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, 2012
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{NATIVE_SYSTEM_HEADER_DIR} (set in
474 @file{config.gcc}, normally @file{/usr/include}) in the search order.
476 Cross compilers do not search either @file{/usr/local/include} or its
480 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
481 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
482 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
483 If you do not define this macro, no component is used.
486 @defmac INCLUDE_DEFAULTS
487 Define this macro if you wish to override the entire default search path
488 for include files. For a native compiler, the default search path
489 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
490 @code{GPLUSPLUS_INCLUDE_DIR}, and
491 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
492 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
493 and specify private search areas for GCC@. The directory
494 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
496 The definition should be an initializer for an array of structures.
497 Each array element should have four elements: the directory name (a
498 string constant), the component name (also a string constant), a flag
499 for C++-only directories,
500 and a flag showing that the includes in the directory don't need to be
501 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
502 the array with a null element.
504 The component name denotes what GNU package the include file is part of,
505 if any, in all uppercase letters. For example, it might be @samp{GCC}
506 or @samp{BINUTILS}. If the package is part of a vendor-supplied
507 operating system, code the component name as @samp{0}.
509 For example, here is the definition used for VAX/VMS:
512 #define INCLUDE_DEFAULTS \
514 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
515 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
516 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
523 Here is the order of prefixes tried for exec files:
527 Any prefixes specified by the user with @option{-B}.
530 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
531 is not set and the compiler has not been installed in the configure-time
532 @var{prefix}, the location in which the compiler has actually been installed.
535 The directories specified by the environment variable @code{COMPILER_PATH}.
538 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
539 in the configured-time @var{prefix}.
542 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
545 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
548 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
552 Here is the order of prefixes tried for startfiles:
556 Any prefixes specified by the user with @option{-B}.
559 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
560 value based on the installed toolchain location.
563 The directories specified by the environment variable @code{LIBRARY_PATH}
564 (or port-specific name; native only, cross compilers do not use this).
567 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
568 in the configured @var{prefix} or this is a native compiler.
571 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
574 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
578 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
579 native compiler, or we have a target system root.
582 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
583 native compiler, or we have a target system root.
586 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
587 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
588 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
591 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
592 compiler, or we have a target system root. The default for this macro is
596 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
597 compiler, or we have a target system root. The default for this macro is
601 @node Run-time Target
602 @section Run-time Target Specification
603 @cindex run-time target specification
604 @cindex predefined macros
605 @cindex target specifications
607 @c prevent bad page break with this line
608 Here are run-time target specifications.
610 @defmac TARGET_CPU_CPP_BUILTINS ()
611 This function-like macro expands to a block of code that defines
612 built-in preprocessor macros and assertions for the target CPU, using
613 the functions @code{builtin_define}, @code{builtin_define_std} and
614 @code{builtin_assert}. When the front end
615 calls this macro it provides a trailing semicolon, and since it has
616 finished command line option processing your code can use those
619 @code{builtin_assert} takes a string in the form you pass to the
620 command-line option @option{-A}, such as @code{cpu=mips}, and creates
621 the assertion. @code{builtin_define} takes a string in the form
622 accepted by option @option{-D} and unconditionally defines the macro.
624 @code{builtin_define_std} takes a string representing the name of an
625 object-like macro. If it doesn't lie in the user's namespace,
626 @code{builtin_define_std} defines it unconditionally. Otherwise, it
627 defines a version with two leading underscores, and another version
628 with two leading and trailing underscores, and defines the original
629 only if an ISO standard was not requested on the command line. For
630 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
631 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
632 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
633 defines only @code{_ABI64}.
635 You can also test for the C dialect being compiled. The variable
636 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
637 or @code{clk_objective_c}. Note that if we are preprocessing
638 assembler, this variable will be @code{clk_c} but the function-like
639 macro @code{preprocessing_asm_p()} will return true, so you might want
640 to check for that first. If you need to check for strict ANSI, the
641 variable @code{flag_iso} can be used. The function-like macro
642 @code{preprocessing_trad_p()} can be used to check for traditional
646 @defmac TARGET_OS_CPP_BUILTINS ()
647 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
648 and is used for the target operating system instead.
651 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
652 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
653 and is used for the target object format. @file{elfos.h} uses this
654 macro to define @code{__ELF__}, so you probably do not need to define
658 @deftypevar {extern int} target_flags
659 This variable is declared in @file{options.h}, which is included before
660 any target-specific headers.
663 @deftypevr {Common Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
664 This variable specifies the initial value of @code{target_flags}.
665 Its default setting is 0.
668 @cindex optional hardware or system features
669 @cindex features, optional, in system conventions
671 @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})
672 This hook is called whenever the user specifies one of the
673 target-specific options described by the @file{.opt} definition files
674 (@pxref{Options}). It has the opportunity to do some option-specific
675 processing and should return true if the option is valid. The default
676 definition does nothing but return true.
678 @var{decoded} specifies the option and its arguments. @var{opts} and
679 @var{opts_set} are the @code{gcc_options} structures to be used for
680 storing option state, and @var{loc} is the location at which the
681 option was passed (@code{UNKNOWN_LOCATION} except for options passed
685 @deftypefn {C Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
686 This target hook is called whenever the user specifies one of the
687 target-specific C language family options described by the @file{.opt}
688 definition files(@pxref{Options}). It has the opportunity to do some
689 option-specific processing and should return true if the option is
690 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
691 default definition does nothing but return false.
693 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
694 options. However, if processing an option requires routines that are
695 only available in the C (and related language) front ends, then you
696 should use @code{TARGET_HANDLE_C_OPTION} instead.
699 @deftypefn {C Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING_OBJECT (tree @var{string})
700 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.
703 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE (const char *@var{classname})
704 Declare that Objective C class @var{classname} is referenced by the current TU.
707 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_CLASS_DEFINITION (const char *@var{classname})
708 Declare that Objective C class @var{classname} is defined by the current TU.
711 @deftypefn {C Target Hook} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref})
712 If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object.
715 @deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list})
716 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.
719 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
720 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
721 but is called when the optimize level is changed via an attribute or
722 pragma or when it is reset at the end of the code affected by the
723 attribute or pragma. It is not called at the beginning of compilation
724 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
725 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
726 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
729 @defmac C_COMMON_OVERRIDE_OPTIONS
730 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
731 but is only used in the C
732 language frontends (C, Objective-C, C++, Objective-C++) and so can be
733 used to alter option flag variables which only exist in those
737 @deftypevr {Common Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
738 Some machines may desire to change what optimizations are performed for
739 various optimization levels. This variable, if defined, describes
740 options to enable at particular sets of optimization levels. These
741 options are processed once
742 just after the optimization level is determined and before the remainder
743 of the command options have been parsed, so may be overridden by other
744 options passed explicitly.
746 This processing is run once at program startup and when the optimization
747 options are changed via @code{#pragma GCC optimize} or by using the
748 @code{optimize} attribute.
751 @deftypefn {Common Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
752 Set target-dependent initial values of fields in @var{opts}.
755 @deftypefn {Common Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
756 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
759 @defmac SWITCHABLE_TARGET
760 Some targets need to switch between substantially different subtargets
761 during compilation. For example, the MIPS target has one subtarget for
762 the traditional MIPS architecture and another for MIPS16. Source code
763 can switch between these two subarchitectures using the @code{mips16}
764 and @code{nomips16} attributes.
766 Such subtargets can differ in things like the set of available
767 registers, the set of available instructions, the costs of various
768 operations, and so on. GCC caches a lot of this type of information
769 in global variables, and recomputing them for each subtarget takes a
770 significant amount of time. The compiler therefore provides a facility
771 for maintaining several versions of the global variables and quickly
772 switching between them; see @file{target-globals.h} for details.
774 Define this macro to 1 if your target needs this facility. The default
778 @node Per-Function Data
779 @section Defining data structures for per-function information.
780 @cindex per-function data
781 @cindex data structures
783 If the target needs to store information on a per-function basis, GCC
784 provides a macro and a couple of variables to allow this. Note, just
785 using statics to store the information is a bad idea, since GCC supports
786 nested functions, so you can be halfway through encoding one function
787 when another one comes along.
789 GCC defines a data structure called @code{struct function} which
790 contains all of the data specific to an individual function. This
791 structure contains a field called @code{machine} whose type is
792 @code{struct machine_function *}, which can be used by targets to point
793 to their own specific data.
795 If a target needs per-function specific data it should define the type
796 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
797 This macro should be used to initialize the function pointer
798 @code{init_machine_status}. This pointer is explained below.
800 One typical use of per-function, target specific data is to create an
801 RTX to hold the register containing the function's return address. This
802 RTX can then be used to implement the @code{__builtin_return_address}
803 function, for level 0.
805 Note---earlier implementations of GCC used a single data area to hold
806 all of the per-function information. Thus when processing of a nested
807 function began the old per-function data had to be pushed onto a
808 stack, and when the processing was finished, it had to be popped off the
809 stack. GCC used to provide function pointers called
810 @code{save_machine_status} and @code{restore_machine_status} to handle
811 the saving and restoring of the target specific information. Since the
812 single data area approach is no longer used, these pointers are no
815 @defmac INIT_EXPANDERS
816 Macro called to initialize any target specific information. This macro
817 is called once per function, before generation of any RTL has begun.
818 The intention of this macro is to allow the initialization of the
819 function pointer @code{init_machine_status}.
822 @deftypevar {void (*)(struct function *)} init_machine_status
823 If this function pointer is non-@code{NULL} it will be called once per
824 function, before function compilation starts, in order to allow the
825 target to perform any target specific initialization of the
826 @code{struct function} structure. It is intended that this would be
827 used to initialize the @code{machine} of that structure.
829 @code{struct machine_function} structures are expected to be freed by GC@.
830 Generally, any memory that they reference must be allocated by using
831 GC allocation, including the structure itself.
835 @section Storage Layout
836 @cindex storage layout
838 Note that the definitions of the macros in this table which are sizes or
839 alignments measured in bits do not need to be constant. They can be C
840 expressions that refer to static variables, such as the @code{target_flags}.
841 @xref{Run-time Target}.
843 @defmac BITS_BIG_ENDIAN
844 Define this macro to have the value 1 if the most significant bit in a
845 byte has the lowest number; otherwise define it to have the value zero.
846 This means that bit-field instructions count from the most significant
847 bit. If the machine has no bit-field instructions, then this must still
848 be defined, but it doesn't matter which value it is defined to. This
849 macro need not be a constant.
851 This macro does not affect the way structure fields are packed into
852 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
855 @defmac BYTES_BIG_ENDIAN
856 Define this macro to have the value 1 if the most significant byte in a
857 word has the lowest number. This macro need not be a constant.
860 @defmac WORDS_BIG_ENDIAN
861 Define this macro to have the value 1 if, in a multiword object, the
862 most significant word has the lowest number. This applies to both
863 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
864 order of words in memory is not the same as the order in registers. This
865 macro need not be a constant.
868 @defmac REG_WORDS_BIG_ENDIAN
869 On some machines, the order of words in a multiword object differs between
870 registers in memory. In such a situation, define this macro to describe
871 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
872 the order of words in memory.
875 @defmac FLOAT_WORDS_BIG_ENDIAN
876 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
877 @code{TFmode} floating point numbers are stored in memory with the word
878 containing the sign bit at the lowest address; otherwise define it to
879 have the value 0. This macro need not be a constant.
881 You need not define this macro if the ordering is the same as for
885 @defmac BITS_PER_UNIT
886 Define this macro to be the number of bits in an addressable storage
887 unit (byte). If you do not define this macro the default is 8.
890 @defmac BITS_PER_WORD
891 Number of bits in a word. If you do not define this macro, the default
892 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
895 @defmac MAX_BITS_PER_WORD
896 Maximum number of bits in a word. If this is undefined, the default is
897 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
898 largest value that @code{BITS_PER_WORD} can have at run-time.
901 @defmac UNITS_PER_WORD
902 Number of storage units in a word; normally the size of a general-purpose
903 register, a power of two from 1 or 8.
906 @defmac MIN_UNITS_PER_WORD
907 Minimum number of units in a word. If this is undefined, the default is
908 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
909 smallest value that @code{UNITS_PER_WORD} can have at run-time.
913 Width of a pointer, in bits. You must specify a value no wider than the
914 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
915 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
916 a value the default is @code{BITS_PER_WORD}.
919 @defmac POINTERS_EXTEND_UNSIGNED
920 A C expression that determines how pointers should be extended from
921 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
922 greater than zero if pointers should be zero-extended, zero if they
923 should be sign-extended, and negative if some other sort of conversion
924 is needed. In the last case, the extension is done by the target's
925 @code{ptr_extend} instruction.
927 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
928 and @code{word_mode} are all the same width.
931 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
932 A macro to update @var{m} and @var{unsignedp} when an object whose type
933 is @var{type} and which has the specified mode and signedness is to be
934 stored in a register. This macro is only called when @var{type} is a
937 On most RISC machines, which only have operations that operate on a full
938 register, define this macro to set @var{m} to @code{word_mode} if
939 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
940 cases, only integer modes should be widened because wider-precision
941 floating-point operations are usually more expensive than their narrower
944 For most machines, the macro definition does not change @var{unsignedp}.
945 However, some machines, have instructions that preferentially handle
946 either signed or unsigned quantities of certain modes. For example, on
947 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
948 sign-extend the result to 64 bits. On such machines, set
949 @var{unsignedp} according to which kind of extension is more efficient.
951 Do not define this macro if it would never modify @var{m}.
954 @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})
955 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
956 function return values. The target hook should return the new mode
957 and possibly change @code{*@var{punsignedp}} if the promotion should
958 change signedness. This function is called only for scalar @emph{or
961 @var{for_return} allows to distinguish the promotion of arguments and
962 return values. If it is @code{1}, a return value is being promoted and
963 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
964 If it is @code{2}, the returned mode should be that of the register in
965 which an incoming parameter is copied, or the outgoing result is computed;
966 then the hook should return the same mode as @code{promote_mode}, though
967 the signedness may be different.
969 @var{type} can be NULL when promoting function arguments of libcalls.
971 The default is to not promote arguments and return values. You can
972 also define the hook to @code{default_promote_function_mode_always_promote}
973 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
976 @defmac PARM_BOUNDARY
977 Normal alignment required for function parameters on the stack, in
978 bits. All stack parameters receive at least this much alignment
979 regardless of data type. On most machines, this is the same as the
983 @defmac STACK_BOUNDARY
984 Define this macro to the minimum alignment enforced by hardware for the
985 stack pointer on this machine. The definition is a C expression for the
986 desired alignment (measured in bits). This value is used as a default
987 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
988 this should be the same as @code{PARM_BOUNDARY}.
991 @defmac PREFERRED_STACK_BOUNDARY
992 Define this macro if you wish to preserve a certain alignment for the
993 stack pointer, greater than what the hardware enforces. The definition
994 is a C expression for the desired alignment (measured in bits). This
995 macro must evaluate to a value equal to or larger than
996 @code{STACK_BOUNDARY}.
999 @defmac INCOMING_STACK_BOUNDARY
1000 Define this macro if the incoming stack boundary may be different
1001 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
1002 to a value equal to or larger than @code{STACK_BOUNDARY}.
1005 @defmac FUNCTION_BOUNDARY
1006 Alignment required for a function entry point, in bits.
1009 @defmac BIGGEST_ALIGNMENT
1010 Biggest alignment that any data type can require on this machine, in
1011 bits. Note that this is not the biggest alignment that is supported,
1012 just the biggest alignment that, when violated, may cause a fault.
1015 @defmac MALLOC_ABI_ALIGNMENT
1016 Alignment, in bits, a C conformant malloc implementation has to
1017 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1020 @defmac ATTRIBUTE_ALIGNED_VALUE
1021 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1022 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1025 @defmac MINIMUM_ATOMIC_ALIGNMENT
1026 If defined, the smallest alignment, in bits, that can be given to an
1027 object that can be referenced in one operation, without disturbing any
1028 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1029 on machines that don't have byte or half-word store operations.
1032 @defmac BIGGEST_FIELD_ALIGNMENT
1033 Biggest alignment that any structure or union field can require on this
1034 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1035 structure and union fields only, unless the field alignment has been set
1036 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1039 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1040 An expression for the alignment of a structure field @var{field} if the
1041 alignment computed in the usual way (including applying of
1042 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1043 alignment) is @var{computed}. It overrides alignment only if the
1044 field alignment has not been set by the
1045 @code{__attribute__ ((aligned (@var{n})))} construct.
1048 @defmac MAX_STACK_ALIGNMENT
1049 Biggest stack alignment guaranteed by the backend. Use this macro
1050 to specify the maximum alignment of a variable on stack.
1052 If not defined, the default value is @code{STACK_BOUNDARY}.
1054 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1055 @c But the fix for PR 32893 indicates that we can only guarantee
1056 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1057 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1060 @defmac MAX_OFILE_ALIGNMENT
1061 Biggest alignment supported by the object file format of this machine.
1062 Use this macro to limit the alignment which can be specified using the
1063 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1064 the default value is @code{BIGGEST_ALIGNMENT}.
1066 On systems that use ELF, the default (in @file{config/elfos.h}) is
1067 the largest supported 32-bit ELF section alignment representable on
1068 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1069 On 32-bit ELF the largest supported section alignment in bits is
1070 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1073 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1074 If defined, a C expression to compute the alignment for a variable in
1075 the static store. @var{type} is the data type, and @var{basic-align} is
1076 the alignment that the object would ordinarily have. The value of this
1077 macro is used instead of that alignment to align the object.
1079 If this macro is not defined, then @var{basic-align} is used.
1082 One use of this macro is to increase alignment of medium-size data to
1083 make it all fit in fewer cache lines. Another is to cause character
1084 arrays to be word-aligned so that @code{strcpy} calls that copy
1085 constants to character arrays can be done inline.
1088 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1089 If defined, a C expression to compute the alignment given to a constant
1090 that is being placed in memory. @var{constant} is the constant and
1091 @var{basic-align} is the alignment that the object would ordinarily
1092 have. The value of this macro is used instead of that alignment to
1095 If this macro is not defined, then @var{basic-align} is used.
1097 The typical use of this macro is to increase alignment for string
1098 constants to be word aligned so that @code{strcpy} calls that copy
1099 constants can be done inline.
1102 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1103 If defined, a C expression to compute the alignment for a variable in
1104 the local store. @var{type} is the data type, and @var{basic-align} is
1105 the alignment that the object would ordinarily have. The value of this
1106 macro is used instead of that alignment to align the object.
1108 If this macro is not defined, then @var{basic-align} is used.
1110 One use of this macro is to increase alignment of medium-size data to
1111 make it all fit in fewer cache lines.
1113 If the value of this macro has a type, it should be an unsigned type.
1116 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_VECTOR_ALIGNMENT (const_tree @var{type})
1117 This hook can be used to define the alignment for a vector of type
1118 @var{type}, in order to comply with a platform ABI. The default is to
1119 require natural alignment for vector types. The alignment returned by
1120 this hook must be a power-of-two multiple of the default alignment of
1121 the vector element type.
1124 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1125 If defined, a C expression to compute the alignment for stack slot.
1126 @var{type} is the data type, @var{mode} is the widest mode available,
1127 and @var{basic-align} is the alignment that the slot would ordinarily
1128 have. The value of this macro is used instead of that alignment to
1131 If this macro is not defined, then @var{basic-align} is used when
1132 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1135 This macro is to set alignment of stack slot to the maximum alignment
1136 of all possible modes which the slot may have.
1138 If the value of this macro has a type, it should be an unsigned type.
1141 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1142 If defined, a C expression to compute the alignment for a local
1143 variable @var{decl}.
1145 If this macro is not defined, then
1146 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1149 One use of this macro is to increase alignment of medium-size data to
1150 make it all fit in fewer cache lines.
1152 If the value of this macro has a type, it should be an unsigned type.
1155 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1156 If defined, a C expression to compute the minimum required alignment
1157 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1158 @var{mode}, assuming normal alignment @var{align}.
1160 If this macro is not defined, then @var{align} will be used.
1163 @defmac EMPTY_FIELD_BOUNDARY
1164 Alignment in bits to be given to a structure bit-field that follows an
1165 empty field such as @code{int : 0;}.
1167 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1170 @defmac STRUCTURE_SIZE_BOUNDARY
1171 Number of bits which any structure or union's size must be a multiple of.
1172 Each structure or union's size is rounded up to a multiple of this.
1174 If you do not define this macro, the default is the same as
1175 @code{BITS_PER_UNIT}.
1178 @defmac STRICT_ALIGNMENT
1179 Define this macro to be the value 1 if instructions will fail to work
1180 if given data not on the nominal alignment. If instructions will merely
1181 go slower in that case, define this macro as 0.
