made to export
10062 class @var{type} to be overruled. Upon entry @var{import_export}
10063 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10064 to be imported and 0 otherwise. This function should return the
10065 modified value and perform any other actions necessary to support the
10066 backend's targeted operating system.
10069 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10070 This hook should return @code{true} if constructors and destructors return
10071 the address of the object created/destroyed. The default is to return
10075 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10076 This hook returns true if the key method for a class (i.e., the method
10077 which, if defined in the current translation unit, causes the virtual
10078 table to be emitted) may be an inline function. Under the standard
10079 Itanium C++ ABI the key method may be an inline function so long as
10080 the function is not declared inline in the class definition. Under
10081 some variants of the ABI, an inline function can never be the key
10082 method. The default is to return @code{true}.
10085 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10086 @var{decl} is a virtual table, virtual table table, typeinfo object, or other similar implicit class data object that will be emitted with external linkage in this translation unit. No ELF visibility has been explicitly specified. If the target needs to specify a visibility other than that of the containing class, use this hook to set @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
10089 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10090 This hook returns true (the default) if virtual tables and other
10091 similar implicit class data objects are always COMDAT if they have
10092 external linkage. If this hook returns false, then class data for
10093 classes whose virtual table will be emitted in only one translation
10094 unit will not be COMDAT.
10097 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10098 This hook returns true (the default) if the RTTI information for
10099 the basic types which is defined in the C++ runtime should always
10100 be COMDAT, false if it should not be COMDAT.
10103 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10104 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10105 should be used to register static destructors when @option{-fuse-cxa-atexit}
10106 is in effect. The default is to return false to use @code{__cxa_atexit}.
10109 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10110 This hook returns true if the target @code{atexit} function can be used
10111 in the same manner as @code{__cxa_atexit} to register C++ static
10112 destructors. This requires that @code{atexit}-registered functions in
10113 shared libraries are run in the correct order when the libraries are
10114 unloaded. The default is to return false.
10117 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10118 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been defined. Use this hook to make adjustments to the class (eg, tweak visibility or perform any other required target modifications).
10121 @node Named Address Spaces
10122 @section Adding support for named address spaces
10123 @cindex named address spaces
10125 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10126 standards committee, @cite{Programming Languages - C - Extensions to
10127 support embedded processors}, specifies a syntax for embedded
10128 processors to specify alternate address spaces. You can configure a
10129 GCC port to support section 5.1 of the draft report to add support for
10130 address spaces other than the default address space. These address
10131 spaces are new keywords that are similar to the @code{volatile} and
10132 @code{const} type attributes.
10134 Pointers to named address spaces can have a different size than
10135 pointers to the generic address space.
10137 For example, the SPU port uses the @code{__ea} address space to refer
10138 to memory in the host processor, rather than memory local to the SPU
10139 processor. Access to memory in the @code{__ea} address space involves
10140 issuing DMA operations to move data between the host processor and the
10141 local processor memory address space. Pointers in the @code{__ea}
10142 address space are either 32 bits or 64 bits based on the
10143 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10146 Internally, address spaces are represented as a small integer in the
10147 range 0 to 15 with address space 0 being reserved for the generic
10150 To register a named address space qualifier keyword with the C front end,
10151 the target may call the @code{c_register_addr_space} routine. For example,
10152 the SPU port uses the following to declare @code{__ea} as the keyword for
10153 named address space #1:
10155 #define ADDR_SPACE_EA 1
10156 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10159 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10160 Define this to return the machine mode to use for pointers to
10161 @var{address_space} if the target supports named address spaces.
10162 The default version of this hook returns @code{ptr_mode} for the
10163 generic address space only.
10166 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10167 Define this to return the machine mode to use for addresses in
10168 @var{address_space} if the target supports named address spaces.
10169 The default version of this hook returns @code{Pmode} for the
10170 generic address space only.
10173 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (enum machine_mode @var{mode}, addr_space_t @var{as})
10174 Define this to return nonzero if the port can handle pointers
10175 with machine mode @var{mode} to address space @var{as}. This target
10176 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10177 except that it includes explicit named address space support. The default
10178 version of this hook returns true for the modes returned by either the
10179 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10180 target hooks for the given address space.
10183 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{exp}, bool @var{strict}, addr_space_t @var{as})
10184 Define this to return true if @var{exp} is a valid address for mode
10185 @var{mode} in the named address space @var{as}. The @var{strict}
10186 parameter says whether strict addressing is in effect after reload has
10187 finished. This target hook is the same as the
10188 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10189 explicit named address space support.
10192 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode}, addr_space_t @var{as})
10193 Define this to modify an invalid address @var{x} to be a valid address
10194 with mode @var{mode} in the named address space @var{as}. This target
10195 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10196 except that it includes explicit named address space support.
10199 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{superset}, addr_space_t @var{subset})
10200 Define this to return whether the @var{subset} named address space is
10201 contained within the @var{superset} named address space. Pointers to
10202 a named address space that is a subset of another named address space
10203 will be converted automatically without a cast if used together in
10204 arithmetic operations. Pointers to a superset address space can be
10205 converted to pointers to a subset address space via explicit casts.
10208 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10209 Define this to convert the pointer expression represented by the RTL
10210 @var{op} with type @var{from_type} that points to a named address
10211 space to a new pointer expression with type @var{to_type} that points
10212 to a different named address space. When this hook it called, it is
10213 guaranteed that one of the two address spaces is a subset of the other,
10214 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10218 @section Miscellaneous Parameters
10219 @cindex parameters, miscellaneous
10221 @c prevent bad page break with this line
10222 Here are several miscellaneous parameters.
10224 @defmac HAS_LONG_COND_BRANCH
10225 Define this boolean macro to indicate whether or not your architecture
10226 has conditional branches that can span all of memory. It is used in
10227 conjunction with an optimization that partitions hot and cold basic
10228 blocks into separate sections of the executable. If this macro is
10229 set to false, gcc will convert any conditional branches that attempt
10230 to cross between sections into unconditional branches or indirect jumps.
10233 @defmac HAS_LONG_UNCOND_BRANCH
10234 Define this boolean macro to indicate whether or not your architecture
10235 has unconditional branches that can span all of memory. It is used in
10236 conjunction with an optimization that partitions hot and cold basic
10237 blocks into separate sections of the executable. If this macro is
10238 set to false, gcc will convert any unconditional branches that attempt
10239 to cross between sections into indirect jumps.
10242 @defmac CASE_VECTOR_MODE
10243 An alias for a machine mode name. This is the machine mode that
10244 elements of a jump-table should have.
10247 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10248 Optional: return the preferred mode for an @code{addr_diff_vec}
10249 when the minimum and maximum offset are known. If you define this,
10250 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10251 To make this work, you also have to define @code{INSN_ALIGN} and
10252 make the alignment for @code{addr_diff_vec} explicit.
10253 The @var{body} argument is provided so that the offset_unsigned and scale
10254 flags can be updated.
10257 @defmac CASE_VECTOR_PC_RELATIVE
10258 Define this macro to be a C expression to indicate when jump-tables
10259 should contain relative addresses. You need not define this macro if
10260 jump-tables never contain relative addresses, or jump-tables should
10261 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10265 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10266 This function return the smallest number of different values for which it
10267 is best to use a jump-table instead of a tree of conditional branches.
10268 The default is four for machines with a @code{casesi} instruction and
10269 five otherwise. This is best for most machines.
10272 @defmac CASE_USE_BIT_TESTS
10273 Define this macro to be a C expression to indicate whether C switch
10274 statements may be implemented by a sequence of bit tests. This is
10275 advantageous on processors that can efficiently implement left shift
10276 of 1 by the number of bits held in a register, but inappropriate on
10277 targets that would require a loop. By default, this macro returns
10278 @code{true} if the target defines an @code{ashlsi3} pattern, and
10279 @code{false} otherwise.
10282 @defmac WORD_REGISTER_OPERATIONS
10283 Define this macro if operations between registers with integral mode
10284 smaller than a word are always performed on the entire register.
10285 Most RISC machines have this property and most CISC machines do not.
10288 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10289 Define this macro to be a C expression indicating when insns that read
10290 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10291 bits outside of @var{mem_mode} to be either the sign-extension or the
10292 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10293 of @var{mem_mode} for which the
10294 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10295 @code{UNKNOWN} for other modes.
10297 This macro is not called with @var{mem_mode} non-integral or with a width
10298 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10299 value in this case. Do not define this macro if it would always return
10300 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10301 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10303 You may return a non-@code{UNKNOWN} value even if for some hard registers
10304 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10305 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10306 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10307 integral mode larger than this but not larger than @code{word_mode}.
10309 You must return @code{UNKNOWN} if for some hard registers that allow this
10310 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10311 @code{word_mode}, but that they can change to another integral mode that
10312 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10315 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10316 Define this macro if loading short immediate values into registers sign
10320 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
10321 Define this macro if the same instructions that convert a floating
10322 point number to a signed fixed point number also convert validly to an
10326 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
10327 When @option{-ffast-math} is in effect, GCC tries to optimize
10328 divisions by the same divisor, by turning them into multiplications by
10329 the reciprocal. This target hook specifies the minimum number of divisions
10330 that should be there for GCC to perform the optimization for a variable
10331 of mode @var{mode}. The default implementation returns 3 if the machine
10332 has an instruction for the division, and 2 if it does not.
10336 The maximum number of bytes that a single instruction can move quickly
10337 between memory and registers or between two memory locations.
10340 @defmac MAX_MOVE_MAX
10341 The maximum number of bytes that a single instruction can move quickly
10342 between memory and registers or between two memory locations. If this
10343 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10344 constant value that is the largest value that @code{MOVE_MAX} can have
10348 @defmac SHIFT_COUNT_TRUNCATED
10349 A C expression that is nonzero if on this machine the number of bits
10350 actually used for the count of a shift operation is equal to the number
10351 of bits needed to represent the size of the object being shifted. When
10352 this macro is nonzero, the compiler will assume that it is safe to omit
10353 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10354 truncates the count of a shift operation. On machines that have
10355 instructions that act on bit-fields at variable positions, which may
10356 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10357 also enables deletion of truncations of the values that serve as
10358 arguments to bit-field instructions.