1184 @defmac PCC_BITFIELD_TYPE_MATTERS
1185 Define this if you wish to imitate the way many other C compilers handle
1186 alignment of bit-fields and the structures that contain them.
1188 The behavior is that the type written for a named bit-field (@code{int},
1189 @code{short}, or other integer type) imposes an alignment for the entire
1190 structure, as if the structure really did contain an ordinary field of
1191 that type. In addition, the bit-field is placed within the structure so
1192 that it would fit within such a field, not crossing a boundary for it.
1194 Thus, on most machines, a named bit-field whose type is written as
1195 @code{int} would not cross a four-byte boundary, and would force
1196 four-byte alignment for the whole structure. (The alignment used may
1197 not be four bytes; it is controlled by the other alignment parameters.)
1199 An unnamed bit-field will not affect the alignment of the containing
1202 If the macro is defined, its definition should be a C expression;
1203 a nonzero value for the expression enables this behavior.
1205 Note that if this macro is not defined, or its value is zero, some
1206 bit-fields may cross more than one alignment boundary. The compiler can
1207 support such references if there are @samp{insv}, @samp{extv}, and
1208 @samp{extzv} insns that can directly reference memory.
1210 The other known way of making bit-fields work is to define
1211 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1212 Then every structure can be accessed with fullwords.
1214 Unless the machine has bit-field instructions or you define
1215 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1216 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1218 If your aim is to make GCC use the same conventions for laying out
1219 bit-fields as are used by another compiler, here is how to investigate
1220 what the other compiler does. Compile and run this program:
1239 printf ("Size of foo1 is %d\n",
1240 sizeof (struct foo1));
1241 printf ("Size of foo2 is %d\n",
1242 sizeof (struct foo2));
1247 If this prints 2 and 5, then the compiler's behavior is what you would
1248 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1251 @defmac BITFIELD_NBYTES_LIMITED
1252 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1253 to aligning a bit-field within the structure.
1256 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1257 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1258 whether unnamed bitfields affect the alignment of the containing
1259 structure. The hook should return true if the structure should inherit
1260 the alignment requirements of an unnamed bitfield's type.
1263 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1264 This target hook should return @code{true} if accesses to volatile bitfields
1265 should use the narrowest mode possible. It should return @code{false} if
1266 these accesses should use the bitfield container type.
1268 The default is @code{!TARGET_STRICT_ALIGN}.
1271 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1272 Return 1 if a structure or array containing @var{field} should be accessed using
1275 If @var{field} is the only field in the structure, @var{mode} is its
1276 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1277 case where structures of one field would require the structure's mode to
1278 retain the field's mode.
1280 Normally, this is not needed.
1283 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1284 Define this macro as an expression for the alignment of a type (given
1285 by @var{type} as a tree node) if the alignment computed in the usual
1286 way is @var{computed} and the alignment explicitly specified was
1289 The default is to use @var{specified} if it is larger; otherwise, use
1290 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1293 @defmac MAX_FIXED_MODE_SIZE
1294 An integer expression for the size in bits of the largest integer
1295 machine mode that should actually be used. All integer machine modes of
1296 this size or smaller can be used for structures and unions with the
1297 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1298 (DImode)} is assumed.
1301 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1302 If defined, an expression of type @code{enum machine_mode} that
1303 specifies the mode of the save area operand of a
1304 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1305 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1306 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1307 having its mode specified.
1309 You need not define this macro if it always returns @code{Pmode}. You
1310 would most commonly define this macro if the
1311 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1315 @defmac STACK_SIZE_MODE
1316 If defined, an expression of type @code{enum machine_mode} that
1317 specifies the mode of the size increment operand of an
1318 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1320 You need not define this macro if it always returns @code{word_mode}.
1321 You would most commonly define this macro if the @code{allocate_stack}
1322 pattern needs to support both a 32- and a 64-bit mode.
1325 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE (void)
1326 This target hook should return the mode to be used for the return value
1327 of compare instructions expanded to libgcc calls. If not defined
1328 @code{word_mode} is returned which is the right choice for a majority of
1332 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1333 This target hook should return the mode to be used for the shift count operand
1334 of shift instructions expanded to libgcc calls. If not defined
1335 @code{word_mode} is returned which is the right choice for a majority of
1339 @deftypefn {Target Hook} {enum machine_mode} TARGET_UNWIND_WORD_MODE (void)
1340 Return machine mode to be used for @code{_Unwind_Word} type.
1341 The default is to use @code{word_mode}.
1344 @defmac ROUND_TOWARDS_ZERO
1345 If defined, this macro should be true if the prevailing rounding
1346 mode is towards zero.
1348 Defining this macro only affects the way @file{libgcc.a} emulates
1349 floating-point arithmetic.
1351 Not defining this macro is equivalent to returning zero.
1354 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1355 This macro should return true if floats with @var{size}
1356 bits do not have a NaN or infinity representation, but use the largest
1357 exponent for normal numbers instead.
1359 Defining this macro only affects the way @file{libgcc.a} emulates
1360 floating-point arithmetic.
1362 The default definition of this macro returns false for all sizes.
1365 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1366 This target hook returns @code{true} if bit-fields in the given
1367 @var{record_type} are to be laid out following the rules of Microsoft
1368 Visual C/C++, namely: (i) a bit-field won't share the same storage
1369 unit with the previous bit-field if their underlying types have
1370 different sizes, and the bit-field will be aligned to the highest
1371 alignment of the underlying types of itself and of the previous
1372 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1373 the whole enclosing structure, even if it is unnamed; except that
1374 (iii) a zero-sized bit-field will be disregarded unless it follows
1375 another bit-field of nonzero size. If this hook returns @code{true},
1376 other macros that control bit-field layout are ignored.
1378 When a bit-field is inserted into a packed record, the whole size
1379 of the underlying type is used by one or more same-size adjacent
1380 bit-fields (that is, if its long:3, 32 bits is used in the record,
1381 and any additional adjacent long bit-fields are packed into the same
1382 chunk of 32 bits. However, if the size changes, a new field of that
1383 size is allocated). In an unpacked record, this is the same as using
1384 alignment, but not equivalent when packing.
1386 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1387 the latter will take precedence. If @samp{__attribute__((packed))} is
1388 used on a single field when MS bit-fields are in use, it will take
1389 precedence for that field, but the alignment of the rest of the structure
1390 may affect its placement.
1393 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1394 Returns true if the target supports decimal floating point.
1397 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1398 Returns true if the target supports fixed-point arithmetic.
1401 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1402 This hook is called just before expansion into rtl, allowing the target
1403 to perform additional initializations or analysis before the expansion.
1404 For example, the rs6000 port uses it to allocate a scratch stack slot
1405 for use in copying SDmode values between memory and floating point
1406 registers whenever the function being expanded has any SDmode
1410 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1411 This hook allows the backend to perform additional instantiations on rtl
1412 that are not actually in any insns yet, but will be later.
1415 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1416 If your target defines any fundamental types, or any types your target
1417 uses should be mangled differently from the default, define this hook
1418 to return the appropriate encoding for these types as part of a C++
1419 mangled name. The @var{type} argument is the tree structure representing
1420 the type to be mangled. The hook may be applied to trees which are
1421 not target-specific fundamental types; it should return @code{NULL}
1422 for all such types, as well as arguments it does not recognize. If the
1423 return value is not @code{NULL}, it must point to a statically-allocated
1426 Target-specific fundamental types might be new fundamental types or
1427 qualified versions of ordinary fundamental types. Encode new
1428 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1429 is the name used for the type in source code, and @var{n} is the
1430 length of @var{name} in decimal. Encode qualified versions of
1431 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1432 @var{name} is the name used for the type qualifier in source code,
1433 @var{n} is the length of @var{name} as above, and @var{code} is the
1434 code used to represent the unqualified version of this type. (See
1435 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1436 codes.) In both cases the spaces are for clarity; do not include any
1437 spaces in your string.
1439 This hook is applied to types prior to typedef resolution. If the mangled
1440 name for a particular type depends only on that type's main variant, you
1441 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1444 The default version of this hook always returns @code{NULL}, which is
1445 appropriate for a target that does not define any new fundamental
1450 @section Layout of Source Language Data Types
1452 These macros define the sizes and other characteristics of the standard
1453 basic data types used in programs being compiled. Unlike the macros in
1454 the previous section, these apply to specific features of C and related
1455 languages, rather than to fundamental aspects of storage layout.
1457 @defmac INT_TYPE_SIZE
1458 A C expression for the size in bits of the type @code{int} on the
1459 target machine. If you don't define this, the default is one word.
1462 @defmac SHORT_TYPE_SIZE
1463 A C expression for the size in bits of the type @code{short} on the
1464 target machine. If you don't define this, the default is half a word.
1465 (If this would be less than one storage unit, it is rounded up to one
1469 @defmac LONG_TYPE_SIZE
1470 A C expression for the size in bits of the type @code{long} on the
1471 target machine. If you don't define this, the default is one word.
1474 @defmac ADA_LONG_TYPE_SIZE
1475 On some machines, the size used for the Ada equivalent of the type
1476 @code{long} by a native Ada compiler differs from that used by C@. In
1477 that situation, define this macro to be a C expression to be used for
1478 the size of that type. If you don't define this, the default is the
1479 value of @code{LONG_TYPE_SIZE}.
1482 @defmac LONG_LONG_TYPE_SIZE
1483 A C expression for the size in bits of the type @code{long long} on the
1484 target machine. If you don't define this, the default is two
1485 words. If you want to support GNU Ada on your machine, the value of this
1486 macro must be at least 64.
1489 @defmac CHAR_TYPE_SIZE
1490 A C expression for the size in bits of the type @code{char} on the
1491 target machine. If you don't define this, the default is
1492 @code{BITS_PER_UNIT}.
1495 @defmac BOOL_TYPE_SIZE
1496 A C expression for the size in bits of the C++ type @code{bool} and
1497 C99 type @code{_Bool} on the target machine. If you don't define
1498 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1501 @defmac FLOAT_TYPE_SIZE
1502 A C expression for the size in bits of the type @code{float} on the
1503 target machine. If you don't define this, the default is one word.
1506 @defmac DOUBLE_TYPE_SIZE
1507 A C expression for the size in bits of the type @code{double} on the
1508 target machine. If you don't define this, the default is two
1512 @defmac LONG_DOUBLE_TYPE_SIZE
1513 A C expression for the size in bits of the type @code{long double} on
1514 the target machine. If you don't define this, the default is two
1518 @defmac SHORT_FRACT_TYPE_SIZE
1519 A C expression for the size in bits of the type @code{short _Fract} on
1520 the target machine. If you don't define this, the default is
1521 @code{BITS_PER_UNIT}.
1524 @defmac FRACT_TYPE_SIZE
1525 A C expression for the size in bits of the type @code{_Fract} on
1526 the target machine. If you don't define this, the default is
1527 @code{BITS_PER_UNIT * 2}.
1530 @defmac LONG_FRACT_TYPE_SIZE
1531 A C expression for the size in bits of the type @code{long _Fract} on
1532 the target machine. If you don't define this, the default is
1533 @code{BITS_PER_UNIT * 4}.
1536 @defmac LONG_LONG_FRACT_TYPE_SIZE
1537 A C expression for the size in bits of the type @code{long long _Fract} on
1538 the target machine. If you don't define this, the default is
1539 @code{BITS_PER_UNIT * 8}.
1542 @defmac SHORT_ACCUM_TYPE_SIZE
1543 A C expression for the size in bits of the type @code{short _Accum} on
1544 the target machine. If you don't define this, the default is
1545 @code{BITS_PER_UNIT * 2}.
1548 @defmac ACCUM_TYPE_SIZE
1549 A C expression for the size in bits of the type @code{_Accum} on
1550 the target machine. If you don't define this, the default is
1551 @code{BITS_PER_UNIT * 4}.
1554 @defmac LONG_ACCUM_TYPE_SIZE
1555 A C expression for the size in bits of the type @code{long _Accum} on
1556 the target machine. If you don't define this, the default is
1557 @code{BITS_PER_UNIT * 8}.
1560 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1561 A C expression for the size in bits of the type @code{long long _Accum} on
1562 the target machine. If you don't define this, the default is
1563 @code{BITS_PER_UNIT * 16}.
1566 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1567 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1568 if you want routines in @file{libgcc2.a} for a size other than
1569 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1570 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1573 @defmac LIBGCC2_HAS_DF_MODE
1574 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1575 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1576 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1577 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1578 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1582 @defmac LIBGCC2_HAS_XF_MODE
1583 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1584 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1585 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1586 is 80 then the default is 1, otherwise it is 0.
1589 @defmac LIBGCC2_HAS_TF_MODE
1590 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1591 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1592 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1593 is 128 then the default is 1, otherwise it is 0.
1596 @defmac LIBGCC2_GNU_PREFIX
1597 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1598 hook and should be defined if that hook is overriden to be true. It
1599 causes function names in libgcc to be changed to use a @code{__gnu_}
1600 prefix for their name rather than the default @code{__}. A port which
1601 uses this macro should also arrange to use @file{t-gnu-prefix} in
1602 the libgcc @file{config.host}.
1609 Define these macros to be the size in bits of the mantissa of
1610 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1611 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1612 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1613 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1614 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1615 @code{DOUBLE_TYPE_SIZE} or
1616 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1619 @defmac TARGET_FLT_EVAL_METHOD
1620 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1621 assuming, if applicable, that the floating-point control word is in its
1622 default state. If you do not define this macro the value of
1623 @code{FLT_EVAL_METHOD} will be zero.
1626 @defmac WIDEST_HARDWARE_FP_SIZE
1627 A C expression for the size in bits of the widest floating-point format
1628 supported by the hardware. If you define this macro, you must specify a
1629 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1630 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1634 @defmac DEFAULT_SIGNED_CHAR
1635 An expression whose value is 1 or 0, according to whether the type
1636 @code{char} should be signed or unsigned by default. The user can
1637 always override this default with the options @option{-fsigned-char}
1638 and @option{-funsigned-char}.
1641 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1642 This target hook should return true if the compiler should give an
1643 @code{enum} type only as many bytes as it takes to represent the range
1644 of possible values of that type. It should return false if all
1645 @code{enum} types should be allocated like @code{int}.
1647 The default is to return false.
1651 A C expression for a string describing the name of the data type to use
1652 for size values. The typedef name @code{size_t} is defined using the
1653 contents of the string.
1655 The string can contain more than one keyword. If so, separate them with
1656 spaces, and write first any length keyword, then @code{unsigned} if
1657 appropriate, and finally @code{int}. The string must exactly match one
1658 of the data type names defined in the function
1659 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1660 omit @code{int} or change the order---that would cause the compiler to
1663 If you don't define this macro, the default is @code{"long unsigned
1667 @defmac PTRDIFF_TYPE
1668 A C expression for a string describing the name of the data type to use
1669 for the result of subtracting two pointers. The typedef name
1670 @code{ptrdiff_t} is defined using the contents of the string. See
1671 @code{SIZE_TYPE} above for more information.
1673 If you don't define this macro, the default is @code{"long int"}.
1677 A C expression for a string describing the name of the data type to use
1678 for wide characters. The typedef name @code{wchar_t} is defined using
1679 the contents of the string. See @code{SIZE_TYPE} above for more
1682 If you don't define this macro, the default is @code{"int"}.
1685 @defmac WCHAR_TYPE_SIZE
1686 A C expression for the size in bits of the data type for wide
1687 characters. This is used in @code{cpp}, which cannot make use of
1692 A C expression for a string describing the name of the data type to
1693 use for wide characters passed to @code{printf} and returned from
1694 @code{getwc}. The typedef name @code{wint_t} is defined using the
1695 contents of the string. See @code{SIZE_TYPE} above for more
1698 If you don't define this macro, the default is @code{"unsigned int"}.
1702 A C expression for a string describing the name of the data type that
1703 can represent any value of any standard or extended signed integer type.
1704 The typedef name @code{intmax_t} is defined using the contents of the
1705 string. See @code{SIZE_TYPE} above for more information.
1707 If you don't define this macro, the default is the first of
1708 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1709 much precision as @code{long long int}.
1712 @defmac UINTMAX_TYPE
1713 A C expression for a string describing the name of the data type that
1714 can represent any value of any standard or extended unsigned integer
1715 type. The typedef name @code{uintmax_t} is defined using the contents
1716 of the string. See @code{SIZE_TYPE} above for more information.
1718 If you don't define this macro, the default is the first of
1719 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1720 unsigned int"} that has as much precision as @code{long long unsigned
1724 @defmac SIG_ATOMIC_TYPE
1730 @defmacx UINT16_TYPE
1731 @defmacx UINT32_TYPE
1732 @defmacx UINT64_TYPE
1733 @defmacx INT_LEAST8_TYPE
1734 @defmacx INT_LEAST16_TYPE
1735 @defmacx INT_LEAST32_TYPE
1736 @defmacx INT_LEAST64_TYPE
1737 @defmacx UINT_LEAST8_TYPE
1738 @defmacx UINT_LEAST16_TYPE
1739 @defmacx UINT_LEAST32_TYPE
1740 @defmacx UINT_LEAST64_TYPE
1741 @defmacx INT_FAST8_TYPE
1742 @defmacx INT_FAST16_TYPE
1743 @defmacx INT_FAST32_TYPE
1744 @defmacx INT_FAST64_TYPE
1745 @defmacx UINT_FAST8_TYPE
1746 @defmacx UINT_FAST16_TYPE
1747 @defmacx UINT_FAST32_TYPE
1748 @defmacx UINT_FAST64_TYPE
1749 @defmacx INTPTR_TYPE
1750 @defmacx UINTPTR_TYPE
1751 C expressions for the standard types @code{sig_atomic_t},
1752 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1753 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1754 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1755 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1756 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1757 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1758 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1759 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1760 @code{SIZE_TYPE} above for more information.
1762 If any of these macros evaluates to a null pointer, the corresponding
1763 type is not supported; if GCC is configured to provide
1764 @code{<stdint.h>} in such a case, the header provided may not conform
1765 to C99, depending on the type in question. The defaults for all of
1766 these macros are null pointers.
1769 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1770 The C++ compiler represents a pointer-to-member-function with a struct
1777 ptrdiff_t vtable_index;
1784 The C++ compiler must use one bit to indicate whether the function that
1785 will be called through a pointer-to-member-function is virtual.
1786 Normally, we assume that the low-order bit of a function pointer must
1787 always be zero. Then, by ensuring that the vtable_index is odd, we can
1788 distinguish which variant of the union is in use. But, on some
1789 platforms function pointers can be odd, and so this doesn't work. In
1790 that case, we use the low-order bit of the @code{delta} field, and shift
1791 the remainder of the @code{delta} field to the left.
1793 GCC will automatically make the right selection about where to store
1794 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1795 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1796 set such that functions always start at even addresses, but the lowest
1797 bit of pointers to functions indicate whether the function at that
1798 address is in ARM or Thumb mode. If this is the case of your
1799 architecture, you should define this macro to
1800 @code{ptrmemfunc_vbit_in_delta}.
1802 In general, you should not have to define this macro. On architectures
1803 in which function addresses are always even, according to
1804 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1805 @code{ptrmemfunc_vbit_in_pfn}.
1808 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1809 Normally, the C++ compiler uses function pointers in vtables. This
1810 macro allows the target to change to use ``function descriptors''
1811 instead. Function descriptors are found on targets for whom a
1812 function pointer is actually a small data structure. Normally the
1813 data structure consists of the actual code address plus a data
1814 pointer to which the function's data is relative.
1816 If vtables are used, the value of this macro should be the number
1817 of words that the function descriptor occupies.
1820 @defmac TARGET_VTABLE_ENTRY_ALIGN
1821 By default, the vtable entries are void pointers, the so the alignment
1822 is the same as pointer alignment. The value of this macro specifies
1823 the alignment of the vtable entry in bits. It should be defined only
1824 when special alignment is necessary. */
1827 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1828 There are a few non-descriptor entries in the vtable at offsets below
1829 zero. If these entries must be padded (say, to preserve the alignment
1830 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1831 of words in each data entry.
1835 @section Register Usage
1836 @cindex register usage
1838 This section explains how to describe what registers the target machine
1839 has, and how (in general) they can be used.
1841 The description of which registers a specific instruction can use is
1842 done with register classes; see @ref{Register Classes}. For information
1843 on using registers to access a stack frame, see @ref{Frame Registers}.
1844 For passing values in registers, see @ref{Register Arguments}.
1845 For returning values in registers, see @ref{Scalar Return}.
1848 * Register Basics:: Number and kinds of registers.