10360 If both types of instructions truncate the count (for shifts) and
10361 position (for bit-field operations), or if no variable-position bit-field
10362 instructions exist, you should define this macro.
10364 However, on some machines, such as the 80386 and the 680x0, truncation
10365 only applies to shift operations and not the (real or pretended)
10366 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10367 such machines. Instead, add patterns to the @file{md} file that include
10368 the implied truncation of the shift instructions.
10370 You need not define this macro if it would always have the value of zero.
10373 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10374 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
10375 This function describes how the standard shift patterns for @var{mode}
10376 deal with shifts by negative amounts or by more than the width of the mode.
10377 @xref{shift patterns}.
10379 On many machines, the shift patterns will apply a mask @var{m} to the
10380 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10381 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10382 this is true for mode @var{mode}, the function should return @var{m},
10383 otherwise it should return 0. A return value of 0 indicates that no
10384 particular behavior is guaranteed.
10386 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10387 @emph{not} apply to general shift rtxes; it applies only to instructions
10388 that are generated by the named shift patterns.
10390 The default implementation of this function returns
10391 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10392 and 0 otherwise. This definition is always safe, but if
10393 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10394 nevertheless truncate the shift count, you may get better code
10398 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10399 A C expression which is nonzero if on this machine it is safe to
10400 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10401 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10402 operating on it as if it had only @var{outprec} bits.
10404 On many machines, this expression can be 1.
10406 @c rearranged this, removed the phrase "it is reported that". this was
10407 @c to fix an overfull hbox. --mew 10feb93
10408 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10409 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10410 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10411 such cases may improve things.
10414 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
10415 The representation of an integral mode can be such that the values
10416 are always extended to a wider integral mode. Return
10417 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10418 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10419 otherwise. (Currently, none of the targets use zero-extended
10420 representation this way so unlike @code{LOAD_EXTEND_OP},
10421 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10422 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10423 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10424 widest integral mode and currently we take advantage of this fact.)
10426 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10427 value even if the extension is not performed on certain hard registers
10428 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10429 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10431 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10432 describe two related properties. If you define
10433 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10434 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10437 In order to enforce the representation of @code{mode},
10438 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10442 @defmac STORE_FLAG_VALUE
10443 A C expression describing the value returned by a comparison operator
10444 with an integral mode and stored by a store-flag instruction
10445 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10446 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10447 comparison operators whose results have a @code{MODE_INT} mode.
10449 A value of 1 or @minus{}1 means that the instruction implementing the
10450 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10451 and 0 when the comparison is false. Otherwise, the value indicates
10452 which bits of the result are guaranteed to be 1 when the comparison is
10453 true. This value is interpreted in the mode of the comparison
10454 operation, which is given by the mode of the first operand in the
10455 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10456 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10459 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10460 generate code that depends only on the specified bits. It can also
10461 replace comparison operators with equivalent operations if they cause
10462 the required bits to be set, even if the remaining bits are undefined.
10463 For example, on a machine whose comparison operators return an
10464 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10465 @samp{0x80000000}, saying that just the sign bit is relevant, the
10469 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10473 can be converted to
10476 (ashift:SI @var{x} (const_int @var{n}))
10480 where @var{n} is the appropriate shift count to move the bit being
10481 tested into the sign bit.
10483 There is no way to describe a machine that always sets the low-order bit
10484 for a true value, but does not guarantee the value of any other bits,
10485 but we do not know of any machine that has such an instruction. If you
10486 are trying to port GCC to such a machine, include an instruction to
10487 perform a logical-and of the result with 1 in the pattern for the
10488 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10490 Often, a machine will have multiple instructions that obtain a value
10491 from a comparison (or the condition codes). Here are rules to guide the
10492 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10497 Use the shortest sequence that yields a valid definition for
10498 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10499 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10500 comparison operators to do so because there may be opportunities to
10501 combine the normalization with other operations.
10504 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10505 slightly preferred on machines with expensive jumps and 1 preferred on
10509 As a second choice, choose a value of @samp{0x80000001} if instructions
10510 exist that set both the sign and low-order bits but do not define the
10514 Otherwise, use a value of @samp{0x80000000}.
10517 Many machines can produce both the value chosen for
10518 @code{STORE_FLAG_VALUE} and its negation in the same number of
10519 instructions. On those machines, you should also define a pattern for
10520 those cases, e.g., one matching
10523 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10526 Some machines can also perform @code{and} or @code{plus} operations on
10527 condition code values with less instructions than the corresponding
10528 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10529 machines, define the appropriate patterns. Use the names @code{incscc}
10530 and @code{decscc}, respectively, for the patterns which perform
10531 @code{plus} or @code{minus} operations on condition code values. See
10532 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10533 find such instruction sequences on other machines.
10535 If this macro is not defined, the default value, 1, is used. You need
10536 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10537 instructions, or if the value generated by these instructions is 1.
10540 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10541 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10542 returned when comparison operators with floating-point results are true.
10543 Define this macro on machines that have comparison operations that return
10544 floating-point values. If there are no such operations, do not define
10548 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10549 A C expression that gives a rtx representing the nonzero true element
10550 for vector comparisons. The returned rtx should be valid for the inner
10551 mode of @var{mode} which is guaranteed to be a vector mode. Define
10552 this macro on machines that have vector comparison operations that
10553 return a vector result. If there are no such operations, do not define
10554 this macro. Typically, this macro is defined as @code{const1_rtx} or
10555 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10556 the compiler optimizing such vector comparison operations for the
10560 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10561 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10562 A C expression that indicates whether the architecture defines a value
10563 for @code{clz} or @code{ctz} with a zero operand.
10564 A result of @code{0} indicates the value is undefined.
10565 If the value is defined for only the RTL expression, the macro should
10566 evaluate to @code{1}; if the value applies also to the corresponding optab
10567 entry (which is normally the case if it expands directly into
10568 the corresponding RTL), then the macro should evaluate to @code{2}.
10569 In the cases where the value is defined, @var{value} should be set to
10572 If this macro is not defined, the value of @code{clz} or
10573 @code{ctz} at zero is assumed to be undefined.
10575 This macro must be defined if the target's expansion for @code{ffs}
10576 relies on a particular value to get correct results. Otherwise it
10577 is not necessary, though it may be used to optimize some corner cases, and
10578 to provide a default expansion for the @code{ffs} optab.
10580 Note that regardless of this macro the ``definedness'' of @code{clz}
10581 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10582 visible to the user. Thus one may be free to adjust the value at will
10583 to match the target expansion of these operations without fear of
10588 An alias for the machine mode for pointers. On most machines, define
10589 this to be the integer mode corresponding to the width of a hardware
10590 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10591 On some machines you must define this to be one of the partial integer
10592 modes, such as @code{PSImode}.
10594 The width of @code{Pmode} must be at least as large as the value of
10595 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10596 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10600 @defmac FUNCTION_MODE
10601 An alias for the machine mode used for memory references to functions
10602 being called, in @code{call} RTL expressions. On most CISC machines,
10603 where an instruction can begin at any byte address, this should be
10604 @code{QImode}. On most RISC machines, where all instructions have fixed
10605 size and alignment, this should be a mode with the same size and alignment
10606 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10609 @defmac STDC_0_IN_SYSTEM_HEADERS
10610 In normal operation, the preprocessor expands @code{__STDC__} to the
10611 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10612 hosts, like Solaris, the system compiler uses a different convention,
10613 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10614 strict conformance to the C Standard.
10616 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10617 convention when processing system header files, but when processing user
10618 files @code{__STDC__} will always expand to 1.
10621 @defmac NO_IMPLICIT_EXTERN_C
10622 Define this macro if the system header files support C++ as well as C@.
10623 This macro inhibits the usual method of using system header files in
10624 C++, which is to pretend that the file's contents are enclosed in
10625 @samp{extern "C" @{@dots{}@}}.
10630 @defmac REGISTER_TARGET_PRAGMAS ()
10631 Define this macro if you want to implement any target-specific pragmas.
10632 If defined, it is a C expression which makes a series of calls to
10633 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10634 for each pragma. The macro may also do any
10635 setup required for the pragmas.
10637 The primary reason to define this macro is to provide compatibility with
10638 other compilers for the same target. In general, we discourage
10639 definition of target-specific pragmas for GCC@.
10641 If the pragma can be implemented by attributes then you should consider
10642 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10644 Preprocessor macros that appear on pragma lines are not expanded. All
10645 @samp{#pragma} directives that do not match any registered pragma are
10646 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10649 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10650 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10652 Each call to @code{c_register_pragma} or
10653 @code{c_register_pragma_with_expansion} establishes one pragma. The
10654 @var{callback} routine will be called when the preprocessor encounters a
10658 #pragma [@var{space}] @var{name} @dots{}
10661 @var{space} is the case-sensitive namespace of the pragma, or
10662 @code{NULL} to put the pragma in the global namespace. The callback
10663 routine receives @var{pfile} as its first argument, which can be passed
10664 on to cpplib's functions if necessary. You can lex tokens after the
10665 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10666 callback will be silently ignored. The end of the line is indicated by
10667 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10668 arguments of pragmas registered with
10669 @code{c_register_pragma_with_expansion} but not on the arguments of
10670 pragmas registered with @code{c_register_pragma}.
10672 Note that the use of @code{pragma_lex} is specific to the C and C++
10673 compilers. It will not work in the Java or Fortran compilers, or any
10674 other language compilers for that matter. Thus if @code{pragma_lex} is going
10675 to be called from target-specific code, it must only be done so when
10676 building the C and C++ compilers. This can be done by defining the
10677 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10678 target entry in the @file{config.gcc} file. These variables should name
10679 the target-specific, language-specific object file which contains the
10680 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10681 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10682 how to build this object file.