1849 * Allocation Order:: Order in which registers are allocated.
1850 * Values in Registers:: What kinds of values each reg can hold.
1851 * Leaf Functions:: Renumbering registers for leaf functions.
1852 * Stack Registers:: Handling a register stack such as 80387.
1855 @node Register Basics
1856 @subsection Basic Characteristics of Registers
1858 @c prevent bad page break with this line
1859 Registers have various characteristics.
1861 @defmac FIRST_PSEUDO_REGISTER
1862 Number of hardware registers known to the compiler. They receive
1863 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1864 pseudo register's number really is assigned the number
1865 @code{FIRST_PSEUDO_REGISTER}.
1868 @defmac FIXED_REGISTERS
1869 @cindex fixed register
1870 An initializer that says which registers are used for fixed purposes
1871 all throughout the compiled code and are therefore not available for
1872 general allocation. These would include the stack pointer, the frame
1873 pointer (except on machines where that can be used as a general
1874 register when no frame pointer is needed), the program counter on
1875 machines where that is considered one of the addressable registers,
1876 and any other numbered register with a standard use.
1878 This information is expressed as a sequence of numbers, separated by
1879 commas and surrounded by braces. The @var{n}th number is 1 if
1880 register @var{n} is fixed, 0 otherwise.
1882 The table initialized from this macro, and the table initialized by
1883 the following one, may be overridden at run time either automatically,
1884 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1885 the user with the command options @option{-ffixed-@var{reg}},
1886 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1889 @defmac CALL_USED_REGISTERS
1890 @cindex call-used register
1891 @cindex call-clobbered register
1892 @cindex call-saved register
1893 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1894 clobbered (in general) by function calls as well as for fixed
1895 registers. This macro therefore identifies the registers that are not
1896 available for general allocation of values that must live across
1899 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1900 automatically saves it on function entry and restores it on function
1901 exit, if the register is used within the function.
1904 @defmac CALL_REALLY_USED_REGISTERS
1905 @cindex call-used register
1906 @cindex call-clobbered register
1907 @cindex call-saved register
1908 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1909 that the entire set of @code{FIXED_REGISTERS} be included.
1910 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1911 This macro is optional. If not specified, it defaults to the value
1912 of @code{CALL_USED_REGISTERS}.
1915 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1916 @cindex call-used register
1917 @cindex call-clobbered register
1918 @cindex call-saved register
1919 A C expression that is nonzero if it is not permissible to store a
1920 value of mode @var{mode} in hard register number @var{regno} across a
1921 call without some part of it being clobbered. For most machines this
1922 macro need not be defined. It is only required for machines that do not
1923 preserve the entire contents of a register across a call.
1927 @findex call_used_regs
1930 @findex reg_class_contents
1931 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1932 This hook may conditionally modify five variables
1933 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1934 @code{reg_names}, and @code{reg_class_contents}, to take into account
1935 any dependence of these register sets on target flags. The first three
1936 of these are of type @code{char []} (interpreted as Boolean vectors).
1937 @code{global_regs} is a @code{const char *[]}, and
1938 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1939 called, @code{fixed_regs}, @code{call_used_regs},
1940 @code{reg_class_contents}, and @code{reg_names} have been initialized
1941 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1942 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1943 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1944 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1945 command options have been applied.
1947 @cindex disabling certain registers
1948 @cindex controlling register usage
1949 If the usage of an entire class of registers depends on the target
1950 flags, you may indicate this to GCC by using this macro to modify
1951 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1952 registers in the classes which should not be used by GCC@. Also define
1953 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1954 to return @code{NO_REGS} if it
1955 is called with a letter for a class that shouldn't be used.
1957 (However, if this class is not included in @code{GENERAL_REGS} and all
1958 of the insn patterns whose constraints permit this class are
1959 controlled by target switches, then GCC will automatically avoid using
1960 these registers when the target switches are opposed to them.)
1963 @defmac INCOMING_REGNO (@var{out})
1964 Define this macro if the target machine has register windows. This C
1965 expression returns the register number as seen by the called function
1966 corresponding to the register number @var{out} as seen by the calling
1967 function. Return @var{out} if register number @var{out} is not an
1971 @defmac OUTGOING_REGNO (@var{in})
1972 Define this macro if the target machine has register windows. This C
1973 expression returns the register number as seen by the calling function
1974 corresponding to the register number @var{in} as seen by the called
1975 function. Return @var{in} if register number @var{in} is not an inbound
1979 @defmac LOCAL_REGNO (@var{regno})
1980 Define this macro if the target machine has register windows. This C
1981 expression returns true if the register is call-saved but is in the
1982 register window. Unlike most call-saved registers, such registers
1983 need not be explicitly restored on function exit or during non-local
1988 If the program counter has a register number, define this as that
1989 register number. Otherwise, do not define it.
1992 @node Allocation Order
1993 @subsection Order of Allocation of Registers
1994 @cindex order of register allocation
1995 @cindex register allocation order
1997 @c prevent bad page break with this line
1998 Registers are allocated in order.
2000 @defmac REG_ALLOC_ORDER
2001 If defined, an initializer for a vector of integers, containing the
2002 numbers of hard registers in the order in which GCC should prefer
2003 to use them (from most preferred to least).
2005 If this macro is not defined, registers are used lowest numbered first
2006 (all else being equal).
2008 One use of this macro is on machines where the highest numbered
2009 registers must always be saved and the save-multiple-registers
2010 instruction supports only sequences of consecutive registers. On such
2011 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2012 the highest numbered allocable register first.
2015 @defmac ADJUST_REG_ALLOC_ORDER
2016 A C statement (sans semicolon) to choose the order in which to allocate
2017 hard registers for pseudo-registers local to a basic block.
2019 Store the desired register order in the array @code{reg_alloc_order}.
2020 Element 0 should be the register to allocate first; element 1, the next
2021 register; and so on.
2023 The macro body should not assume anything about the contents of
2024 @code{reg_alloc_order} before execution of the macro.
2026 On most machines, it is not necessary to define this macro.
2029 @defmac HONOR_REG_ALLOC_ORDER
2030 Normally, IRA tries to estimate the costs for saving a register in the
2031 prologue and restoring it in the epilogue. This discourages it from
2032 using call-saved registers. If a machine wants to ensure that IRA
2033 allocates registers in the order given by REG_ALLOC_ORDER even if some
2034 call-saved registers appear earlier than call-used ones, this macro
2038 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2039 In some case register allocation order is not enough for the
2040 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2041 If this macro is defined, it should return a floating point value
2042 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2043 be increased by approximately the pseudo's usage frequency times the
2044 value returned by this macro. Not defining this macro is equivalent
2045 to having it always return @code{0.0}.
2047 On most machines, it is not necessary to define this macro.
2050 @node Values in Registers
2051 @subsection How Values Fit in Registers
2053 This section discusses the macros that describe which kinds of values
2054 (specifically, which machine modes) each register can hold, and how many
2055 consecutive registers are needed for a given mode.
2057 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2058 A C expression for the number of consecutive hard registers, starting
2059 at register number @var{regno}, required to hold a value of mode
2060 @var{mode}. This macro must never return zero, even if a register
2061 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2062 and/or CANNOT_CHANGE_MODE_CLASS instead.
2064 On a machine where all registers are exactly one word, a suitable
2065 definition of this macro is
2068 #define HARD_REGNO_NREGS(REGNO, MODE) \
2069 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2074 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2075 A C expression that is nonzero if a value of mode @var{mode}, stored
2076 in memory, ends with padding that causes it to take up more space than
2077 in registers starting at register number @var{regno} (as determined by
2078 multiplying GCC's notion of the size of the register when containing
2079 this mode by the number of registers returned by
2080 @code{HARD_REGNO_NREGS}). By default this is zero.
2082 For example, if a floating-point value is stored in three 32-bit
2083 registers but takes up 128 bits in memory, then this would be
2086 This macros only needs to be defined if there are cases where
2087 @code{subreg_get_info}
2088 would otherwise wrongly determine that a @code{subreg} can be
2089 represented by an offset to the register number, when in fact such a
2090 @code{subreg} would contain some of the padding not stored in
2091 registers and so not be representable.
2094 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2095 For values of @var{regno} and @var{mode} for which
2096 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2097 returning the greater number of registers required to hold the value
2098 including any padding. In the example above, the value would be four.
2101 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2102 Define this macro if the natural size of registers that hold values
2103 of mode @var{mode} is not the word size. It is a C expression that
2104 should give the natural size in bytes for the specified mode. It is
2105 used by the register allocator to try to optimize its results. This
2106 happens for example on SPARC 64-bit where the natural size of
2107 floating-point registers is still 32-bit.
2110 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2111 A C expression that is nonzero if it is permissible to store a value
2112 of mode @var{mode} in hard register number @var{regno} (or in several
2113 registers starting with that one). For a machine where all registers
2114 are equivalent, a suitable definition is
2117 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2120 You need not include code to check for the numbers of fixed registers,
2121 because the allocation mechanism considers them to be always occupied.
2123 @cindex register pairs
2124 On some machines, double-precision values must be kept in even/odd
2125 register pairs. You can implement that by defining this macro to reject
2126 odd register numbers for such modes.
2128 The minimum requirement for a mode to be OK in a register is that the
2129 @samp{mov@var{mode}} instruction pattern support moves between the
2130 register and other hard register in the same class and that moving a
2131 value into the register and back out not alter it.
2133 Since the same instruction used to move @code{word_mode} will work for
2134 all narrower integer modes, it is not necessary on any machine for
2135 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2136 you define patterns @samp{movhi}, etc., to take advantage of this. This
2137 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2138 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2141 Many machines have special registers for floating point arithmetic.
2142 Often people assume that floating point machine modes are allowed only
2143 in floating point registers. This is not true. Any registers that
2144 can hold integers can safely @emph{hold} a floating point machine
2145 mode, whether or not floating arithmetic can be done on it in those
2146 registers. Integer move instructions can be used to move the values.
2148 On some machines, though, the converse is true: fixed-point machine
2149 modes may not go in floating registers. This is true if the floating
2150 registers normalize any value stored in them, because storing a
2151 non-floating value there would garble it. In this case,
2152 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2153 floating registers. But if the floating registers do not automatically
2154 normalize, if you can store any bit pattern in one and retrieve it
2155 unchanged without a trap, then any machine mode may go in a floating
2156 register, so you can define this macro to say so.
2158 The primary significance of special floating registers is rather that
2159 they are the registers acceptable in floating point arithmetic
2160 instructions. However, this is of no concern to
2161 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2162 constraints for those instructions.
2164 On some machines, the floating registers are especially slow to access,
2165 so that it is better to store a value in a stack frame than in such a
2166 register if floating point arithmetic is not being done. As long as the
2167 floating registers are not in class @code{GENERAL_REGS}, they will not
2168 be used unless some pattern's constraint asks for one.
2171 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2172 A C expression that is nonzero if it is OK to rename a hard register
2173 @var{from} to another hard register @var{to}.
2175 One common use of this macro is to prevent renaming of a register to
2176 another register that is not saved by a prologue in an interrupt
2179 The default is always nonzero.
2182 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2183 A C expression that is nonzero if a value of mode
2184 @var{mode1} is accessible in mode @var{mode2} without copying.
2186 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2187 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2188 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2189 should be nonzero. If they differ for any @var{r}, you should define
2190 this macro to return zero unless some other mechanism ensures the
2191 accessibility of the value in a narrower mode.
2193 You should define this macro to return nonzero in as many cases as
2194 possible since doing so will allow GCC to perform better register
2198 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2199 This target hook should return @code{true} if it is OK to use a hard register
2200 @var{regno} as scratch reg in peephole2.
2202 One common use of this macro is to prevent using of a register that
2203 is not saved by a prologue in an interrupt handler.
2205 The default version of this hook always returns @code{true}.
2208 @defmac AVOID_CCMODE_COPIES
2209 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2210 registers. You should only define this macro if support for copying to/from
2211 @code{CCmode} is incomplete.
2214 @node Leaf Functions
2215 @subsection Handling Leaf Functions
2217 @cindex leaf functions
2218 @cindex functions, leaf
2219 On some machines, a leaf function (i.e., one which makes no calls) can run
2220 more efficiently if it does not make its own register window. Often this
2221 means it is required to receive its arguments in the registers where they
2222 are passed by the caller, instead of the registers where they would
2225 The special treatment for leaf functions generally applies only when
2226 other conditions are met; for example, often they may use only those
2227 registers for its own variables and temporaries. We use the term ``leaf
2228 function'' to mean a function that is suitable for this special
2229 handling, so that functions with no calls are not necessarily ``leaf
2232 GCC assigns register numbers before it knows whether the function is
2233 suitable for leaf function treatment. So it needs to renumber the
2234 registers in order to output a leaf function. The following macros
2237 @defmac LEAF_REGISTERS
2238 Name of a char vector, indexed by hard register number, which
2239 contains 1 for a register that is allowable in a candidate for leaf
2242 If leaf function treatment involves renumbering the registers, then the
2243 registers marked here should be the ones before renumbering---those that
2244 GCC would ordinarily allocate. The registers which will actually be
2245 used in the assembler code, after renumbering, should not be marked with 1
2248 Define this macro only if the target machine offers a way to optimize
2249 the treatment of leaf functions.
2252 @defmac LEAF_REG_REMAP (@var{regno})
2253 A C expression whose value is the register number to which @var{regno}
2254 should be renumbered, when a function is treated as a leaf function.
2256 If @var{regno} is a register number which should not appear in a leaf
2257 function before renumbering, then the expression should yield @minus{}1, which
2258 will cause the compiler to abort.
2260 Define this macro only if the target machine offers a way to optimize the
2261 treatment of leaf functions, and registers need to be renumbered to do
2265 @findex current_function_is_leaf
2266 @findex current_function_uses_only_leaf_regs
2267 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2268 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2269 specially. They can test the C variable @code{current_function_is_leaf}
2270 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2271 set prior to local register allocation and is valid for the remaining
2272 compiler passes. They can also test the C variable
2273 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2274 functions which only use leaf registers.
2275 @code{current_function_uses_only_leaf_regs} is valid after all passes
2276 that modify the instructions have been run and is only useful if
2277 @code{LEAF_REGISTERS} is defined.
2278 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2279 @c of the next paragraph?! --mew 2feb93
2281 @node Stack Registers
2282 @subsection Registers That Form a Stack
2284 There are special features to handle computers where some of the
2285 ``registers'' form a stack. Stack registers are normally written by
2286 pushing onto the stack, and are numbered relative to the top of the
2289 Currently, GCC can only handle one group of stack-like registers, and
2290 they must be consecutively numbered. Furthermore, the existing
2291 support for stack-like registers is specific to the 80387 floating
2292 point coprocessor. If you have a new architecture that uses
2293 stack-like registers, you will need to do substantial work on
2294 @file{reg-stack.c} and write your machine description to cooperate
2295 with it, as well as defining these macros.
2298 Define this if the machine has any stack-like registers.
2301 @defmac STACK_REG_COVER_CLASS
2302 This is a cover class containing the stack registers. Define this if
2303 the machine has any stack-like registers.
2306 @defmac FIRST_STACK_REG
2307 The number of the first stack-like register. This one is the top
2311 @defmac LAST_STACK_REG
2312 The number of the last stack-like register. This one is the bottom of
2316 @node Register Classes
2317 @section Register Classes
2318 @cindex register class definitions
2319 @cindex class definitions, register
2321 On many machines, the numbered registers are not all equivalent.
2322 For example, certain registers may not be allowed for indexed addressing;
2323 certain registers may not be allowed in some instructions. These machine
2324 restrictions are described to the compiler using @dfn{register classes}.
2326 You define a number of register classes, giving each one a name and saying
2327 which of the registers belong to it. Then you can specify register classes
2328 that are allowed as operands to particular instruction patterns.
2332 In general, each register will belong to several classes. In fact, one
2333 class must be named @code{ALL_REGS} and contain all the registers. Another
2334 class must be named @code{NO_REGS} and contain no registers. Often the
2335 union of two classes will be another class; however, this is not required.
2337 @findex GENERAL_REGS
2338 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2339 terribly special about the name, but the operand constraint letters
2340 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2341 the same as @code{ALL_REGS}, just define it as a macro which expands
2344 Order the classes so that if class @var{x} is contained in class @var{y}
2345 then @var{x} has a lower class number than @var{y}.
2347 The way classes other than @code{GENERAL_REGS} are specified in operand
2348 constraints is through machine-dependent operand constraint letters.
2349 You can define such letters to correspond to various classes, then use
2350 them in operand constraints.
2352 You must define the narrowest register classes for allocatable
2353 registers, so that each class either has no subclasses, or that for
2354 some mode, the move cost between registers within the class is
2355 cheaper than moving a register in the class to or from memory
2358 You should define a class for the union of two classes whenever some
2359 instruction allows both classes. For example, if an instruction allows
2360 either a floating point (coprocessor) register or a general register for a
2361 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2362 which includes both of them. Otherwise you will get suboptimal code,
2363 or even internal compiler errors when reload cannot find a register in the
2364 class computed via @code{reg_class_subunion}.
2366 You must also specify certain redundant information about the register
2367 classes: for each class, which classes contain it and which ones are
2368 contained in it; for each pair of classes, the largest class contained
2371 When a value occupying several consecutive registers is expected in a
2372 certain class, all the registers used must belong to that class.
2373 Therefore, register classes cannot be used to enforce a requirement for
2374 a register pair to start with an even-numbered register. The way to
2375 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2377 Register classes used for input-operands of bitwise-and or shift
2378 instructions have a special requirement: each such class must have, for
2379 each fixed-point machine mode, a subclass whose registers can transfer that
2380 mode to or from memory. For example, on some machines, the operations for
2381 single-byte values (@code{QImode}) are limited to certain registers. When
2382 this is so, each register class that is used in a bitwise-and or shift
2383 instruction must have a subclass consisting of registers from which
2384 single-byte values can be loaded or stored. This is so that
2385 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2387 @deftp {Data type} {enum reg_class}
2388 An enumerated type that must be defined with all the register class names
2389 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2390 must be the last register class, followed by one more enumerated value,
2391 @code{LIM_REG_CLASSES}, which is not a register class but rather
2392 tells how many classes there are.
2394 Each register class has a number, which is the value of casting
2395 the class name to type @code{int}. The number serves as an index
2396 in many of the tables described below.
2399 @defmac N_REG_CLASSES
2400 The number of distinct register classes, defined as follows:
2403 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2407 @defmac REG_CLASS_NAMES
2408 An initializer containing the names of the register classes as C string
2409 constants. These names are used in writing some of the debugging dumps.
2412 @defmac REG_CLASS_CONTENTS
2413 An initializer containing the contents of the register classes, as integers
2414 which are bit masks. The @var{n}th integer specifies the contents of class
2415 @var{n}. The way the integer @var{mask} is interpreted is that
2416 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2418 When the machine has more than 32 registers, an integer does not suffice.
2419 Then the integers are replaced by sub-initializers, braced groupings containing
2420 several integers. Each sub-initializer must be suitable as an initializer
2421 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2422 In this situation, the first integer in each sub-initializer corresponds to
2423 registers 0 through 31, the second integer to registers 32 through 63, and
2427 @defmac REGNO_REG_CLASS (@var{regno})
2428 A C expression whose value is a register class containing hard register
2429 @var{regno}. In general there is more than one such class; choose a class
2430 which is @dfn{minimal}, meaning that no smaller class also contains the
2434 @defmac BASE_REG_CLASS
2435 A macro whose definition is the name of the class to which a valid
2436 base register must belong. A base register is one used in an address
2437 which is the register value plus a displacement.
2440 @defmac MODE_BASE_REG_CLASS (@var{mode})
2441 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2442 the selection of a base register in a mode dependent manner. If
2443 @var{mode} is VOIDmode then it should return the same value as
2444 @code{BASE_REG_CLASS}.
2447 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2448 A C expression whose value is the register class to which a valid
2449 base register must belong in order to be used in a base plus index
2450 register address. You should define this macro if base plus index
2451 addresses have different requirements than other base register uses.
2454 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2455 A C expression whose value is the register class to which a valid
2456 base register for a memory reference in mode @var{mode} to address
2457 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2458 define the context in which the base register occurs. @var{outer_code} is
2459 the code of the immediately enclosing expression (@code{MEM} for the top level
2460 of an address, @code{ADDRESS} for something that occurs in an
2461 @code{address_operand}). @var{index_code} is the code of the corresponding
2462 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2465 @defmac INDEX_REG_CLASS
2466 A macro whose definition is the name of the class to which a valid
2467 index register must belong. An index register is one used in an
2468 address where its value is either multiplied by a scale factor or
2469 added to another register (as well as added to a displacement).
2472 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2473 A C expression which is nonzero if register number @var{num} is
2474 suitable for use as a base register in operand addresses.
2477 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2478 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2479 that expression may examine the mode of the memory reference in
2480 @var{mode}. You should define this macro if the mode of the memory
2481 reference affects whether a register may be used as a base register. If
2482 you define this macro, the compiler will use it instead of
2483 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2484 addresses that appear outside a @code{MEM}, i.e., as an
2485 @code{address_operand}.
2488 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2489 A C expression which is nonzero if register number @var{num} is suitable for
2490 use as a base register in base plus index operand addresses, accessing
2491 memory in mode @var{mode}. It may be either a suitable hard register or a
2492 pseudo register that has been allocated such a hard register. You should
2493 define this macro if base plus index addresses have different requirements
2494 than other base register uses.
2496 Use of this macro is deprecated; please use the more general
2497 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2500 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2501 A C expression which is nonzero if register number @var{num} is
2502 suitable for use as a base register in operand addresses, accessing
2503 memory in mode @var{mode} in address space @var{address_space}.
2504 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2505 that that expression may examine the context in which the register
2506 appears in the memory reference. @var{outer_code} is the code of the
2507 immediately enclosing expression (@code{MEM} if at the top level of the
2508 address, @code{ADDRESS} for something that occurs in an
2509 @code{address_operand}). @var{index_code} is the code of the
2510 corresponding index expression if @var{outer_code} is @code{PLUS};
2511 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2512 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2515 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2516 A C expression which is nonzero if register number @var{num} is
2517 suitable for use as an index register in operand addresses. It may be
2518 either a suitable hard register or a pseudo register that has been
2519 allocated such a hard register.
2521 The difference between an index register and a base register is that
2522 the index register may be scaled. If an address involves the sum of
2523 two registers, neither one of them scaled, then either one may be
2524 labeled the ``base'' and the other the ``index''; but whichever
2525 labeling is used must fit the machine's constraints of which registers
2526 may serve in each capacity. The compiler will try both labelings,
2527 looking for one that is valid, and will reload one or both registers
2528 only if neither labeling works.
2531 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2532 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.
2535 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2536 A target hook that places additional restrictions on the register class
2537 to use when it is necessary to copy value @var{x} into a register in class
2538 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2539 another, smaller class.
2541 The default version of this hook always returns value of @code{rclass} argument.
2543 Sometimes returning a more restrictive class makes better code. For
2544 example, on the 68000, when @var{x} is an integer constant that is in range
2545 for a @samp{moveq} instruction, the value of this macro is always
2546 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2547 Requiring a data register guarantees that a @samp{moveq} will be used.
2549 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2550 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2551 loaded into some register class. By returning @code{NO_REGS} you can
2552 force @var{x} into a memory location. For example, rs6000 can load
2553 immediate values into general-purpose registers, but does not have an
2554 instruction for loading an immediate value into a floating-point
2555 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2556 @var{x} is a floating-point constant. If the constant can't be loaded
2557 into any kind of register, code generation will be better if
2558 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2559 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2561 If an insn has pseudos in it after register allocation, reload will go
2562 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2563 to find the best one. Returning @code{NO_REGS}, in this case, makes
2564 reload add a @code{!} in front of the constraint: the x86 back-end uses
2565 this feature to discourage usage of 387 registers when math is done in
2566 the SSE registers (and vice versa).
2569 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2570 A C expression that places additional restrictions on the register class
2571 to use when it is necessary to copy value @var{x} into a register in class
2572 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2573 another, smaller class. On many machines, the following definition is
2577 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2580 Sometimes returning a more restrictive class makes better code. For
2581 example, on the 68000, when @var{x} is an integer constant that is in range
2582 for a @samp{moveq} instruction, the value of this macro is always
2583 @code{DATA_REGS} as long as @var{class} includes the data registers.
2584 Requiring a data register guarantees that a @samp{moveq} will be used.
2586 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2587 @var{class} is if @var{x} is a legitimate constant which cannot be
2588 loaded into some register class. By returning @code{NO_REGS} you can
2589 force @var{x} into a memory location. For example, rs6000 can load
2590 immediate values into general-purpose registers, but does not have an
2591 instruction for loading an immediate value into a floating-point
2592 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2593 @var{x} is a floating-point constant. If the constant can't be loaded
2594 into any kind of register, code generation will be better if
2595 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2596 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2598 If an insn has pseudos in it after register allocation, reload will go
2599 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2600 to find the best one. Returning @code{NO_REGS}, in this case, makes
2601 reload add a @code{!} in front of the constraint: the x86 back-end uses
2602 this feature to discourage usage of 387 registers when math is done in
2603 the SSE registers (and vice versa).
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 @deftypefn {Target Hook} {unsigned char} TARGET_CLASS_MAX_NREGS (reg_class_t @var{rclass}, enum machine_mode @var{mode})
2843 A target hook returns the maximum number of consecutive registers
2844 of class @var{rclass} 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 returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass},
2848 @var{mode})} target hook should be the maximum value of
2849 @code{HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno}
2850 values in the class @var{rclass}.
2852 This target hook helps control the handling of multiple-word values
2855 The default version of this target hook returns the size of @var{mode}
2859 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2860 A C expression for the maximum number of consecutive registers
2861 of class @var{class} needed to hold a value of mode @var{mode}.
2863 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2864 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2865 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2866 @var{mode})} for all @var{regno} values in the class @var{class}.
2868 This macro helps control the handling of multiple-word values
2872 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2873 If defined, a C expression that returns nonzero for a @var{class} for which
2874 a change from mode @var{from} to mode @var{to} is invalid.
2876 For the example, loading 32-bit integer or floating-point objects into
2877 floating-point registers on the Alpha extends them to 64 bits.
2878 Therefore loading a 64-bit object and then storing it as a 32-bit object
2879 does not store the low-order 32 bits, as would be the case for a normal
2880 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2884 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2885 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2886 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2890 @node Old Constraints
2891 @section Obsolete Macros for Defining Constraints
2892 @cindex defining constraints, obsolete method
2893 @cindex constraints, defining, obsolete method
2895 Machine-specific constraints can be defined with these macros instead
2896 of the machine description constructs described in @ref{Define
2897 Constraints}. This mechanism is obsolete. New ports should not use
2898 it; old ports should convert to the new mechanism.
2900 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2901 For the constraint at the start of @var{str}, which starts with the letter
2902 @var{c}, return the length. This allows you to have register class /
2903 constant / extra constraints that are longer than a single letter;
2904 you don't need to define this macro if you can do with single-letter
2905 constraints only. The definition of this macro should use
2906 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2907 to handle specially.
2908 There are some sanity checks in genoutput.c that check the constraint lengths
2909 for the md file, so you can also use this macro to help you while you are
2910 transitioning from a byzantine single-letter-constraint scheme: when you
2911 return a negative length for a constraint you want to re-use, genoutput
2912 will complain about every instance where it is used in the md file.
2915 @defmac REG_CLASS_FROM_LETTER (@var{char})
2916 A C expression which defines the machine-dependent operand constraint
2917 letters for register classes. If @var{char} is such a letter, the
2918 value should be the register class corresponding to it. Otherwise,
2919 the value should be @code{NO_REGS}. The register letter @samp{r},
2920 corresponding to class @code{GENERAL_REGS}, will not be passed
2921 to this macro; you do not need to handle it.
2924 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2925 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2926 passed in @var{str}, so that you can use suffixes to distinguish between
2930 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2931 A C expression that defines the machine-dependent operand constraint
2932 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2933 particular ranges of integer values. If @var{c} is one of those
2934 letters, the expression should check that @var{value}, an integer, is in
2935 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2936 not one of those letters, the value should be 0 regardless of
2940 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2941 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2942 string passed in @var{str}, so that you can use suffixes to distinguish
2943 between different variants.
2946 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2947 A C expression that defines the machine-dependent operand constraint
2948 letters that specify particular ranges of @code{const_double} values
2949 (@samp{G} or @samp{H}).
2951 If @var{c} is one of those letters, the expression should check that
2952 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2953 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2954 letters, the value should be 0 regardless of @var{value}.
2956 @code{const_double} is used for all floating-point constants and for
2957 @code{DImode} fixed-point constants. A given letter can accept either
2958 or both kinds of values. It can use @code{GET_MODE} to distinguish
2959 between these kinds.
2962 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2963 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2964 string passed in @var{str}, so that you can use suffixes to distinguish
2965 between different variants.
2968 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2969 A C expression that defines the optional machine-dependent constraint
2970 letters that can be used to segregate specific types of operands, usually
2971 memory references, for the target machine. Any letter that is not
2972 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2973 @code{REG_CLASS_FROM_CONSTRAINT}
2974 may be used. Normally this macro will not be defined.
2976 If it is required for a particular target machine, it should return 1
2977 if @var{value} corresponds to the operand type represented by the
2978 constraint letter @var{c}. If @var{c} is not defined as an extra
2979 constraint, the value returned should be 0 regardless of @var{value}.
2981 For example, on the ROMP, load instructions cannot have their output
2982 in r0 if the memory reference contains a symbolic address. Constraint
2983 letter @samp{Q} is defined as representing a memory address that does
2984 @emph{not} contain a symbolic address. An alternative is specified with
2985 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2986 alternative specifies @samp{m} on the input and a register class that
2987 does not include r0 on the output.
2990 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2991 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2992 in @var{str}, so that you can use suffixes to distinguish between different
2996 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2997 A C expression that defines the optional machine-dependent constraint
2998 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2999 be treated like memory constraints by the reload pass.
3001 It should return 1 if the operand type represented by the constraint
3002 at the start of @var{str}, the first letter of which is the letter @var{c},
3003 comprises a subset of all memory references including
3004 all those whose address is simply a base register. This allows the reload
3005 pass to reload an operand, if it does not directly correspond to the operand
3006 type of @var{c}, by copying its address into a base register.
3008 For example, on the S/390, some instructions do not accept arbitrary
3009 memory references, but only those that do not make use of an index
3010 register. The constraint letter @samp{Q} is defined via
3011 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
3012 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
3013 a @samp{Q} constraint can handle any memory operand, because the
3014 reload pass knows it can be reloaded by copying the memory address
3015 into a base register if required. This is analogous to the way
3016 an @samp{o} constraint can handle any memory operand.
3019 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
3020 A C expression that defines the optional machine-dependent constraint
3021 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
3022 @code{EXTRA_CONSTRAINT_STR}, that should
3023 be treated like address constraints by the reload pass.
3025 It should return 1 if the operand type represented by the constraint
3026 at the start of @var{str}, which starts with the letter @var{c}, comprises
3027 a subset of all memory addresses including
3028 all those that consist of just a base register. This allows the reload
3029 pass to reload an operand, if it does not directly correspond to the operand
3030 type of @var{str}, by copying it into a base register.
3032 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3033 be used with the @code{address_operand} predicate. It is treated
3034 analogously to the @samp{p} constraint.
3037 @node Stack and Calling
3038 @section Stack Layout and Calling Conventions
3039 @cindex calling conventions
3041 @c prevent bad page break with this line
3042 This describes the stack layout and calling conventions.
3046 * Exception Handling::
3051 * Register Arguments::
3053 * Aggregate Return::
3058 * Stack Smashing Protection::
3062 @subsection Basic Stack Layout
3063 @cindex stack frame layout
3064 @cindex frame layout
3066 @c prevent bad page break with this line
3067 Here is the basic stack layout.
3069 @defmac STACK_GROWS_DOWNWARD
3070 Define this macro if pushing a word onto the stack moves the stack
3071 pointer to a smaller address.
3073 When we say, ``define this macro if @dots{}'', it means that the
3074 compiler checks this macro only with @code{#ifdef} so the precise
3075 definition used does not matter.
3078 @defmac STACK_PUSH_CODE
3079 This macro defines the operation used when something is pushed
3080 on the stack. In RTL, a push operation will be
3081 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3083 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3084 and @code{POST_INC}. Which of these is correct depends on
3085 the stack direction and on whether the stack pointer points
3086 to the last item on the stack or whether it points to the
3087 space for the next item on the stack.
3089 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3090 defined, which is almost always right, and @code{PRE_INC} otherwise,
3091 which is often wrong.
3094 @defmac FRAME_GROWS_DOWNWARD
3095 Define this macro to nonzero value if the addresses of local variable slots
3096 are at negative offsets from the frame pointer.
3099 @defmac ARGS_GROW_DOWNWARD
3100 Define this macro if successive arguments to a function occupy decreasing
3101 addresses on the stack.
3104 @defmac STARTING_FRAME_OFFSET
3105 Offset from the frame pointer to the first local variable slot to be allocated.
3107 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3108 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3109 Otherwise, it is found by adding the length of the first slot to the
3110 value @code{STARTING_FRAME_OFFSET}.
3111 @c i'm not sure if the above is still correct.. had to change it to get
3112 @c rid of an overfull. --mew 2feb93
3115 @defmac STACK_ALIGNMENT_NEEDED
3116 Define to zero to disable final alignment of the stack during reload.
3117 The nonzero default for this macro is suitable for most ports.
3119 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3120 is a register save block following the local block that doesn't require
3121 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3122 stack alignment and do it in the backend.
3125 @defmac STACK_POINTER_OFFSET
3126 Offset from the stack pointer register to the first location at which
3127 outgoing arguments are placed. If not specified, the default value of
3128 zero is used. This is the proper value for most machines.
3130 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3131 the first location at which outgoing arguments are placed.
3134 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3135 Offset from the argument pointer register to the first argument's
3136 address. On some machines it may depend on the data type of the
3139 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3140 the first argument's address.
3143 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3144 Offset from the stack pointer register to an item dynamically allocated
3145 on the stack, e.g., by @code{alloca}.
3147 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3148 length of the outgoing arguments. The default is correct for most
3149 machines. See @file{function.c} for details.
3152 @defmac INITIAL_FRAME_ADDRESS_RTX
3153 A C expression whose value is RTL representing the address of the initial
3154 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3155 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3156 default value will be used. Define this macro in order to make frame pointer
3157 elimination work in the presence of @code{__builtin_frame_address (count)} and
3158 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3161 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3162 A C expression whose value is RTL representing the address in a stack
3163 frame where the pointer to the caller's frame is stored. Assume that
3164 @var{frameaddr} is an RTL expression for the address of the stack frame
3167 If you don't define this macro, the default is to return the value
3168 of @var{frameaddr}---that is, the stack frame address is also the
3169 address of the stack word that points to the previous frame.
3172 @defmac SETUP_FRAME_ADDRESSES
3173 If defined, a C expression that produces the machine-specific code to
3174 setup the stack so that arbitrary frames can be accessed. For example,
3175 on the SPARC, we must flush all of the register windows to the stack
3176 before we can access arbitrary stack frames. You will seldom need to
3180 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3181 This target hook should return an rtx that is used to store
3182 the address of the current frame into the built in @code{setjmp} buffer.
3183 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3184 machines. One reason you may need to define this target hook is if
3185 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3188 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3189 A C expression whose value is RTL representing the value of the frame
3190 address for the current frame. @var{frameaddr} is the frame pointer
3191 of the current frame. This is used for __builtin_frame_address.
3192 You need only define this macro if the frame address is not the same
3193 as the frame pointer. Most machines do not need to define it.
3196 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3197 A C expression whose value is RTL representing the value of the return
3198 address for the frame @var{count} steps up from the current frame, after
3199 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3200 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3201 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3203 The value of the expression must always be the correct address when
3204 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3205 determine the return address of other frames.
3208 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3209 Define this if the return address of a particular stack frame is accessed
3210 from the frame pointer of the previous stack frame.
3213 @defmac INCOMING_RETURN_ADDR_RTX
3214 A C expression whose value is RTL representing the location of the
3215 incoming return address at the beginning of any function, before the
3216 prologue. This RTL is either a @code{REG}, indicating that the return
3217 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3220 You only need to define this macro if you want to support call frame
3221 debugging information like that provided by DWARF 2.
3223 If this RTL is a @code{REG}, you should also define
3224 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3227 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3228 A C expression whose value is an integer giving a DWARF 2 column
3229 number that may be used as an alternative return column. The column
3230 must not correspond to any gcc hard register (that is, it must not
3231 be in the range of @code{DWARF_FRAME_REGNUM}).
3233 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3234 general register, but an alternative column needs to be used for signal
3235 frames. Some targets have also used different frame return columns
3239 @defmac DWARF_ZERO_REG
3240 A C expression whose value is an integer giving a DWARF 2 register
3241 number that is considered to always have the value zero. This should
3242 only be defined if the target has an architected zero register, and
3243 someone decided it was a good idea to use that register number to
3244 terminate the stack backtrace. New ports should avoid this.
3247 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3248 This target hook allows the backend to emit frame-related insns that
3249 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3250 info engine will invoke it on insns of the form
3252 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3256 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3258 to let the backend emit the call frame instructions. @var{label} is
3259 the CFI label attached to the insn, @var{pattern} is the pattern of
3260 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3263 @defmac INCOMING_FRAME_SP_OFFSET
3264 A C expression whose value is an integer giving the offset, in bytes,
3265 from the value of the stack pointer register to the top of the stack
3266 frame at the beginning of any function, before the prologue. The top of
3267 the frame is defined to be the value of the stack pointer in the
3268 previous frame, just before the call instruction.
3270 You only need to define this macro if you want to support call frame
3271 debugging information like that provided by DWARF 2.
3274 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3275 A C expression whose value is an integer giving the offset, in bytes,
3276 from the argument pointer to the canonical frame address (cfa). The
3277 final value should coincide with that calculated by
3278 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3279 during virtual register instantiation.
3281 The default value for this macro is
3282 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3283 which is correct for most machines; in general, the arguments are found
3284 immediately before the stack frame. Note that this is not the case on
3285 some targets that save registers into the caller's frame, such as SPARC
3286 and rs6000, and so such targets need to define this macro.
3288 You only need to define this macro if the default is incorrect, and you
3289 want to support call frame debugging information like that provided by
3293 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3294 If defined, a C expression whose value is an integer giving the offset
3295 in bytes from the frame pointer to the canonical frame address (cfa).
3296 The final value should coincide with that calculated by
3297 @code{INCOMING_FRAME_SP_OFFSET}.
3299 Normally the CFA is calculated as an offset from the argument pointer,
3300 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3301 variable due to the ABI, this may not be possible. If this macro is
3302 defined, it implies that the virtual register instantiation should be
3303 based on the frame pointer instead of the argument pointer. Only one
3304 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3308 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3309 If defined, a C expression whose value is an integer giving the offset
3310 in bytes from the canonical frame address (cfa) to the frame base used
3311 in DWARF 2 debug information. The default is zero. A different value
3312 may reduce the size of debug information on some ports.
3315 @node Exception Handling
3316 @subsection Exception Handling Support
3317 @cindex exception handling
3319 @defmac EH_RETURN_DATA_REGNO (@var{N})
3320 A C expression whose value is the @var{N}th register number used for
3321 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3322 @var{N} registers are usable.
3324 The exception handling library routines communicate with the exception
3325 handlers via a set of agreed upon registers. Ideally these registers
3326 should be call-clobbered; it is possible to use call-saved registers,
3327 but may negatively impact code size. The target must support at least
3328 2 data registers, but should define 4 if there are enough free registers.
3330 You must define this macro if you want to support call frame exception
3331 handling like that provided by DWARF 2.
3334 @defmac EH_RETURN_STACKADJ_RTX
3335 A C expression whose value is RTL representing a location in which
3336 to store a stack adjustment to be applied before function return.
3337 This is used to unwind the stack to an exception handler's call frame.
3338 It will be assigned zero on code paths that return normally.
3340 Typically this is a call-clobbered hard register that is otherwise
3341 untouched by the epilogue, but could also be a stack slot.
3343 Do not define this macro if the stack pointer is saved and restored
3344 by the regular prolog and epilog code in the call frame itself; in
3345 this case, the exception handling library routines will update the
3346 stack location to be restored in place. Otherwise, you must define
3347 this macro if you want to support call frame exception handling like
3348 that provided by DWARF 2.
3351 @defmac EH_RETURN_HANDLER_RTX
3352 A C expression whose value is RTL representing a location in which
3353 to store the address of an exception handler to which we should
3354 return. It will not be assigned on code paths that return normally.
3356 Typically this is the location in the call frame at which the normal
3357 return address is stored. For targets that return by popping an
3358 address off the stack, this might be a memory address just below
3359 the @emph{target} call frame rather than inside the current call
3360 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3361 been assigned, so it may be used to calculate the location of the
3364 Some targets have more complex requirements than storing to an
3365 address calculable during initial code generation. In that case
3366 the @code{eh_return} instruction pattern should be used instead.