10685 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10686 Define this macro if macros should be expanded in the
10687 arguments of @samp{#pragma pack}.
10690 @deftypevr {Target Hook} bool TARGET_HANDLE_PRAGMA_EXTERN_PREFIX
10691 True if @code{#pragma extern_prefix} is to be supported.
10694 @defmac TARGET_DEFAULT_PACK_STRUCT
10695 If your target requires a structure packing default other than 0 (meaning
10696 the machine default), define this macro to the necessary value (in bytes).
10697 This must be a value that would also be valid to use with
10698 @samp{#pragma pack()} (that is, a small power of two).
10701 @defmac DOLLARS_IN_IDENTIFIERS
10702 Define this macro to control use of the character @samp{$} in
10703 identifier names for the C family of languages. 0 means @samp{$} is
10704 not allowed by default; 1 means it is allowed. 1 is the default;
10705 there is no need to define this macro in that case.
10708 @defmac NO_DOLLAR_IN_LABEL
10709 Define this macro if the assembler does not accept the character
10710 @samp{$} in label names. By default constructors and destructors in
10711 G++ have @samp{$} in the identifiers. If this macro is defined,
10712 @samp{.} is used instead.
10715 @defmac NO_DOT_IN_LABEL
10716 Define this macro if the assembler does not accept the character
10717 @samp{.} in label names. By default constructors and destructors in G++
10718 have names that use @samp{.}. If this macro is defined, these names
10719 are rewritten to avoid @samp{.}.
10722 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10723 Define this macro as a C expression that is nonzero if it is safe for the
10724 delay slot scheduler to place instructions in the delay slot of @var{insn},
10725 even if they appear to use a resource set or clobbered in @var{insn}.
10726 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10727 every @code{call_insn} has this behavior. On machines where some @code{insn}
10728 or @code{jump_insn} is really a function call and hence has this behavior,
10729 you should define this macro.
10731 You need not define this macro if it would always return zero.
10734 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10735 Define this macro as a C expression that is nonzero if it is safe for the
10736 delay slot scheduler to place instructions in the delay slot of @var{insn},
10737 even if they appear to set or clobber a resource referenced in @var{insn}.
10738 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10739 some @code{insn} or @code{jump_insn} is really a function call and its operands
10740 are registers whose use is actually in the subroutine it calls, you should
10741 define this macro. Doing so allows the delay slot scheduler to move
10742 instructions which copy arguments into the argument registers into the delay
10743 slot of @var{insn}.
10745 You need not define this macro if it would always return zero.
10748 @defmac MULTIPLE_SYMBOL_SPACES
10749 Define this macro as a C expression that is nonzero if, in some cases,
10750 global symbols from one translation unit may not be bound to undefined
10751 symbols in another translation unit without user intervention. For
10752 instance, under Microsoft Windows symbols must be explicitly imported
10753 from shared libraries (DLLs).
10755 You need not define this macro if it would always evaluate to zero.
10758 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10759 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10760 any hard regs the port wishes to automatically clobber for an asm.
10761 It should return the result of the last @code{tree_cons} used to add a
10762 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10763 corresponding parameters to the asm and may be inspected to avoid
10764 clobbering a register that is an input or output of the asm. You can use
10765 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10766 for overlap with regards to asm-declared registers.
10769 @defmac MATH_LIBRARY
10770 Define this macro as a C string constant for the linker argument to link
10771 in the system math library, minus the initial @samp{"-l"}, or
10772 @samp{""} if the target does not have a
10773 separate math library.
10775 You need only define this macro if the default of @samp{"m"} is wrong.
10778 @defmac LIBRARY_PATH_ENV
10779 Define this macro as a C string constant for the environment variable that
10780 specifies where the linker should look for libraries.
10782 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10786 @defmac TARGET_POSIX_IO
10787 Define this macro if the target supports the following POSIX@ file
10788 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10789 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10790 to use file locking when exiting a program, which avoids race conditions
10791 if the program has forked. It will also create directories at run-time
10792 for cross-profiling.
10795 @defmac MAX_CONDITIONAL_EXECUTE
10797 A C expression for the maximum number of instructions to execute via
10798 conditional execution instructions instead of a branch. A value of
10799 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10800 1 if it does use cc0.
10803 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10804 Used if the target needs to perform machine-dependent modifications on the
10805 conditionals used for turning basic blocks into conditionally executed code.
10806 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10807 contains information about the currently processed blocks. @var{true_expr}
10808 and @var{false_expr} are the tests that are used for converting the
10809 then-block and the else-block, respectively. Set either @var{true_expr} or
10810 @var{false_expr} to a null pointer if the tests cannot be converted.
10813 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10814 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10815 if-statements into conditions combined by @code{and} and @code{or} operations.
10816 @var{bb} contains the basic block that contains the test that is currently
10817 being processed and about to be turned into a condition.
10820 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10821 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10822 be converted to conditional execution format. @var{ce_info} points to
10823 a data structure, @code{struct ce_if_block}, which contains information
10824 about the currently processed blocks.