3368 If you want to support call frame exception handling, you must
3369 define either this macro or the @code{eh_return} instruction pattern.
3372 @defmac RETURN_ADDR_OFFSET
3373 If defined, an integer-valued C expression for which rtl will be generated
3374 to add it to the exception handler address before it is searched in the
3375 exception handling tables, and to subtract it again from the address before
3376 using it to return to the exception handler.
3379 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3380 This macro chooses the encoding of pointers embedded in the exception
3381 handling sections. If at all possible, this should be defined such
3382 that the exception handling section will not require dynamic relocations,
3383 and so may be read-only.
3385 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3386 @var{global} is true if the symbol may be affected by dynamic relocations.
3387 The macro should return a combination of the @code{DW_EH_PE_*} defines
3388 as found in @file{dwarf2.h}.
3390 If this macro is not defined, pointers will not be encoded but
3391 represented directly.
3394 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3395 This macro allows the target to emit whatever special magic is required
3396 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3397 Generic code takes care of pc-relative and indirect encodings; this must
3398 be defined if the target uses text-relative or data-relative encodings.
3400 This is a C statement that branches to @var{done} if the format was
3401 handled. @var{encoding} is the format chosen, @var{size} is the number
3402 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3406 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3407 This macro allows the target to add CPU and operating system specific
3408 code to the call-frame unwinder for use when there is no unwind data
3409 available. The most common reason to implement this macro is to unwind
3410 through signal frames.
3412 This macro is called from @code{uw_frame_state_for} in
3413 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3414 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3415 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3416 for the address of the code being executed and @code{context->cfa} for
3417 the stack pointer value. If the frame can be decoded, the register
3418 save addresses should be updated in @var{fs} and the macro should
3419 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3420 the macro should evaluate to @code{_URC_END_OF_STACK}.
3422 For proper signal handling in Java this macro is accompanied by
3423 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3426 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3427 This macro allows the target to add operating system specific code to the
3428 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3429 usually used for signal or interrupt frames.
3431 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3432 @var{context} is an @code{_Unwind_Context};
3433 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3434 for the abi and context in the @code{.unwabi} directive. If the
3435 @code{.unwabi} directive can be handled, the register save addresses should
3436 be updated in @var{fs}.
3439 @defmac TARGET_USES_WEAK_UNWIND_INFO
3440 A C expression that evaluates to true if the target requires unwind
3441 info to be given comdat linkage. Define it to be @code{1} if comdat
3442 linkage is necessary. The default is @code{0}.
3445 @node Stack Checking
3446 @subsection Specifying How Stack Checking is Done
3448 GCC will check that stack references are within the boundaries of the
3449 stack, if the option @option{-fstack-check} is specified, in one of
3454 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3455 will assume that you have arranged for full stack checking to be done
3456 at appropriate places in the configuration files. GCC will not do
3457 other special processing.
3460 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3461 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3462 that you have arranged for static stack checking (checking of the
3463 static stack frame of functions) to be done at appropriate places
3464 in the configuration files. GCC will only emit code to do dynamic
3465 stack checking (checking on dynamic stack allocations) using the third
3469 If neither of the above are true, GCC will generate code to periodically
3470 ``probe'' the stack pointer using the values of the macros defined below.
3473 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3474 GCC will change its allocation strategy for large objects if the option
3475 @option{-fstack-check} is specified: they will always be allocated
3476 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3478 @defmac STACK_CHECK_BUILTIN
3479 A nonzero value if stack checking is done by the configuration files in a
3480 machine-dependent manner. You should define this macro if stack checking
3481 is required by the ABI of your machine or if you would like to do stack
3482 checking in some more efficient way than the generic approach. The default
3483 value of this macro is zero.
3486 @defmac STACK_CHECK_STATIC_BUILTIN
3487 A nonzero value if static stack checking is done by the configuration files
3488 in a machine-dependent manner. You should define this macro if you would
3489 like to do static stack checking in some more efficient way than the generic
3490 approach. The default value of this macro is zero.
3493 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3494 An integer specifying the interval at which GCC must generate stack probe
3495 instructions, defined as 2 raised to this integer. You will normally
3496 define this macro so that the interval be no larger than the size of
3497 the ``guard pages'' at the end of a stack area. The default value
3498 of 12 (4096-byte interval) is suitable for most systems.
3501 @defmac STACK_CHECK_MOVING_SP
3502 An integer which is nonzero if GCC should move the stack pointer page by page
3503 when doing probes. This can be necessary on systems where the stack pointer
3504 contains the bottom address of the memory area accessible to the executing
3505 thread at any point in time. In this situation an alternate signal stack
3506 is required in order to be able to recover from a stack overflow. The
3507 default value of this macro is zero.
3510 @defmac STACK_CHECK_PROTECT
3511 The number of bytes of stack needed to recover from a stack overflow, for
3512 languages where such a recovery is supported. The default value of 75 words
3513 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3514 8192 bytes with other exception handling mechanisms should be adequate for
3518 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3519 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3520 in the opposite case.
3522 @defmac STACK_CHECK_MAX_FRAME_SIZE
3523 The maximum size of a stack frame, in bytes. GCC will generate probe
3524 instructions in non-leaf functions to ensure at least this many bytes of
3525 stack are available. If a stack frame is larger than this size, stack
3526 checking will not be reliable and GCC will issue a warning. The
3527 default is chosen so that GCC only generates one instruction on most
3528 systems. You should normally not change the default value of this macro.
3531 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3532 GCC uses this value to generate the above warning message. It
3533 represents the amount of fixed frame used by a function, not including
3534 space for any callee-saved registers, temporaries and user variables.
3535 You need only specify an upper bound for this amount and will normally
3536 use the default of four words.
3539 @defmac STACK_CHECK_MAX_VAR_SIZE
3540 The maximum size, in bytes, of an object that GCC will place in the
3541 fixed area of the stack frame when the user specifies
3542 @option{-fstack-check}.
3543 GCC computed the default from the values of the above macros and you will
3544 normally not need to override that default.
3548 @node Frame Registers
3549 @subsection Registers That Address the Stack Frame
3551 @c prevent bad page break with this line
3552 This discusses registers that address the stack frame.
3554 @defmac STACK_POINTER_REGNUM
3555 The register number of the stack pointer register, which must also be a
3556 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3557 the hardware determines which register this is.
3560 @defmac FRAME_POINTER_REGNUM
3561 The register number of the frame pointer register, which is used to
3562 access automatic variables in the stack frame. On some machines, the
3563 hardware determines which register this is. On other machines, you can
3564 choose any register you wish for this purpose.
3567 @defmac HARD_FRAME_POINTER_REGNUM
3568 On some machines the offset between the frame pointer and starting
3569 offset of the automatic variables is not known until after register
3570 allocation has been done (for example, because the saved registers are
3571 between these two locations). On those machines, define
3572 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3573 be used internally until the offset is known, and define
3574 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3575 used for the frame pointer.
3577 You should define this macro only in the very rare circumstances when it
3578 is not possible to calculate the offset between the frame pointer and
3579 the automatic variables until after register allocation has been
3580 completed. When this macro is defined, you must also indicate in your
3581 definition of @code{ELIMINABLE_REGS} how to eliminate
3582 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3583 or @code{STACK_POINTER_REGNUM}.
3585 Do not define this macro if it would be the same as
3586 @code{FRAME_POINTER_REGNUM}.
3589 @defmac ARG_POINTER_REGNUM
3590 The register number of the arg pointer register, which is used to access
3591 the function's argument list. On some machines, this is the same as the
3592 frame pointer register. On some machines, the hardware determines which
3593 register this is. On other machines, you can choose any register you
3594 wish for this purpose. If this is not the same register as the frame
3595 pointer register, then you must mark it as a fixed register according to
3596 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3597 (@pxref{Elimination}).
3600 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3601 Define this to a preprocessor constant that is nonzero if
3602 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3603 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3604 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3605 definition is not suitable for use in preprocessor conditionals.
3608 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3609 Define this to a preprocessor constant that is nonzero if
3610 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3611 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3612 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3613 definition is not suitable for use in preprocessor conditionals.
3616 @defmac RETURN_ADDRESS_POINTER_REGNUM
3617 The register number of the return address pointer register, which is used to
3618 access the current function's return address from the stack. On some
3619 machines, the return address is not at a fixed offset from the frame
3620 pointer or stack pointer or argument pointer. This register can be defined
3621 to point to the return address on the stack, and then be converted by
3622 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3624 Do not define this macro unless there is no other way to get the return
3625 address from the stack.
3628 @defmac STATIC_CHAIN_REGNUM
3629 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3630 Register numbers used for passing a function's static chain pointer. If
3631 register windows are used, the register number as seen by the called
3632 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3633 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3634 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3637 The static chain register need not be a fixed register.
3639 If the static chain is passed in memory, these macros should not be
3640 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3643 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl}, bool @var{incoming_p})
3644 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3645 targets that may use different static chain locations for different
3646 nested functions. This may be required if the target has function
3647 attributes that affect the calling conventions of the function and
3648 those calling conventions use different static chain locations.
3650 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3652 If the static chain is passed in memory, this hook should be used to
3653 provide rtx giving @code{mem} expressions that denote where they are stored.
3654 Often the @code{mem} expression as seen by the caller will be at an offset
3655 from the stack pointer and the @code{mem} expression as seen by the callee
3656 will be at an offset from the frame pointer.
3657 @findex stack_pointer_rtx
3658 @findex frame_pointer_rtx
3659 @findex arg_pointer_rtx
3660 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3661 @code{arg_pointer_rtx} will have been initialized and should be used
3662 to refer to those items.
3665 @defmac DWARF_FRAME_REGISTERS
3666 This macro specifies the maximum number of hard registers that can be
3667 saved in a call frame. This is used to size data structures used in
3668 DWARF2 exception handling.
3670 Prior to GCC 3.0, this macro was needed in order to establish a stable
3671 exception handling ABI in the face of adding new hard registers for ISA
3672 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3673 in the number of hard registers. Nevertheless, this macro can still be
3674 used to reduce the runtime memory requirements of the exception handling
3675 routines, which can be substantial if the ISA contains a lot of
3676 registers that are not call-saved.
3678 If this macro is not defined, it defaults to
3679 @code{FIRST_PSEUDO_REGISTER}.
3682 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3684 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3685 for backward compatibility in pre GCC 3.0 compiled code.
3687 If this macro is not defined, it defaults to
3688 @code{DWARF_FRAME_REGISTERS}.
3691 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3693 Define this macro if the target's representation for dwarf registers
3694 is different than the internal representation for unwind column.
3695 Given a dwarf register, this macro should return the internal unwind
3696 column number to use instead.
3698 See the PowerPC's SPE target for an example.
3701 @defmac DWARF_FRAME_REGNUM (@var{regno})
3703 Define this macro if the target's representation for dwarf registers
3704 used in .eh_frame or .debug_frame is different from that used in other
3705 debug info sections. Given a GCC hard register number, this macro
3706 should return the .eh_frame register number. The default is
3707 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3711 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3713 Define this macro to map register numbers held in the call frame info
3714 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3715 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3716 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3717 return @code{@var{regno}}.
3721 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3723 Define this macro if the target stores register values as
3724 @code{_Unwind_Word} type in unwind context. It should be defined if
3725 target register size is larger than the size of @code{void *}. The
3726 default is to store register values as @code{void *} type.
3730 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3732 Define this macro to be 1 if the target always uses extended unwind
3733 context with version, args_size and by_value fields. If it is undefined,
3734 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3735 defined and 0 otherwise.
3740 @subsection Eliminating Frame Pointer and Arg Pointer
3742 @c prevent bad page break with this line
3743 This is about eliminating the frame pointer and arg pointer.
3745 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3746 This target hook should return @code{true} if a function must have and use
3747 a frame pointer. This target hook is called in the reload pass. If its return
3748 value is @code{true} the function will have a frame pointer.
3750 This target hook can in principle examine the current function and decide
3751 according to the facts, but on most machines the constant @code{false} or the
3752 constant @code{true} suffices. Use @code{false} when the machine allows code
3753 to be generated with no frame pointer, and doing so saves some time or space.
3754 Use @code{true} when there is no possible advantage to avoiding a frame
3757 In certain cases, the compiler does not know how to produce valid code
3758 without a frame pointer. The compiler recognizes those cases and
3759 automatically gives the function a frame pointer regardless of what
3760 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3763 In a function that does not require a frame pointer, the frame pointer
3764 register can be allocated for ordinary usage, unless you mark it as a
3765 fixed register. See @code{FIXED_REGISTERS} for more information.
3767 Default return value is @code{false}.
3770 @findex get_frame_size
3771 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3772 A C statement to store in the variable @var{depth-var} the difference
3773 between the frame pointer and the stack pointer values immediately after
3774 the function prologue. The value would be computed from information
3775 such as the result of @code{get_frame_size ()} and the tables of
3776 registers @code{regs_ever_live} and @code{call_used_regs}.
3778 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3779 need not be defined. Otherwise, it must be defined even if
3780 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3781 case, you may set @var{depth-var} to anything.
3784 @defmac ELIMINABLE_REGS
3785 If defined, this macro specifies a table of register pairs used to
3786 eliminate unneeded registers that point into the stack frame. If it is not
3787 defined, the only elimination attempted by the compiler is to replace
3788 references to the frame pointer with references to the stack pointer.
3790 The definition of this macro is a list of structure initializations, each
3791 of which specifies an original and replacement register.
3793 On some machines, the position of the argument pointer is not known until
3794 the compilation is completed. In such a case, a separate hard register
3795 must be used for the argument pointer. This register can be eliminated by
3796 replacing it with either the frame pointer or the argument pointer,
3797 depending on whether or not the frame pointer has been eliminated.
3799 In this case, you might specify:
3801 #define ELIMINABLE_REGS \
3802 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3803 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3804 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3807 Note that the elimination of the argument pointer with the stack pointer is
3808 specified first since that is the preferred elimination.
3811 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3812 This target hook should returns @code{true} if the compiler is allowed to
3813 try to replace register number @var{from_reg} with register number
3814 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3815 is defined, and will usually be @code{true}, since most of the cases
3816 preventing register elimination are things that the compiler already
3819 Default return value is @code{true}.
3822 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3823 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3824 specifies the initial difference between the specified pair of
3825 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3829 @node Stack Arguments
3830 @subsection Passing Function Arguments on the Stack
3831 @cindex arguments on stack
3832 @cindex stack arguments
3834 The macros in this section control how arguments are passed
3835 on the stack. See the following section for other macros that
3836 control passing certain arguments in registers.
3838 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3839 This target hook returns @code{true} if an argument declared in a
3840 prototype as an integral type smaller than @code{int} should actually be
3841 passed as an @code{int}. In addition to avoiding errors in certain
3842 cases of mismatch, it also makes for better code on certain machines.
3843 The default is to not promote prototypes.
3847 A C expression. If nonzero, push insns will be used to pass
3849 If the target machine does not have a push instruction, set it to zero.
3850 That directs GCC to use an alternate strategy: to
3851 allocate the entire argument block and then store the arguments into
3852 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3855 @defmac PUSH_ARGS_REVERSED
3856 A C expression. If nonzero, function arguments will be evaluated from
3857 last to first, rather than from first to last. If this macro is not
3858 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3859 and args grow in opposite directions, and 0 otherwise.
3862 @defmac PUSH_ROUNDING (@var{npushed})
3863 A C expression that is the number of bytes actually pushed onto the
3864 stack when an instruction attempts to push @var{npushed} bytes.
3866 On some machines, the definition
3869 #define PUSH_ROUNDING(BYTES) (BYTES)
3873 will suffice. But on other machines, instructions that appear
3874 to push one byte actually push two bytes in an attempt to maintain
3875 alignment. Then the definition should be
3878 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3881 If the value of this macro has a type, it should be an unsigned type.
3884 @findex current_function_outgoing_args_size
3885 @defmac ACCUMULATE_OUTGOING_ARGS
3886 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3887 will be computed and placed into the variable
3888 @code{current_function_outgoing_args_size}. No space will be pushed
3889 onto the stack for each call; instead, the function prologue should
3890 increase the stack frame size by this amount.
3892 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3896 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3897 Define this macro if functions should assume that stack space has been
3898 allocated for arguments even when their values are passed in
3901 The value of this macro is the size, in bytes, of the area reserved for
3902 arguments passed in registers for the function represented by @var{fndecl},
3903 which can be zero if GCC is calling a library function.
3904 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3907 This space can be allocated by the caller, or be a part of the
3908 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3911 @c above is overfull. not sure what to do. --mew 5feb93 did
3912 @c something, not sure if it looks good. --mew 10feb93
3914 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3915 Define this to a nonzero value if it is the responsibility of the
3916 caller to allocate the area reserved for arguments passed in registers
3917 when calling a function of @var{fntype}. @var{fntype} may be NULL
3918 if the function called is a library function.
3920 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3921 whether the space for these arguments counts in the value of
3922 @code{current_function_outgoing_args_size}.
3925 @defmac STACK_PARMS_IN_REG_PARM_AREA
3926 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3927 stack parameters don't skip the area specified by it.
3928 @c i changed this, makes more sens and it should have taken care of the
3929 @c overfull.. not as specific, tho. --mew 5feb93
3931 Normally, when a parameter is not passed in registers, it is placed on the
3932 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3933 suppresses this behavior and causes the parameter to be passed on the
3934 stack in its natural location.
3937 @deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size})
3938 This target hook returns the number of bytes of its own arguments that
3939 a function pops on returning, or 0 if the function pops no arguments
3940 and the caller must therefore pop them all after the function returns.
3942 @var{fundecl} is a C variable whose value is a tree node that describes
3943 the function in question. Normally it is a node of type
3944 @code{FUNCTION_DECL} that describes the declaration of the function.
3945 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3947 @var{funtype} is a C variable whose value is a tree node that
3948 describes the function in question. Normally it is a node of type
3949 @code{FUNCTION_TYPE} that describes the data type of the function.
3950 From this it is possible to obtain the data types of the value and
3951 arguments (if known).
3953 When a call to a library function is being considered, @var{fundecl}
3954 will contain an identifier node for the library function. Thus, if
3955 you need to distinguish among various library functions, you can do so
3956 by their names. Note that ``library function'' in this context means
3957 a function used to perform arithmetic, whose name is known specially
3958 in the compiler and was not mentioned in the C code being compiled.
3960 @var{size} is the number of bytes of arguments passed on the
3961 stack. If a variable number of bytes is passed, it is zero, and
3962 argument popping will always be the responsibility of the calling function.
3964 On the VAX, all functions always pop their arguments, so the definition
3965 of this macro is @var{size}. On the 68000, using the standard
3966 calling convention, no functions pop their arguments, so the value of
3967 the macro is always 0 in this case. But an alternative calling
3968 convention is available in which functions that take a fixed number of
3969 arguments pop them but other functions (such as @code{printf}) pop
3970 nothing (the caller pops all). When this convention is in use,
3971 @var{funtype} is examined to determine whether a function takes a fixed
3972 number of arguments.
3975 @defmac CALL_POPS_ARGS (@var{cum})
3976 A C expression that should indicate the number of bytes a call sequence
3977 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3978 when compiling a function call.
3980 @var{cum} is the variable in which all arguments to the called function
3981 have been accumulated.
3983 On certain architectures, such as the SH5, a call trampoline is used
3984 that pops certain registers off the stack, depending on the arguments
3985 that have been passed to the function. Since this is a property of the
3986 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3990 @node Register Arguments
3991 @subsection Passing Arguments in Registers
3992 @cindex arguments in registers
3993 @cindex registers arguments
3995 This section describes the macros which let you control how various
3996 types of arguments are passed in registers or how they are arranged in
3999 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4000 Return an RTX indicating whether a function argument is passed in a
4001 register and if so, which register.
4003 The arguments are @var{ca}, which summarizes all the previous
4004 arguments; @var{mode}, the machine mode of the argument; @var{type},
4005 the data type of the argument as a tree node or 0 if that is not known
4006 (which happens for C support library functions); and @var{named},
4007 which is @code{true} for an ordinary argument and @code{false} for
4008 nameless arguments that correspond to @samp{@dots{}} in the called
4009 function's prototype. @var{type} can be an incomplete type if a
4010 syntax error has previously occurred.