10827 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10828 A C expression to perform any final machine dependent modifications in
10829 converting code to conditional execution. The involved basic blocks
10830 can be found in the @code{struct ce_if_block} structure that is pointed
10831 to by @var{ce_info}.
10834 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10835 A C expression to cancel any machine dependent modifications in
10836 converting code to conditional execution. The involved basic blocks
10837 can be found in the @code{struct ce_if_block} structure that is pointed
10838 to by @var{ce_info}.
10841 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10842 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10843 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10846 @defmac IFCVT_EXTRA_FIELDS
10847 If defined, it should expand to a set of field declarations that will be
10848 added to the @code{struct ce_if_block} structure. These should be initialized
10849 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10852 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
10853 If non-null, this hook performs a target-specific pass over the
10854 instruction stream. The compiler will run it at all optimization levels,
10855 just before the point at which it normally does delayed-branch scheduling.
10857 The exact purpose of the hook varies from target to target. Some use
10858 it to do transformations that are necessary for correctness, such as
10859 laying out in-function constant pools or avoiding hardware hazards.
10860 Others use it as an opportunity to do some machine-dependent optimizations.
10862 You need not implement the hook if it has nothing to do. The default
10863 definition is null.
10866 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
10867 Define this hook if you have any machine-specific built-in functions
10868 that need to be defined. It should be a function that performs the
10871 Machine specific built-in functions can be useful to expand special machine
10872 instructions that would otherwise not normally be generated because
10873 they have no equivalent in the source language (for example, SIMD vector
10874 instructions or prefetch instructions).
10876 To create a built-in function, call the function
10877 @code{lang_hooks.builtin_function}
10878 which is defined by the language front end. You can use any type nodes set
10879 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
10880 only language front ends that use those two functions will call
10881 @samp{TARGET_INIT_BUILTINS}.
10884 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
10885 Define this hook if you have any machine-specific built-in functions
10886 that need to be defined. It should be a function that returns the
10887 builtin function declaration for the builtin function code @var{code}.
10888 If there is no such builtin and it cannot be initialized at this time
10889 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10890 If @var{code} is out of range the function should return
10891 @code{error_mark_node}.
10894 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10896 Expand a call to a machine specific built-in function that was set up by
10897 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10898 function call; the result should go to @var{target} if that is
10899 convenient, and have mode @var{mode} if that is convenient.
10900 @var{subtarget} may be used as the target for computing one of
10901 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10902 ignored. This function should return the result of the call to the
10906 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
10907 Select a replacement for a machine specific built-in function that
10908 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10909 @emph{before} regular type checking, and so allows the target to
10910 implement a crude form of function overloading. @var{fndecl} is the
10911 declaration of the built-in function. @var{arglist} is the list of
10912 arguments passed to the built-in function. The result is a
10913 complete expression that implements the operation, usually
10914 another @code{CALL_EXPR}.
10915 @var{arglist} really has type @samp{VEC(tree,gc)*}
10918 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
10919 Fold a call to a machine specific built-in function that was set up by
10920 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10921 built-in function. @var{n_args} is the number of arguments passed to
10922 the function; the arguments themselves are pointed to by @var{argp}.
10923 The result is another tree containing a simplified expression for the
10924 call's result. If @var{ignore} is true the value will be ignored.
10927 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const_rtx @var{insn})
10929 Take an instruction in @var{insn} and return NULL if it is valid within a
10930 low-overhead loop, otherwise return a string explaining why doloop
10931 could not be applied.
10933 Many targets use special registers for low-overhead looping. For any
10934 instruction that clobbers these this function should return a string indicating
10935 the reason why the doloop could not be applied.
10936 By default, the RTL loop optimizer does not use a present doloop pattern for
10937 loops containing function calls or branch on table instructions.
10940 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10942 Take a branch insn in @var{branch1} and another in @var{branch2}.
10943 Return true if redirecting @var{branch1} to the destination of
10944 @var{branch2} is possible.
10946 On some targets, branches may have a limited range. Optimizing the
10947 filling of delay slots can result in branches being redirected, and this
10948 may in turn cause a branch offset to overflow.
10951 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
10952 This target hook returns @code{true} if @var{x} is considered to be commutative.
10953 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10954 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10955 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10958 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
10960 When the initial value of a hard register has been copied in a pseudo
10961 register, it is often not necessary to actually allocate another register
10962 to this pseudo register, because the original hard register or a stack slot
10963 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10964 is called at the start of register allocation once for each hard register
10965 that had its initial value copied by using
10966 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10967 Possible values are @code{NULL_RTX}, if you don't want
10968 to do any special allocation, a @code{REG} rtx---that would typically be
10969 the hard register itself, if it is known not to be clobbered---or a
10971 If you are returning a @code{MEM}, this is only a hint for the allocator;
10972 it might decide to use another register anyways.
10973 You may use @code{current_function_leaf_function} in the hook, functions
10974 that use @code{REG_N_SETS}, to determine if the hard
10975 register in question will not be clobbered.
10976 The default value of this hook is @code{NULL}, which disables any special
10980 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
10981 This target hook returns nonzero if @var{x}, an @code{unspec} or
10982 @code{unspec_volatile} operation, might cause a trap. Targets can use
10983 this hook to enhance precision of analysis for @code{unspec} and
10984 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10985 to analyze inner elements of @var{x} in which case @var{flags} should be
10989 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
10990 The compiler invokes this hook whenever it changes its current function
10991 context (@code{cfun}). You can define this function if
10992 the back end needs to perform any initialization or reset actions on a
10993 per-function basis. For example, it may be used to implement function
10994 attributes that affect register usage or code generation patterns.
10995 The argument @var{decl} is the declaration for the new function context,
10996 and may be null to indicate that the compiler has left a function context
10997 and is returning to processing at the top level.
10998 The default hook function does nothing.
11000 GCC sets @code{cfun} to a dummy function context during initialization of
11001 some parts of the back end. The hook function is not invoked in this
11002 situation; you need not worry about the hook being invoked recursively,
11003 or when the back end is in a partially-initialized state.
11004 @code{cfun} might be @code{NULL} to indicate processing at top level,
11005 outside of any function scope.
11008 @defmac TARGET_OBJECT_SUFFIX
11009 Define this macro to be a C string representing the suffix for object
11010 files on your target machine. If you do not define this macro, GCC will
11011 use @samp{.o} as the suffix for object files.
11014 @defmac TARGET_EXECUTABLE_SUFFIX
11015 Define this macro to be a C string representing the suffix to be
11016 automatically added to executable files on your target machine. If you
11017 do not define this macro, GCC will use the null string as the suffix for
11021 @defmac COLLECT_EXPORT_LIST
11022 If defined, @code{collect2} will scan the individual object files
11023 specified on its command line and create an export list for the linker.
11024 Define this macro for systems like AIX, where the linker discards
11025 object files that are not referenced from @code{main} and uses export
11029 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11030 Define this macro to a C expression representing a variant of the
11031 method call @var{mdecl}, if Java Native Interface (JNI) methods
11032 must be invoked differently from other methods on your target.
11033 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11034 the @code{stdcall} calling convention and this macro is then
11035 defined as this expression:
11038 build_type_attribute_variant (@var{mdecl},
11040 (get_identifier ("stdcall"),
11045 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11046 This target hook returns @code{true} past the point in which new jump
11047 instructions could be created. On machines that require a register for
11048 every jump such as the SHmedia ISA of SH5, this point would typically be
11049 reload, so this target hook should be defined to a function such as:
11053 cannot_modify_jumps_past_reload_p ()
11055 return (reload_completed || reload_in_progress);
11060 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11061 This target hook returns a register class for which branch target register
11062 optimizations should be applied. All registers in this class should be
11063 usable interchangeably. After reload, registers in this class will be
11064 re-allocated and loads will be hoisted out of loops and be subjected
11065 to inter-block scheduling.
11068 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11069 Branch target register optimization will by default exclude callee-saved
11071 that are not already live during the current function; if this target hook
11072 returns true, they will be included. The target code must than make sure
11073 that all target registers in the class returned by
11074 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11075 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11076 epilogues have already been generated. Note, even if you only return
11077 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11078 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11079 to reserve space for caller-saved target registers.
11082 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11083 This target hook returns true if the target supports conditional execution.
11084 This target hook is required only when the target has several different
11085 modes and they have different conditional execution capability, such as ARM.
11088 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop})
11089 This target hook returns a new value for the number of times @var{loop}
11090 should be unrolled. The parameter @var{nunroll} is the number of times
11091 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11092 the loop, which is going to be checked for unrolling. This target hook
11093 is required only when the target has special constraints like maximum
11094 number of memory accesses.
11097 @defmac POWI_MAX_MULTS
11098 If defined, this macro is interpreted as a signed integer C expression
11099 that specifies the maximum number of floating point multiplications
11100 that should be emitted when expanding exponentiation by an integer
11101 constant inline. When this value is defined, exponentiation requiring
11102 more than this number of multiplications is implemented by calling the
11103 system library's @code{pow}, @code{powf} or @code{powl} routines.
11104 The default value places no upper bound on the multiplication count.
11107 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11108 This target hook should register any extra include files for the
11109 target. The parameter @var{stdinc} indicates if normal include files
11110 are present. The parameter @var{sysroot} is the system root directory.
11111 The parameter @var{iprefix} is the prefix for the gcc directory.
11114 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11115 This target hook should register any extra include files for the
11116 target before any standard headers. The parameter @var{stdinc}
11117 indicates if normal include files are present. The parameter
11118 @var{sysroot} is the system root directory. The parameter
11119 @var{iprefix} is the prefix for the gcc directory.
11122 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11123 This target hook should register special include paths for the target.
11124 The parameter @var{path} is the include to register. On Darwin
11125 systems, this is used for Framework includes, which have semantics
11126 that are different from @option{-I}.