4012 The return value is usually either a @code{reg} RTX for the hard
4013 register in which to pass the argument, or zero to pass the argument
4016 The value of the expression can also be a @code{parallel} RTX@. This is
4017 used when an argument is passed in multiple locations. The mode of the
4018 @code{parallel} should be the mode of the entire argument. The
4019 @code{parallel} holds any number of @code{expr_list} pairs; each one
4020 describes where part of the argument is passed. In each
4021 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
4022 register in which to pass this part of the argument, and the mode of the
4023 register RTX indicates how large this part of the argument is. The
4024 second operand of the @code{expr_list} is a @code{const_int} which gives
4025 the offset in bytes into the entire argument of where this part starts.
4026 As a special exception the first @code{expr_list} in the @code{parallel}
4027 RTX may have a first operand of zero. This indicates that the entire
4028 argument is also stored on the stack.
4030 The last time this hook is called, it is called with @code{MODE ==
4031 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
4032 pattern as operands 2 and 3 respectively.
4034 @cindex @file{stdarg.h} and register arguments
4035 The usual way to make the ISO library @file{stdarg.h} work on a
4036 machine where some arguments are usually passed in registers, is to
4037 cause nameless arguments to be passed on the stack instead. This is
4038 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
4039 @var{named} is @code{false}.
4041 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
4042 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
4043 You may use the hook @code{targetm.calls.must_pass_in_stack}
4044 in the definition of this macro to determine if this argument is of a
4045 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4046 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
4047 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4048 defined, the argument will be computed in the stack and then loaded into
4052 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, const_tree @var{type})
4053 This target hook should return @code{true} if we should not pass @var{type}
4054 solely in registers. The file @file{expr.h} defines a
4055 definition that is usually appropriate, refer to @file{expr.h} for additional
4059 @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})
4060 Define this hook if the target machine has ``register windows'', so
4061 that the register in which a function sees an arguments is not
4062 necessarily the same as the one in which the caller passed the
4065 For such machines, @code{TARGET_FUNCTION_ARG} computes the register in
4066 which the caller passes the value, and
4067 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4068 fashion to tell the function being called where the arguments will
4071 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4072 @code{TARGET_FUNCTION_ARG} serves both purposes.
4075 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4076 This target hook returns the number of bytes at the beginning of an
4077 argument that must be put in registers. The value must be zero for
4078 arguments that are passed entirely in registers or that are entirely
4079 pushed on the stack.
4081 On some machines, certain arguments must be passed partially in
4082 registers and partially in memory. On these machines, typically the
4083 first few words of arguments are passed in registers, and the rest
4084 on the stack. If a multi-word argument (a @code{double} or a
4085 structure) crosses that boundary, its first few words must be passed
4086 in registers and the rest must be pushed. This macro tells the
4087 compiler when this occurs, and how many bytes should go in registers.
4089 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4090 register to be used by the caller for this argument; likewise
4091 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4094 @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})
4095 This target hook should return @code{true} if an argument at the
4096 position indicated by @var{cum} should be passed by reference. This
4097 predicate is queried after target independent reasons for being
4098 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4100 If the hook returns true, a copy of that argument is made in memory and a
4101 pointer to the argument is passed instead of the argument itself.
4102 The pointer is passed in whatever way is appropriate for passing a pointer
4106 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4107 The function argument described by the parameters to this hook is
4108 known to be passed by reference. The hook should return true if the
4109 function argument should be copied by the callee instead of copied
4112 For any argument for which the hook returns true, if it can be
4113 determined that the argument is not modified, then a copy need
4116 The default version of this hook always returns false.
4119 @defmac CUMULATIVE_ARGS
4120 A C type for declaring a variable that is used as the first argument
4121 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4122 target machines, the type @code{int} suffices and can hold the number
4123 of bytes of argument so far.
4125 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4126 arguments that have been passed on the stack. The compiler has other
4127 variables to keep track of that. For target machines on which all
4128 arguments are passed on the stack, there is no need to store anything in
4129 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4130 should not be empty, so use @code{int}.
4133 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4134 If defined, this macro is called before generating any code for a
4135 function, but after the @var{cfun} descriptor for the function has been
4136 created. The back end may use this macro to update @var{cfun} to
4137 reflect an ABI other than that which would normally be used by default.
4138 If the compiler is generating code for a compiler-generated function,
4139 @var{fndecl} may be @code{NULL}.
4142 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4143 A C statement (sans semicolon) for initializing the variable
4144 @var{cum} for the state at the beginning of the argument list. The
4145 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4146 is the tree node for the data type of the function which will receive
4147 the args, or 0 if the args are to a compiler support library function.
4148 For direct calls that are not libcalls, @var{fndecl} contain the
4149 declaration node of the function. @var{fndecl} is also set when
4150 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4151 being compiled. @var{n_named_args} is set to the number of named
4152 arguments, including a structure return address if it is passed as a
4153 parameter, when making a call. When processing incoming arguments,
4154 @var{n_named_args} is set to @minus{}1.
4156 When processing a call to a compiler support library function,
4157 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4158 contains the name of the function, as a string. @var{libname} is 0 when
4159 an ordinary C function call is being processed. Thus, each time this
4160 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4161 never both of them at once.
4164 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4165 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4166 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4167 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4168 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4169 0)} is used instead.
4172 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4173 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4174 finding the arguments for the function being compiled. If this macro is
4175 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4177 The value passed for @var{libname} is always 0, since library routines
4178 with special calling conventions are never compiled with GCC@. The
4179 argument @var{libname} exists for symmetry with
4180 @code{INIT_CUMULATIVE_ARGS}.
4181 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4182 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4185 @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})
4186 This hook updates the summarizer variable pointed to by @var{ca} to
4187 advance past an argument in the argument list. The values @var{mode},
4188 @var{type} and @var{named} describe that argument. Once this is done,
4189 the variable @var{cum} is suitable for analyzing the @emph{following}
4190 argument with @code{TARGET_FUNCTION_ARG}, etc.
4192 This hook need not do anything if the argument in question was passed
4193 on the stack. The compiler knows how to track the amount of stack space
4194 used for arguments without any special help.
4197 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4198 If defined, a C expression that is the number of bytes to add to the
4199 offset of the argument passed in memory. This is needed for the SPU,
4200 which passes @code{char} and @code{short} arguments in the preferred
4201 slot that is in the middle of the quad word instead of starting at the
4205 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4206 If defined, a C expression which determines whether, and in which direction,
4207 to pad out an argument with extra space. The value should be of type
4208 @code{enum direction}: either @code{upward} to pad above the argument,
4209 @code{downward} to pad below, or @code{none} to inhibit padding.
4211 The @emph{amount} of padding is not controlled by this macro, but by the
4212 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
4213 always just enough to reach the next multiple of that boundary.
4215 This macro has a default definition which is right for most systems.
4216 For little-endian machines, the default is to pad upward. For
4217 big-endian machines, the default is to pad downward for an argument of
4218 constant size shorter than an @code{int}, and upward otherwise.
4221 @defmac PAD_VARARGS_DOWN
4222 If defined, a C expression which determines whether the default
4223 implementation of va_arg will attempt to pad down before reading the
4224 next argument, if that argument is smaller than its aligned space as
4225 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4226 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4229 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4230 Specify padding for the last element of a block move between registers and
4231 memory. @var{first} is nonzero if this is the only element. Defining this
4232 macro allows better control of register function parameters on big-endian
4233 machines, without using @code{PARALLEL} rtl. In particular,
4234 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4235 registers, as there is no longer a "wrong" part of a register; For example,
4236 a three byte aggregate may be passed in the high part of a register if so
4240 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4241 This hook returns the alignment boundary, in bits, of an argument
4242 with the specified mode and type. The default hook returns
4243 @code{PARM_BOUNDARY} for all arguments.
4246 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_ROUND_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4247 Normally, the size of an argument is rounded up to @code{PARM_BOUNDARY},
4248 which is the default value for this hook. You can define this hook to
4249 return a different value if an argument size must be rounded to a larger
4253 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4254 A C expression that is nonzero if @var{regno} is the number of a hard
4255 register in which function arguments are sometimes passed. This does
4256 @emph{not} include implicit arguments such as the static chain and
4257 the structure-value address. On many machines, no registers can be
4258 used for this purpose since all function arguments are pushed on the
4262 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4263 This hook should return true if parameter of type @var{type} are passed
4264 as two scalar parameters. By default, GCC will attempt to pack complex
4265 arguments into the target's word size. Some ABIs require complex arguments
4266 to be split and treated as their individual components. For example, on
4267 AIX64, complex floats should be passed in a pair of floating point
4268 registers, even though a complex float would fit in one 64-bit floating
4271 The default value of this hook is @code{NULL}, which is treated as always
4275 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4276 This hook returns a type node for @code{va_list} for the target.
4277 The default version of the hook returns @code{void*}.
4280 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4281 This target hook is used in function @code{c_common_nodes_and_builtins}
4282 to iterate through the target specific builtin types for va_list. The
4283 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4284 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4286 The arguments @var{pname} and @var{ptree} are used to store the result of
4287 this macro and are set to the name of the va_list builtin type and its
4289 If the return value of this macro is zero, then there is no more element.
4290 Otherwise the @var{IDX} should be increased for the next call of this
4291 macro to iterate through all types.
4294 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4295 This hook returns the va_list type of the calling convention specified by
4297 The default version of this hook returns @code{va_list_type_node}.
4300 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4301 This hook returns the va_list type of the calling convention specified by the
4302 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4306 @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})
4307 This hook performs target-specific gimplification of
4308 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4309 arguments to @code{va_arg}; the latter two are as in
4310 @code{gimplify.c:gimplify_expr}.
4313 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4314 Define this to return nonzero if the port can handle pointers
4315 with machine mode @var{mode}. The default version of this
4316 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4319 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (struct ao_ref_s *@var{ref})
4320 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.
4323 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4324 Define this to return nonzero if the port is prepared to handle
4325 insns involving scalar mode @var{mode}. For a scalar mode to be
4326 considered supported, all the basic arithmetic and comparisons
4329 The default version of this hook returns true for any mode
4330 required to handle the basic C types (as defined by the port).
4331 Included here are the double-word arithmetic supported by the
4332 code in @file{optabs.c}.
4335 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4336 Define this to return nonzero if the port is prepared to handle
4337 insns involving vector mode @var{mode}. At the very least, it
4338 must have move patterns for this mode.
4341 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (enum machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4342 Return true if GCC should try to use a scalar mode to store an array
4343 of @var{nelems} elements, given that each element has mode @var{mode}.
4344 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4345 and allows GCC to use any defined integer mode.
4347 One use of this hook is to support vector load and store operations
4348 that operate on several homogeneous vectors. For example, ARM NEON
4349 has operations like:
4352 int8x8x3_t vld3_s8 (const int8_t *)
4355 where the return type is defined as:
4358 typedef struct int8x8x3_t
4364 If this hook allows @code{val} to have a scalar mode, then
4365 @code{int8x8x3_t} can have the same mode. GCC can then store
4366 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4369 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (enum machine_mode @var{mode})
4370 Define this to return nonzero for machine modes for which the port has
4371 small register classes. If this target hook returns nonzero for a given
4372 @var{mode}, the compiler will try to minimize the lifetime of registers
4373 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4374 In this case, the hook is expected to return nonzero if it returns nonzero
4377 On some machines, it is risky to let hard registers live across arbitrary
4378 insns. Typically, these machines have instructions that require values
4379 to be in specific registers (like an accumulator), and reload will fail
4380 if the required hard register is used for another purpose across such an
4383 Passes before reload do not know which hard registers will be used
4384 in an instruction, but the machine modes of the registers set or used in
4385 the instruction are already known. And for some machines, register
4386 classes are small for, say, integer registers but not for floating point
4387 registers. For example, the AMD x86-64 architecture requires specific
4388 registers for the legacy x86 integer instructions, but there are many
4389 SSE registers for floating point operations. On such targets, a good
4390 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4391 machine modes but zero for the SSE register classes.
4393 The default version of this hook returns false for any mode. It is always
4394 safe to redefine this hook to return with a nonzero value. But if you
4395 unnecessarily define it, you will reduce the amount of optimizations
4396 that can be performed in some cases. If you do not define this hook
4397 to return a nonzero value when it is required, the compiler will run out
4398 of spill registers and print a fatal error message.
4401 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
4402 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.
4406 @subsection How Scalar Function Values Are Returned
4407 @cindex return values in registers
4408 @cindex values, returned by functions
4409 @cindex scalars, returned as values
4411 This section discusses the macros that control returning scalars as
4412 values---values that can fit in registers.
4414 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4416 Define this to return an RTX representing the place where a function
4417 returns or receives a value of data type @var{ret_type}, a tree node
4418 representing a data type. @var{fn_decl_or_type} is a tree node
4419 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4420 function being called. If @var{outgoing} is false, the hook should
4421 compute the register in which the caller will see the return value.
4422 Otherwise, the hook should return an RTX representing the place where
4423 a function returns a value.
4425 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4426 (Actually, on most machines, scalar values are returned in the same
4427 place regardless of mode.) The value of the expression is usually a
4428 @code{reg} RTX for the hard register where the return value is stored.
4429 The value can also be a @code{parallel} RTX, if the return value is in
4430 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4431 @code{parallel} form. Note that the callee will populate every
4432 location specified in the @code{parallel}, but if the first element of
4433 the @code{parallel} contains the whole return value, callers will use
4434 that element as the canonical location and ignore the others. The m68k
4435 port uses this type of @code{parallel} to return pointers in both
4436 @samp{%a0} (the canonical location) and @samp{%d0}.
4438 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4439 the same promotion rules specified in @code{PROMOTE_MODE} if
4440 @var{valtype} is a scalar type.
4442 If the precise function being called is known, @var{func} is a tree
4443 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4444 pointer. This makes it possible to use a different value-returning
4445 convention for specific functions when all their calls are
4448 Some target machines have ``register windows'' so that the register in
4449 which a function returns its value is not the same as the one in which
4450 the caller sees the value. For such machines, you should return
4451 different RTX depending on @var{outgoing}.
4453 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4454 aggregate data types, because these are returned in another way. See
4455 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4458 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4459 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4460 a new target instead.
4463 @defmac LIBCALL_VALUE (@var{mode})
4464 A C expression to create an RTX representing the place where a library
4465 function returns a value of mode @var{mode}.
4467 Note that ``library function'' in this context means a compiler
4468 support routine, used to perform arithmetic, whose name is known
4469 specially by the compiler and was not mentioned in the C code being
4473 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (enum machine_mode @var{mode}, const_rtx @var{fun})
4474 Define this hook if the back-end needs to know the name of the libcall
4475 function in order to determine where the result should be returned.
4477 The mode of the result is given by @var{mode} and the name of the called
4478 library function is given by @var{fun}. The hook should return an RTX
4479 representing the place where the library function result will be returned.
4481 If this hook is not defined, then LIBCALL_VALUE will be used.
4484 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4485 A C expression that is nonzero if @var{regno} is the number of a hard
4486 register in which the values of called function may come back.
4488 A register whose use for returning values is limited to serving as the
4489 second of a pair (for a value of type @code{double}, say) need not be
4490 recognized by this macro. So for most machines, this definition
4494 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4497 If the machine has register windows, so that the caller and the called
4498 function use different registers for the return value, this macro
4499 should recognize only the caller's register numbers.
4501 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4502 for a new target instead.
4505 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4506 A target hook that return @code{true} if @var{regno} is the number of a hard
4507 register in which the values of called function may come back.
4509 A register whose use for returning values is limited to serving as the
4510 second of a pair (for a value of type @code{double}, say) need not be
4511 recognized by this target hook.
4513 If the machine has register windows, so that the caller and the called
4514 function use different registers for the return value, this target hook
4515 should recognize only the caller's register numbers.
4517 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4520 @defmac APPLY_RESULT_SIZE
4521 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4522 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4523 saving and restoring an arbitrary return value.
4526 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4527 This hook should return true if values of type @var{type} are returned
4528 at the most significant end of a register (in other words, if they are
4529 padded at the least significant end). You can assume that @var{type}
4530 is returned in a register; the caller is required to check this.
4532 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4533 be able to hold the complete return value. For example, if a 1-, 2-
4534 or 3-byte structure is returned at the most significant end of a
4535 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4539 @node Aggregate Return
4540 @subsection How Large Values Are Returned
4541 @cindex aggregates as return values
4542 @cindex large return values
4543 @cindex returning aggregate values
4544 @cindex structure value address
4546 When a function value's mode is @code{BLKmode} (and in some other
4547 cases), the value is not returned according to
4548 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4549 caller passes the address of a block of memory in which the value
4550 should be stored. This address is called the @dfn{structure value
4553 This section describes how to control returning structure values in
4556 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4557 This target hook should return a nonzero value to say to return the
4558 function value in memory, just as large structures are always returned.
4559 Here @var{type} will be the data type of the value, and @var{fntype}
4560 will be the type of the function doing the returning, or @code{NULL} for
4563 Note that values of mode @code{BLKmode} must be explicitly handled
4564 by this function. Also, the option @option{-fpcc-struct-return}
4565 takes effect regardless of this macro. On most systems, it is
4566 possible to leave the hook undefined; this causes a default
4567 definition to be used, whose value is the constant 1 for @code{BLKmode}
4568 values, and 0 otherwise.
4570 Do not use this hook to indicate that structures and unions should always
4571 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4575 @defmac DEFAULT_PCC_STRUCT_RETURN
4576 Define this macro to be 1 if all structure and union return values must be
4577 in memory. Since this results in slower code, this should be defined
4578 only if needed for compatibility with other compilers or with an ABI@.
4579 If you define this macro to be 0, then the conventions used for structure
4580 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4583 If not defined, this defaults to the value 1.
4586 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4587 This target hook should return the location of the structure value
4588 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4589 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4590 be @code{NULL}, for libcalls. You do not need to define this target
4591 hook if the address is always passed as an ``invisible'' first
4594 On some architectures the place where the structure value address
4595 is found by the called function is not the same place that the
4596 caller put it. This can be due to register windows, or it could
4597 be because the function prologue moves it to a different place.
4598 @var{incoming} is @code{1} or @code{2} when the location is needed in
4599 the context of the called function, and @code{0} in the context of
4602 If @var{incoming} is nonzero and the address is to be found on the
4603 stack, return a @code{mem} which refers to the frame pointer. If
4604 @var{incoming} is @code{2}, the result is being used to fetch the
4605 structure value address at the beginning of a function. If you need
4606 to emit adjusting code, you should do it at this point.
4609 @defmac PCC_STATIC_STRUCT_RETURN
4610 Define this macro if the usual system convention on the target machine
4611 for returning structures and unions is for the called function to return
4612 the address of a static variable containing the value.
4614 Do not define this if the usual system convention is for the caller to
4615 pass an address to the subroutine.
4617 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4618 nothing when you use @option{-freg-struct-return} mode.
4621 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4622 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.
4625 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_ARG_MODE (int @var{regno})
4626 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.
4630 @subsection Caller-Saves Register Allocation
4632 If you enable it, GCC can save registers around function calls. This
4633 makes it possible to use call-clobbered registers to hold variables that
4634 must live across calls.
4636 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4637 A C expression to determine whether it is worthwhile to consider placing
4638 a pseudo-register in a call-clobbered hard register and saving and
4639 restoring it around each function call. The expression should be 1 when
4640 this is worth doing, and 0 otherwise.
4642 If you don't define this macro, a default is used which is good on most
4643 machines: @code{4 * @var{calls} < @var{refs}}.
4646 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4647 A C expression specifying which mode is required for saving @var{nregs}
4648 of a pseudo-register in call-clobbered hard register @var{regno}. If
4649 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4650 returned. For most machines this macro need not be defined since GCC
4651 will select the smallest suitable mode.
4654 @node Function Entry
4655 @subsection Function Entry and Exit
4656 @cindex function entry and exit
4660 This section describes the macros that output function entry
4661 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4663 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4664 If defined, a function that outputs the assembler code for entry to a
4665 function. The prologue is responsible for setting up the stack frame,
4666 initializing the frame pointer register, saving registers that must be
4667 saved, and allocating @var{size} additional bytes of storage for the
4668 local variables. @var{size} is an integer. @var{file} is a stdio
4669 stream to which the assembler code should be output.
4671 The label for the beginning of the function need not be output by this
4672 macro. That has already been done when the macro is run.
4674 @findex regs_ever_live
4675 To determine which registers to save, the macro can refer to the array
4676 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4677 @var{r} is used anywhere within the function. This implies the function
4678 prologue should save register @var{r}, provided it is not one of the
4679 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4680 @code{regs_ever_live}.)
4682 On machines that have ``register windows'', the function entry code does
4683 not save on the stack the registers that are in the windows, even if
4684 they are supposed to be preserved by function calls; instead it takes
4685 appropriate steps to ``push'' the register stack, if any non-call-used
4686 registers are used in the function.