11129 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11130 This target macro returns @code{true} if it is safe to use a local alias
11131 for a virtual function @var{fndecl} when constructing thunks,
11132 @code{false} otherwise. By default, the macro returns @code{true} for all
11133 functions, if a target supports aliases (i.e.@: defines
11134 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11137 @defmac TARGET_FORMAT_TYPES
11138 If defined, this macro is the name of a global variable containing
11139 target-specific format checking information for the @option{-Wformat}
11140 option. The default is to have no target-specific format checks.
11143 @defmac TARGET_N_FORMAT_TYPES
11144 If defined, this macro is the number of entries in
11145 @code{TARGET_FORMAT_TYPES}.
11148 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11149 If defined, this macro is the name of a global variable containing
11150 target-specific format overrides for the @option{-Wformat} option. The
11151 default is to have no target-specific format overrides. If defined,
11152 @code{TARGET_FORMAT_TYPES} must be defined, too.
11155 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11156 If defined, this macro specifies the number of entries in
11157 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11160 @defmac TARGET_OVERRIDES_FORMAT_INIT
11161 If defined, this macro specifies the optional initialization
11162 routine for target specific customizations of the system printf
11163 and scanf formatter settings.
11166 @deftypevr {Target Hook} bool TARGET_RELAXED_ORDERING
11167 If set to @code{true}, means that the target's memory model does not
11168 guarantee that loads which do not depend on one another will access
11169 main memory in the order of the instruction stream; if ordering is
11170 important, an explicit memory barrier must be used. This is true of
11171 many recent processors which implement a policy of ``relaxed,''
11172 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11173 and ia64. The default is @code{false}.
11176 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11177 If defined, this macro returns the diagnostic message when it is
11178 illegal to pass argument @var{val} to function @var{funcdecl}
11179 with prototype @var{typelist}.
11182 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11183 If defined, this macro returns the diagnostic message when it is
11184 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11185 if validity should be determined by the front end.
11188 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11189 If defined, this macro returns the diagnostic message when it is
11190 invalid to apply operation @var{op} (where unary plus is denoted by
11191 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11192 if validity should be determined by the front end.
11195 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11196 If defined, this macro returns the diagnostic message when it is
11197 invalid to apply operation @var{op} to operands of types @var{type1}
11198 and @var{type2}, or @code{NULL} if validity should be determined by
11202 @deftypefn {Target Hook} {const char *} TARGET_INVALID_PARAMETER_TYPE (const_tree @var{type})
11203 If defined, this macro returns the diagnostic message when it is
11204 invalid for functions to include parameters of type @var{type},
11205 or @code{NULL} if validity should be determined by
11206 the front end. This is currently used only by the C and C++ front ends.
11209 @deftypefn {Target Hook} {const char *} TARGET_INVALID_RETURN_TYPE (const_tree @var{type})
11210 If defined, this macro returns the diagnostic message when it is
11211 invalid for functions to have return type @var{type},
11212 or @code{NULL} if validity should be determined by
11213 the front end. This is currently used only by the C and C++ front ends.
11216 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11217 If defined, this target hook returns the type to which values of
11218 @var{type} should be promoted when they appear in expressions,
11219 analogous to the integer promotions, or @code{NULL_TREE} to use the
11220 front end's normal promotion rules. This hook is useful when there are
11221 target-specific types with special promotion rules.
11222 This is currently used only by the C and C++ front ends.
11225 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11226 If defined, this hook returns the result of converting @var{expr} to
11227 @var{type}. It should return the converted expression,
11228 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11229 This hook is useful when there are target-specific types with special
11231 This is currently used only by the C and C++ front ends.
11234 @defmac TARGET_USE_JCR_SECTION
11235 This macro determines whether to use the JCR section to register Java
11236 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11237 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11241 This macro determines the size of the objective C jump buffer for the
11242 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11245 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11246 Define this macro if any target-specific attributes need to be attached
11247 to the functions in @file{libgcc} that provide low-level support for
11248 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11249 and the associated definitions of those functions.
11252 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11253 Define this macro to update the current function stack boundary if
11257 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11258 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11259 different argument pointer register is needed to access the function's
11260 argument list due to stack realignment. Return @code{NULL} if no DRAP
11264 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11265 When optimization is disabled, this hook indicates whether or not
11266 arguments should be allocated to stack slots. Normally, GCC allocates
11267 stacks slots for arguments when not optimizing in order to make
11268 debugging easier. However, when a function is declared with
11269 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11270 cannot safely move arguments from the registers in which they are passed
11271 to the stack. Therefore, this hook should return true in general, but
11272 false for naked functions. The default implementation always returns true.
11275 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11276 On some architectures it can take multiple instructions to synthesize
11277 a constant. If there is another constant already in a register that
11278 is close enough in value then it is preferable that the new constant
11279 is computed from this register using immediate addition or
11280 subtraction. We accomplish this through CSE. Besides the value of
11281 the constant we also add a lower and an upper constant anchor to the
11282 available expressions. These are then queried when encountering new
11283 constants. The anchors are computed by rounding the constant up and
11284 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11285 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11286 accepted by immediate-add plus one. We currently assume that the
11287 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11288 MIPS, where add-immediate takes a 16-bit signed value,
11289 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11290 is zero, which disables this optimization. @end deftypevr