4688 @findex frame_pointer_needed
4689 On machines where functions may or may not have frame-pointers, the
4690 function entry code must vary accordingly; it must set up the frame
4691 pointer if one is wanted, and not otherwise. To determine whether a
4692 frame pointer is in wanted, the macro can refer to the variable
4693 @code{frame_pointer_needed}. The variable's value will be 1 at run
4694 time in a function that needs a frame pointer. @xref{Elimination}.
4696 The function entry code is responsible for allocating any stack space
4697 required for the function. This stack space consists of the regions
4698 listed below. In most cases, these regions are allocated in the
4699 order listed, with the last listed region closest to the top of the
4700 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4701 the highest address if it is not defined). You can use a different order
4702 for a machine if doing so is more convenient or required for
4703 compatibility reasons. Except in cases where required by standard
4704 or by a debugger, there is no reason why the stack layout used by GCC
4705 need agree with that used by other compilers for a machine.
4708 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4709 If defined, a function that outputs assembler code at the end of a
4710 prologue. This should be used when the function prologue is being
4711 emitted as RTL, and you have some extra assembler that needs to be
4712 emitted. @xref{prologue instruction pattern}.
4715 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4716 If defined, a function that outputs assembler code at the start of an
4717 epilogue. This should be used when the function epilogue is being
4718 emitted as RTL, and you have some extra assembler that needs to be
4719 emitted. @xref{epilogue instruction pattern}.
4722 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4723 If defined, a function that outputs the assembler code for exit from a
4724 function. The epilogue is responsible for restoring the saved
4725 registers and stack pointer to their values when the function was
4726 called, and returning control to the caller. This macro takes the
4727 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4728 registers to restore are determined from @code{regs_ever_live} and
4729 @code{CALL_USED_REGISTERS} in the same way.
4731 On some machines, there is a single instruction that does all the work
4732 of returning from the function. On these machines, give that
4733 instruction the name @samp{return} and do not define the macro
4734 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4736 Do not define a pattern named @samp{return} if you want the
4737 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4738 switches to control whether return instructions or epilogues are used,
4739 define a @samp{return} pattern with a validity condition that tests the
4740 target switches appropriately. If the @samp{return} pattern's validity
4741 condition is false, epilogues will be used.
4743 On machines where functions may or may not have frame-pointers, the
4744 function exit code must vary accordingly. Sometimes the code for these
4745 two cases is completely different. To determine whether a frame pointer
4746 is wanted, the macro can refer to the variable
4747 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4748 a function that needs a frame pointer.
4750 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4751 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4752 The C variable @code{current_function_is_leaf} is nonzero for such a
4753 function. @xref{Leaf Functions}.
4755 On some machines, some functions pop their arguments on exit while
4756 others leave that for the caller to do. For example, the 68020 when
4757 given @option{-mrtd} pops arguments in functions that take a fixed
4758 number of arguments.
4760 @findex current_function_pops_args
4761 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4762 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4763 needs to know what was decided. The number of bytes of the current
4764 function's arguments that this function should pop is available in
4765 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4770 @findex current_function_pretend_args_size
4771 A region of @code{current_function_pretend_args_size} bytes of
4772 uninitialized space just underneath the first argument arriving on the
4773 stack. (This may not be at the very start of the allocated stack region
4774 if the calling sequence has pushed anything else since pushing the stack
4775 arguments. But usually, on such machines, nothing else has been pushed
4776 yet, because the function prologue itself does all the pushing.) This
4777 region is used on machines where an argument may be passed partly in
4778 registers and partly in memory, and, in some cases to support the
4779 features in @code{<stdarg.h>}.
4782 An area of memory used to save certain registers used by the function.
4783 The size of this area, which may also include space for such things as
4784 the return address and pointers to previous stack frames, is
4785 machine-specific and usually depends on which registers have been used
4786 in the function. Machines with register windows often do not require
4790 A region of at least @var{size} bytes, possibly rounded up to an allocation
4791 boundary, to contain the local variables of the function. On some machines,
4792 this region and the save area may occur in the opposite order, with the
4793 save area closer to the top of the stack.
4796 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4797 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4798 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4799 argument lists of the function. @xref{Stack Arguments}.
4802 @defmac EXIT_IGNORE_STACK
4803 Define this macro as a C expression that is nonzero if the return
4804 instruction or the function epilogue ignores the value of the stack
4805 pointer; in other words, if it is safe to delete an instruction to
4806 adjust the stack pointer before a return from the function. The
4809 Note that this macro's value is relevant only for functions for which
4810 frame pointers are maintained. It is never safe to delete a final
4811 stack adjustment in a function that has no frame pointer, and the
4812 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4815 @defmac EPILOGUE_USES (@var{regno})
4816 Define this macro as a C expression that is nonzero for registers that are
4817 used by the epilogue or the @samp{return} pattern. The stack and frame
4818 pointer registers are already assumed to be used as needed.
4821 @defmac EH_USES (@var{regno})
4822 Define this macro as a C expression that is nonzero for registers that are
4823 used by the exception handling mechanism, and so should be considered live
4824 on entry to an exception edge.
4827 @defmac DELAY_SLOTS_FOR_EPILOGUE
4828 Define this macro if the function epilogue contains delay slots to which
4829 instructions from the rest of the function can be ``moved''. The
4830 definition should be a C expression whose value is an integer
4831 representing the number of delay slots there.
4834 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4835 A C expression that returns 1 if @var{insn} can be placed in delay
4836 slot number @var{n} of the epilogue.
4838 The argument @var{n} is an integer which identifies the delay slot now
4839 being considered (since different slots may have different rules of
4840 eligibility). It is never negative and is always less than the number
4841 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4842 If you reject a particular insn for a given delay slot, in principle, it
4843 may be reconsidered for a subsequent delay slot. Also, other insns may
4844 (at least in principle) be considered for the so far unfilled delay
4847 @findex current_function_epilogue_delay_list
4848 @findex final_scan_insn
4849 The insns accepted to fill the epilogue delay slots are put in an RTL
4850 list made with @code{insn_list} objects, stored in the variable
4851 @code{current_function_epilogue_delay_list}. The insn for the first
4852 delay slot comes first in the list. Your definition of the macro
4853 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4854 outputting the insns in this list, usually by calling
4855 @code{final_scan_insn}.
4857 You need not define this macro if you did not define
4858 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4861 @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})
4862 A function that outputs the assembler code for a thunk
4863 function, used to implement C++ virtual function calls with multiple
4864 inheritance. The thunk acts as a wrapper around a virtual function,
4865 adjusting the implicit object parameter before handing control off to
4868 First, emit code to add the integer @var{delta} to the location that
4869 contains the incoming first argument. Assume that this argument
4870 contains a pointer, and is the one used to pass the @code{this} pointer
4871 in C++. This is the incoming argument @emph{before} the function prologue,
4872 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4873 all other incoming arguments.
4875 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4876 made after adding @code{delta}. In particular, if @var{p} is the
4877 adjusted pointer, the following adjustment should be made:
4880 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4883 After the additions, emit code to jump to @var{function}, which is a
4884 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4885 not touch the return address. Hence returning from @var{FUNCTION} will
4886 return to whoever called the current @samp{thunk}.
4888 The effect must be as if @var{function} had been called directly with
4889 the adjusted first argument. This macro is responsible for emitting all
4890 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4891 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4893 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4894 have already been extracted from it.) It might possibly be useful on
4895 some targets, but probably not.
4897 If you do not define this macro, the target-independent code in the C++
4898 front end will generate a less efficient heavyweight thunk that calls
4899 @var{function} instead of jumping to it. The generic approach does
4900 not support varargs.
4903 @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})
4904 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4905 to output the assembler code for the thunk function specified by the
4906 arguments it is passed, and false otherwise. In the latter case, the
4907 generic approach will be used by the C++ front end, with the limitations
4912 @subsection Generating Code for Profiling
4913 @cindex profiling, code generation
4915 These macros will help you generate code for profiling.
4917 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4918 A C statement or compound statement to output to @var{file} some
4919 assembler code to call the profiling subroutine @code{mcount}.
4922 The details of how @code{mcount} expects to be called are determined by
4923 your operating system environment, not by GCC@. To figure them out,
4924 compile a small program for profiling using the system's installed C
4925 compiler and look at the assembler code that results.
4927 Older implementations of @code{mcount} expect the address of a counter
4928 variable to be loaded into some register. The name of this variable is
4929 @samp{LP} followed by the number @var{labelno}, so you would generate
4930 the name using @samp{LP%d} in a @code{fprintf}.
4933 @defmac PROFILE_HOOK
4934 A C statement or compound statement to output to @var{file} some assembly
4935 code to call the profiling subroutine @code{mcount} even the target does
4936 not support profiling.
4939 @defmac NO_PROFILE_COUNTERS
4940 Define this macro to be an expression with a nonzero value if the
4941 @code{mcount} subroutine on your system does not need a counter variable
4942 allocated for each function. This is true for almost all modern
4943 implementations. If you define this macro, you must not use the
4944 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4947 @defmac PROFILE_BEFORE_PROLOGUE
4948 Define this macro if the code for function profiling should come before
4949 the function prologue. Normally, the profiling code comes after.
4953 @subsection Permitting tail calls
4956 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4957 True if it is ok to do sibling call optimization for the specified
4958 call expression @var{exp}. @var{decl} will be the called function,
4959 or @code{NULL} if this is an indirect call.
4961 It is not uncommon for limitations of calling conventions to prevent
4962 tail calls to functions outside the current unit of translation, or
4963 during PIC compilation. The hook is used to enforce these restrictions,
4964 as the @code{sibcall} md pattern can not fail, or fall over to a
4965 ``normal'' call. The criteria for successful sibling call optimization
4966 may vary greatly between different architectures.
4969 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4970 Add any hard registers to @var{regs} that are live on entry to the
4971 function. This hook only needs to be defined to provide registers that
4972 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4973 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4974 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4975 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4978 @deftypefn {Target Hook} void TARGET_SET_UP_BY_PROLOGUE (struct hard_reg_set_container *@var{})
4979 This hook should add additional registers that are computed by the prologue to the hard regset for shrink-wrapping optimization purposes.
4982 @node Stack Smashing Protection
4983 @subsection Stack smashing protection
4984 @cindex stack smashing protection
4986 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4987 This hook returns a @code{DECL} node for the external variable to use
4988 for the stack protection guard. This variable is initialized by the
4989 runtime to some random value and is used to initialize the guard value
4990 that is placed at the top of the local stack frame. The type of this
4991 variable must be @code{ptr_type_node}.
4993 The default version of this hook creates a variable called
4994 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4997 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4998 This hook returns a tree expression that alerts the runtime that the
4999 stack protect guard variable has been modified. This expression should
5000 involve a call to a @code{noreturn} function.
5002 The default version of this hook invokes a function called
5003 @samp{__stack_chk_fail}, taking no arguments. This function is
5004 normally defined in @file{libgcc2.c}.
5007 @deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
5008 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
5012 @section Implementing the Varargs Macros
5013 @cindex varargs implementation
5015 GCC comes with an implementation of @code{<varargs.h>} and
5016 @code{<stdarg.h>} that work without change on machines that pass arguments
5017 on the stack. Other machines require their own implementations of
5018 varargs, and the two machine independent header files must have
5019 conditionals to include it.
5021 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
5022 the calling convention for @code{va_start}. The traditional
5023 implementation takes just one argument, which is the variable in which
5024 to store the argument pointer. The ISO implementation of
5025 @code{va_start} takes an additional second argument. The user is
5026 supposed to write the last named argument of the function here.
5028 However, @code{va_start} should not use this argument. The way to find
5029 the end of the named arguments is with the built-in functions described
5032 @defmac __builtin_saveregs ()
5033 Use this built-in function to save the argument registers in memory so
5034 that the varargs mechanism can access them. Both ISO and traditional
5035 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
5036 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
5038 On some machines, @code{__builtin_saveregs} is open-coded under the
5039 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
5040 other machines, it calls a routine written in assembler language,
5041 found in @file{libgcc2.c}.
5043 Code generated for the call to @code{__builtin_saveregs} appears at the
5044 beginning of the function, as opposed to where the call to
5045 @code{__builtin_saveregs} is written, regardless of what the code is.
5046 This is because the registers must be saved before the function starts
5047 to use them for its own purposes.
5048 @c i rewrote the first sentence above to fix an overfull hbox. --mew
5052 @defmac __builtin_next_arg (@var{lastarg})
5053 This builtin returns the address of the first anonymous stack
5054 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
5055 returns the address of the location above the first anonymous stack
5056 argument. Use it in @code{va_start} to initialize the pointer for
5057 fetching arguments from the stack. Also use it in @code{va_start} to
5058 verify that the second parameter @var{lastarg} is the last named argument
5059 of the current function.
5062 @defmac __builtin_classify_type (@var{object})
5063 Since each machine has its own conventions for which data types are
5064 passed in which kind of register, your implementation of @code{va_arg}
5065 has to embody these conventions. The easiest way to categorize the
5066 specified data type is to use @code{__builtin_classify_type} together
5067 with @code{sizeof} and @code{__alignof__}.
5069 @code{__builtin_classify_type} ignores the value of @var{object},
5070 considering only its data type. It returns an integer describing what
5071 kind of type that is---integer, floating, pointer, structure, and so on.
5073 The file @file{typeclass.h} defines an enumeration that you can use to
5074 interpret the values of @code{__builtin_classify_type}.
5077 These machine description macros help implement varargs:
5079 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5080 If defined, this hook produces the machine-specific code for a call to
5081 @code{__builtin_saveregs}. This code will be moved to the very
5082 beginning of the function, before any parameter access are made. The
5083 return value of this function should be an RTX that contains the value
5084 to use as the return of @code{__builtin_saveregs}.
5087 @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})
5088 This target hook offers an alternative to using
5089 @code{__builtin_saveregs} and defining the hook
5090 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5091 register arguments into the stack so that all the arguments appear to
5092 have been passed consecutively on the stack. Once this is done, you can
5093 use the standard implementation of varargs that works for machines that
5094 pass all their arguments on the stack.
5096 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5097 structure, containing the values that are obtained after processing the
5098 named arguments. The arguments @var{mode} and @var{type} describe the
5099 last named argument---its machine mode and its data type as a tree node.
5101 The target hook should do two things: first, push onto the stack all the
5102 argument registers @emph{not} used for the named arguments, and second,
5103 store the size of the data thus pushed into the @code{int}-valued
5104 variable pointed to by @var{pretend_args_size}. The value that you
5105 store here will serve as additional offset for setting up the stack
5108 Because you must generate code to push the anonymous arguments at
5109 compile time without knowing their data types,
5110 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5111 have just a single category of argument register and use it uniformly
5114 If the argument @var{second_time} is nonzero, it means that the
5115 arguments of the function are being analyzed for the second time. This
5116 happens for an inline function, which is not actually compiled until the
5117 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5118 not generate any instructions in this case.
5121 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca})
5122 Define this hook to return @code{true} if the location where a function
5123 argument is passed depends on whether or not it is a named argument.
5125 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5126 is set for varargs and stdarg functions. If this hook returns
5127 @code{true}, the @var{named} argument is always true for named
5128 arguments, and false for unnamed arguments. If it returns @code{false},
5129 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5130 then all arguments are treated as named. Otherwise, all named arguments
5131 except the last are treated as named.
5133 You need not define this hook if it always returns @code{false}.
5136 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca})
5137 If you need to conditionally change ABIs so that one works with
5138 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5139 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5140 defined, then define this hook to return @code{true} if
5141 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5142 Otherwise, you should not define this hook.
5146 @section Trampolines for Nested Functions
5147 @cindex trampolines for nested functions
5148 @cindex nested functions, trampolines for
5150 A @dfn{trampoline} is a small piece of code that is created at run time
5151 when the address of a nested function is taken. It normally resides on
5152 the stack, in the stack frame of the containing function. These macros
5153 tell GCC how to generate code to allocate and initialize a
5156 The instructions in the trampoline must do two things: load a constant
5157 address into the static chain register, and jump to the real address of
5158 the nested function. On CISC machines such as the m68k, this requires
5159 two instructions, a move immediate and a jump. Then the two addresses
5160 exist in the trampoline as word-long immediate operands. On RISC
5161 machines, it is often necessary to load each address into a register in
5162 two parts. Then pieces of each address form separate immediate
5165 The code generated to initialize the trampoline must store the variable
5166 parts---the static chain value and the function address---into the
5167 immediate operands of the instructions. On a CISC machine, this is
5168 simply a matter of copying each address to a memory reference at the
5169 proper offset from the start of the trampoline. On a RISC machine, it
5170 may be necessary to take out pieces of the address and store them
5173 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5174 This hook is called by @code{assemble_trampoline_template} to output,
5175 on the stream @var{f}, assembler code for a block of data that contains
5176 the constant parts of a trampoline. This code should not include a
5177 label---the label is taken care of automatically.
5179 If you do not define this hook, it means no template is needed
5180 for the target. Do not define this hook on systems where the block move
5181 code to copy the trampoline into place would be larger than the code
5182 to generate it on the spot.
5185 @defmac TRAMPOLINE_SECTION
5186 Return the section into which the trampoline template is to be placed
5187 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5190 @defmac TRAMPOLINE_SIZE
5191 A C expression for the size in bytes of the trampoline, as an integer.
5194 @defmac TRAMPOLINE_ALIGNMENT
5195 Alignment required for trampolines, in bits.
5197 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5198 is used for aligning trampolines.
5201 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5202 This hook is called to initialize a trampoline.
5203 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5204 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5205 RTX for the static chain value that should be passed to the function
5208 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5209 first thing this hook should do is emit a block move into @var{m_tramp}
5210 from the memory block returned by @code{assemble_trampoline_template}.
5211 Note that the block move need only cover the constant parts of the
5212 trampoline. If the target isolates the variable parts of the trampoline
5213 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5215 If the target requires any other actions, such as flushing caches or
5216 enabling stack execution, these actions should be performed after
5217 initializing the trampoline proper.
5220 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5221 This hook should perform any machine-specific adjustment in
5222 the address of the trampoline. Its argument contains the address of the
5223 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5224 the address to be used for a function call should be different from the
5225 address at which the template was stored, the different address should
5226 be returned; otherwise @var{addr} should be returned unchanged.
5227 If this hook is not defined, @var{addr} will be used for function calls.
5230 Implementing trampolines is difficult on many machines because they have
5231 separate instruction and data caches. Writing into a stack location
5232 fails to clear the memory in the instruction cache, so when the program
5233 jumps to that location, it executes the old contents.
5235 Here are two possible solutions. One is to clear the relevant parts of
5236 the instruction cache whenever a trampoline is set up. The other is to
5237 make all trampolines identical, by having them jump to a standard
5238 subroutine. The former technique makes trampoline execution faster; the
5239 latter makes initialization faster.
5241 To clear the instruction cache when a trampoline is initialized, define
5242 the following macro.
5244 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5245 If defined, expands to a C expression clearing the @emph{instruction
5246 cache} in the specified interval. The definition of this macro would
5247 typically be a series of @code{asm} statements. Both @var{beg} and
5248 @var{end} are both pointer expressions.
5251 To use a standard subroutine, define the following macro. In addition,
5252 you must make sure that the instructions in a trampoline fill an entire
5253 cache line with identical instructions, or else ensure that the
5254 beginning of the trampoline code is always aligned at the same point in
5255 its cache line. Look in @file{m68k.h} as a guide.
5257 @defmac TRANSFER_FROM_TRAMPOLINE
5258 Define this macro if trampolines need a special subroutine to do their
5259 work. The macro should expand to a series of @code{asm} statements
5260 which will be compiled with GCC@. They go in a library function named
5261 @code{__transfer_from_trampoline}.
5263 If you need to avoid executing the ordinary prologue code of a compiled
5264 C function when you jump to the subroutine, you can do so by placing a
5265 special label of your own in the assembler code. Use one @code{asm}
5266 statement to generate an assembler label, and another to make the label
5267 global. Then trampolines can use that label to jump directly to your
5268 special assembler code.
5272 @section Implicit Calls to Library Routines
5273 @cindex library subroutine names
5274 @cindex @file{libgcc.a}
5276 @c prevent bad page break with this line
5277 Here is an explanation of implicit calls to library routines.
5279 @defmac DECLARE_LIBRARY_RENAMES
5280 This macro, if defined, should expand to a piece of C code that will get
5281 expanded when compiling functions for libgcc.a. It can be used to
5282 provide alternate names for GCC's internal library functions if there
5283 are ABI-mandated names that the compiler should provide.
5286 @findex set_optab_libfunc
5287 @findex init_one_libfunc
5288 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5289 This hook should declare additional library routines or rename
5290 existing ones, using the functions @code{set_optab_libfunc} and
5291 @code{init_one_libfunc} defined in @file{optabs.c}.
5292 @code{init_optabs} calls this macro after initializing all the normal
5295 The default is to do nothing. Most ports don't need to define this hook.
5298 @deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX
5299 If false (the default), internal library routines start with two
5300 underscores. If set to true, these routines start with @code{__gnu_}
5301 instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This
5302 currently only affects functions defined in @file{libgcc2.c}. If this
5303 is set to true, the @file{tm.h} file must also
5304 @code{#define LIBGCC2_GNU_PREFIX}.
5307 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5308 This macro should return @code{true} if the library routine that
5309 implements the floating point comparison operator @var{comparison} in
5310 mode @var{mode} will return a boolean, and @var{false} if it will
5313 GCC's own floating point libraries return tristates from the
5314 comparison operators, so the default returns false always. Most ports
5315 don't need to define this macro.
5318 @defmac TARGET_LIB_INT_CMP_BIASED
5319 This macro should evaluate to @code{true} if the integer comparison
5320 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5321 operand is smaller than the second, 1 to indicate that they are equal,
5322 and 2 to indicate that the first operand is greater than the second.
5323 If this macro evaluates to @code{false} the comparison functions return
5324 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5325 in @file{libgcc.a}, you do not need to define this macro.
5328 @cindex @code{EDOM}, implicit usage
5331 The value of @code{EDOM} on the target machine, as a C integer constant
5332 expression. If you don't define this macro, GCC does not attempt to
5333 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5334 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5337 If you do not define @code{TARGET_EDOM}, then compiled code reports
5338 domain errors by calling the library function and letting it report the
5339 error. If mathematical functions on your system use @code{matherr} when
5340 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5341 that @code{matherr} is used normally.
5344 @cindex @code{errno}, implicit usage
5345 @defmac GEN_ERRNO_RTX
5346 Define this macro as a C expression to create an rtl expression that
5347 refers to the global ``variable'' @code{errno}. (On certain systems,
5348 @code{errno} may not actually be a variable.) If you don't define this
5349 macro, a reasonable default is used.
5352 @cindex C99 math functions, implicit usage
5353 @defmac TARGET_C99_FUNCTIONS
5354 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5355 @code{sinf} and similarly for other functions defined by C99 standard. The
5356 default is zero because a number of existing systems lack support for these
5357 functions in their runtime so this macro needs to be redefined to one on
5358 systems that do support the C99 runtime.
5361 @cindex sincos math function, implicit usage
5362 @defmac TARGET_HAS_SINCOS
5363 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5364 and @code{cos} with the same argument to a call to @code{sincos}. The
5365 default is zero. The target has to provide the following functions:
5367 void sincos(double x, double *sin, double *cos);
5368 void sincosf(float x, float *sin, float *cos);
5369 void sincosl(long double x, long double *sin, long double *cos);
5373 @defmac NEXT_OBJC_RUNTIME
5374 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
5375 by default. This calling convention involves passing the object, the selector
5376 and the method arguments all at once to the method-lookup library function.
5377 This is the usual setting when targeting Darwin/Mac OS X systems, which have
5378 the NeXT runtime installed.
5380 If the macro is set to 0, the "GNU" Objective-C message sending convention
5381 will be used by default. This convention passes just the object and the
5382 selector to the method-lookup function, which returns a pointer to the method.
5384 In either case, it remains possible to select code-generation for the alternate
5385 scheme, by means of compiler command line switches.
5388 @node Addressing Modes
5389 @section Addressing Modes
5390 @cindex addressing modes
5392 @c prevent bad page break with this line
5393 This is about addressing modes.
5395 @defmac HAVE_PRE_INCREMENT
5396 @defmacx HAVE_PRE_DECREMENT
5397 @defmacx HAVE_POST_INCREMENT
5398 @defmacx HAVE_POST_DECREMENT
5399 A C expression that is nonzero if the machine supports pre-increment,
5400 pre-decrement, post-increment, or post-decrement addressing respectively.
5403 @defmac HAVE_PRE_MODIFY_DISP
5404 @defmacx HAVE_POST_MODIFY_DISP
5405 A C expression that is nonzero if the machine supports pre- or
5406 post-address side-effect generation involving constants other than
5407 the size of the memory operand.
5410 @defmac HAVE_PRE_MODIFY_REG
5411 @defmacx HAVE_POST_MODIFY_REG
5412 A C expression that is nonzero if the machine supports pre- or
5413 post-address side-effect generation involving a register displacement.
5416 @defmac CONSTANT_ADDRESS_P (@var{x})
5417 A C expression that is 1 if the RTX @var{x} is a constant which
5418 is a valid address. On most machines the default definition of
5419 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5420 is acceptable, but a few machines are more restrictive as to which
5421 constant addresses are supported.
5424 @defmac CONSTANT_P (@var{x})
5425 @code{CONSTANT_P}, which is defined by target-independent code,
5426 accepts integer-values expressions whose values are not explicitly
5427 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5428 expressions and @code{const} arithmetic expressions, in addition to
5429 @code{const_int} and @code{const_double} expressions.
5432 @defmac MAX_REGS_PER_ADDRESS
5433 A number, the maximum number of registers that can appear in a valid
5434 memory address. Note that it is up to you to specify a value equal to
5435 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5439 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5440 A function that returns whether @var{x} (an RTX) is a legitimate memory
5441 address on the target machine for a memory operand of mode @var{mode}.
5443 Legitimate addresses are defined in two variants: a strict variant and a
5444 non-strict one. The @var{strict} parameter chooses which variant is
5445 desired by the caller.
5447 The strict variant is used in the reload pass. It must be defined so
5448 that any pseudo-register that has not been allocated a hard register is
5449 considered a memory reference. This is because in contexts where some
5450 kind of register is required, a pseudo-register with no hard register
5451 must be rejected. For non-hard registers, the strict variant should look
5452 up the @code{reg_renumber} array; it should then proceed using the hard
5453 register number in the array, or treat the pseudo as a memory reference
5454 if the array holds @code{-1}.
5456 The non-strict variant is used in other passes. It must be defined to
5457 accept all pseudo-registers in every context where some kind of
5458 register is required.
5460 Normally, constant addresses which are the sum of a @code{symbol_ref}
5461 and an integer are stored inside a @code{const} RTX to mark them as
5462 constant. Therefore, there is no need to recognize such sums
5463 specifically as legitimate addresses. Normally you would simply
5464 recognize any @code{const} as legitimate.
5466 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5467 sums that are not marked with @code{const}. It assumes that a naked
5468 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5469 naked constant sums as illegitimate addresses, so that none of them will
5470 be given to @code{PRINT_OPERAND_ADDRESS}.
5472 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5473 On some machines, whether a symbolic address is legitimate depends on
5474 the section that the address refers to. On these machines, define the
5475 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5476 into the @code{symbol_ref}, and then check for it here. When you see a
5477 @code{const}, you will have to look inside it to find the
5478 @code{symbol_ref} in order to determine the section. @xref{Assembler
5481 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5482 Some ports are still using a deprecated legacy substitute for
5483 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5487 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5491 and should @code{goto @var{label}} if the address @var{x} is a valid
5492 address on the target machine for a memory operand of mode @var{mode}.
5494 @findex REG_OK_STRICT
5495 Compiler source files that want to use the strict variant of this
5496 macro define the macro @code{REG_OK_STRICT}. You should use an
5497 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5498 that case and the non-strict variant otherwise.
5500 Using the hook is usually simpler because it limits the number of
5501 files that are recompiled when changes are made.
5504 @defmac TARGET_MEM_CONSTRAINT
5505 A single character to be used instead of the default @code{'m'}
5506 character for general memory addresses. This defines the constraint
5507 letter which matches the memory addresses accepted by
5508 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5509 support new address formats in your back end without changing the
5510 semantics of the @code{'m'} constraint. This is necessary in order to
5511 preserve functionality of inline assembly constructs using the
5512 @code{'m'} constraint.
5515 @defmac FIND_BASE_TERM (@var{x})
5516 A C expression to determine the base term of address @var{x},
5517 or to provide a simplified version of @var{x} from which @file{alias.c}
5518 can easily find the base term. This macro is used in only two places:
5519 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5521 It is always safe for this macro to not be defined. It exists so
5522 that alias analysis can understand machine-dependent addresses.
5524 The typical use of this macro is to handle addresses containing
5525 a label_ref or symbol_ref within an UNSPEC@.
5528 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode})
5529 This hook is given an invalid memory address @var{x} for an
5530 operand of mode @var{mode} and should try to return a valid memory
5533 @findex break_out_memory_refs
5534 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5535 and @var{oldx} will be the operand that was given to that function to produce
5538 The code of the hook should not alter the substructure of
5539 @var{x}. If it transforms @var{x} into a more legitimate form, it
5540 should return the new @var{x}.
5542 It is not necessary for this hook to come up with a legitimate address.
5543 The compiler has standard ways of doing so in all cases. In fact, it
5544 is safe to omit this hook or make it return @var{x} if it cannot find
5545 a valid way to legitimize the address. But often a machine-dependent
5546 strategy can generate better code.
5549 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5550 A C compound statement that attempts to replace @var{x}, which is an address
5551 that needs reloading, with a valid memory address for an operand of mode
5552 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5553 It is not necessary to define this macro, but it might be useful for
5554 performance reasons.
5556 For example, on the i386, it is sometimes possible to use a single
5557 reload register instead of two by reloading a sum of two pseudo
5558 registers into a register. On the other hand, for number of RISC
5559 processors offsets are limited so that often an intermediate address
5560 needs to be generated in order to address a stack slot. By defining
5561 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5562 generated for adjacent some stack slots can be made identical, and thus
5565 @emph{Note}: This macro should be used with caution. It is necessary
5566 to know something of how reload works in order to effectively use this,
5567 and it is quite easy to produce macros that build in too much knowledge
5568 of reload internals.
5570 @emph{Note}: This macro must be able to reload an address created by a
5571 previous invocation of this macro. If it fails to handle such addresses
5572 then the compiler may generate incorrect code or abort.
5575 The macro definition should use @code{push_reload} to indicate parts that
5576 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5577 suitable to be passed unaltered to @code{push_reload}.
5579 The code generated by this macro must not alter the substructure of
5580 @var{x}. If it transforms @var{x} into a more legitimate form, it
5581 should assign @var{x} (which will always be a C variable) a new value.
5582 This also applies to parts that you change indirectly by calling
5585 @findex strict_memory_address_p
5586 The macro definition may use @code{strict_memory_address_p} to test if
5587 the address has become legitimate.
5590 If you want to change only a part of @var{x}, one standard way of doing
5591 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5592 single level of rtl. Thus, if the part to be changed is not at the
5593 top level, you'll need to replace first the top level.
5594 It is not necessary for this macro to come up with a legitimate
5595 address; but often a machine-dependent strategy can generate better code.
5598 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr})
5599 This hook returns @code{true} if memory address @var{addr} can have
5600 different meanings depending on the machine mode of the memory
5601 reference it is used for or if the address is valid for some modes
5604 Autoincrement and autodecrement addresses typically have mode-dependent
5605 effects because the amount of the increment or decrement is the size
5606 of the operand being addressed. Some machines have other mode-dependent
5607 addresses. Many RISC machines have no mode-dependent addresses.
5609 You may assume that @var{addr} is a valid address for the machine.
5611 The default version of this hook returns @code{false}.
5614 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5615 A C statement or compound statement with a conditional @code{goto
5616 @var{label};} executed if memory address @var{x} (an RTX) can have
5617 different meanings depending on the machine mode of the memory
5618 reference it is used for or if the address is valid for some modes
5621 Autoincrement and autodecrement addresses typically have mode-dependent
5622 effects because the amount of the increment or decrement is the size
5623 of the operand being addressed. Some machines have other mode-dependent
5624 addresses. Many RISC machines have no mode-dependent addresses.
5626 You may assume that @var{addr} is a valid address for the machine.
5628 These are obsolete macros, replaced by the
5629 @code{TARGET_MODE_DEPENDENT_ADDRESS_P} target hook.
5632 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5633 This hook returns true if @var{x} is a legitimate constant for a
5634 @var{mode}-mode immediate operand on the target machine. You can assume that
5635 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5637 The default definition returns true.
5640 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5641 This hook is used to undo the possibly obfuscating effects of the
5642 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5643 macros. Some backend implementations of these macros wrap symbol
5644 references inside an @code{UNSPEC} rtx to represent PIC or similar
5645 addressing modes. This target hook allows GCC's optimizers to understand
5646 the semantics of these opaque @code{UNSPEC}s by converting them back
5647 into their original form.
5650 @deftypefn {Target Hook} bool TARGET_CONST_NOT_OK_FOR_DEBUG_P (rtx @var{x})
5651 This hook should return true if @var{x} should not be emitted into
5655 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (enum machine_mode @var{mode}, rtx @var{x})
5656 This hook should return true if @var{x} is of a form that cannot (or
5657 should not) be spilled to the constant pool. @var{mode} is the mode
5660 The default version of this hook returns false.
5662 The primary reason to define this hook is to prevent reload from
5663 deciding that a non-legitimate constant would be better reloaded
5664 from the constant pool instead of spilling and reloading a register
5665 holding the constant. This restriction is often true of addresses
5666 of TLS symbols for various targets.
5669 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, const_rtx @var{x})
5670 This hook should return true if pool entries for constant @var{x} can
5671 be placed in an @code{object_block} structure. @var{mode} is the mode
5674 The default version returns false for all constants.
5677 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (unsigned @var{fn}, bool @var{md_fn}, bool @var{sqrt})
5678 This hook should return the DECL of a function that implements reciprocal of
5679 the builtin function with builtin function code @var{fn}, or
5680 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5681 when @var{fn} is a code of a machine-dependent builtin function. When
5682 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5683 of a square root function are performed, and only reciprocals of @code{sqrt}
5687 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5688 This hook should return the DECL of a function @var{f} that given an
5689 address @var{addr} as an argument returns a mask @var{m} that can be
5690 used to extract from two vectors the relevant data that resides in
5691 @var{addr} in case @var{addr} is not properly aligned.
5693 The autovectorizer, when vectorizing a load operation from an address
5694 @var{addr} that may be unaligned, will generate two vector loads from
5695 the two aligned addresses around @var{addr}. It then generates a
5696 @code{REALIGN_LOAD} operation to extract the relevant data from the
5697 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5698 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5699 the third argument, @var{OFF}, defines how the data will be extracted
5700 from these two vectors: if @var{OFF} is 0, then the returned vector is
5701 @var{v2}; otherwise, the returned vector is composed from the last
5702 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5703 @var{OFF} elements of @var{v2}.
5705 If this hook is defined, the autovectorizer will generate a call
5706 to @var{f} (using the DECL tree that this hook returns) and will
5707 use the return value of @var{f} as the argument @var{OFF} to
5708 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5709 should comply with the semantics expected by @code{REALIGN_LOAD}
5711 If this hook is not defined, then @var{addr} will be used as
5712 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5713 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5716 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5717 This hook should return the DECL of a function @var{f} that implements
5718 widening multiplication of the even elements of two input vectors of type @var{x}.
5720 If this hook is defined, the autovectorizer will use it along with the
5721 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5722 widening multiplication in cases that the order of the results does not have to be
5723 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5724 @code{widen_mult_hi/lo} idioms will be used.
5727 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5728 This hook should return the DECL of a function @var{f} that implements
5729 widening multiplication of the odd elements of two input vectors of type @var{x}.
5731 If this hook is defined, the autovectorizer will use it along with the
5732 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5733 widening multiplication in cases that the order of the results does not have to be
5734 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5735 @code{widen_mult_hi/lo} idioms will be used.
5738 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5739 Returns cost of different scalar or vector statements for vectorization cost model.
5740 For vector memory operations the cost may depend on type (@var{vectype}) and
5741 misalignment value (@var{misalign}).
5744 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5745 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5748 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VEC_PERM_CONST_OK (enum @var{machine_mode}, const unsigned char *@var{sel})
5749 Return true if a vector created for @code{vec_perm_const} is valid.
5752 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5753 This hook should return the DECL of a function that implements conversion of the
5754 input vector of type @var{src_type} to type @var{dest_type}.
5755 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5756 specifies how the conversion is to be applied
5757 (truncation, rounding, etc.).
5759 If this hook is defined, the autovectorizer will use the
5760 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5761 conversion. Otherwise, it will return @code{NULL_TREE}.
5764 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5765 This hook should return the decl of a function that implements the
5766 vectorized variant of the builtin function with builtin function code
5767 @var{code} or @code{NULL_TREE} if such a function is not available.
5768 The value of @var{fndecl} is the builtin function declaration. The
5769 return type of the vectorized function shall be of vector type
5770 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5773 @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})
5774 This hook should return true if the target supports misaligned vector
5775 store/load of a specific factor denoted in the @var{misalignment}
5776 parameter. The vector store/load should be of machine mode @var{mode} and
5777 the elements in the vectors should be of type @var{type}. @var{is_packed}
5778 parameter is true if the memory access is defined in a packed struct.
5781 @deftypefn {Target Hook} {enum machine_mode} TARGET_VECTORIZE_PREFERRED_SIMD_MODE (enum machine_mode @var{mode})
5782 This hook should return the preferred mode for vectorizing scalar
5783 mode @var{mode}. The default is
5784 equal to @code{word_mode}, because the vectorizer can do some
5785 transformations even in absence of specialized @acronym{SIMD} hardware.
5788 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void)
5789 This hook should return a mask of sizes that should be iterated over
5790 after trying to autovectorize using the vector size derived from the
5791 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5792 The default is zero which means to not iterate over other vector sizes.
5795 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_TM_LOAD (tree)
5796 This hook should return the built-in decl needed to load a vector of the given type within a transaction.
5799 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_TM_STORE (tree)
5800 This hook should return the built-in decl needed to store a vector of the given type within a transaction.
5803 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_GATHER (const_tree @var{mem_vectype}, const_tree @var{index_type}, int @var{scale})
5804 Target builtin that implements vector gather operation. @var{mem_vectype}
5805 is the vector type of the load and @var{index_type} is scalar type of
5806 the index, scaled by @var{scale}.
5807 The default is @code{NULL_TREE} which means to not vectorize gather
5811 @node Anchored Addresses
5812 @section Anchored Addresses
5813 @cindex anchored addresses
5814 @cindex @option{-fsection-anchors}
5816 GCC usually addresses every static object as a separate entity.
5817 For example, if we have:
5821 int foo (void) @{ return a + b + c; @}
5824 the code for @code{foo} will usually calculate three separate symbolic
5825 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5826 it would be better to calculate just one symbolic address and access
5827 the three variables relative to it. The equivalent pseudocode would
5833 register int *xr = &x;
5834 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5838 (which isn't valid C). We refer to shared addresses like @code{x} as
5839 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5841 The hooks below describe the target properties that GCC needs to know
5842 in order to make effective use of section anchors. It won't use
5843 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5844 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5846 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5847 The minimum offset that should be applied to a section anchor.
5848 On most targets, it should be the smallest offset that can be
5849 applied to a base register while still giving a legitimate address
5850 for every mode. The default value is 0.
5853 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5854 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5855 offset that should be applied to section anchors. The default
5859 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5860 Write the assembly code to define section anchor @var{x}, which is a
5861 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5862 The hook is called with the assembly output position set to the beginning
5863 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5865 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5866 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5867 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5868 is @code{NULL}, which disables the use of section anchors altogether.
5871 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
5872 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5873 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5874 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5876 The default version is correct for most targets, but you might need to
5877 intercept this hook to handle things like target-specific attributes
5878 or target-specific sections.
5881 @node Condition Code
5882 @section Condition Code Status
5883 @cindex condition code status
5885 The macros in this section can be split in two families, according to the
5886 two ways of representing condition codes in GCC.
5888 The first representation is the so called @code{(cc0)} representation
5889 (@pxref{Jump Patterns}), where all instructions can have an implicit
5890 clobber of the condition codes. The second is the condition code
5891 register representation, which provides better schedulability for
5892 architectures that do have a condition code register, but on which
5893 most instructions do not affect it. The latter category includes
5896 The implicit clobbering poses a strong restriction on the placement of
5897 the definition and use of the condition code, which need to be in adjacent
5898 insns for machines using @code{(cc0)}. This can prevent important
5899 optimizations on some machines. For example, on the IBM RS/6000, there
5900 is a delay for taken branches unless the condition code register is set
5901 three instructions earlier than the conditional branch. The instruction
5902 scheduler cannot perform this optimization if it is not permitted to
5903 separate the definition and use of the condition code register.
5905 For this reason, it is possible and suggested to use a register to
5906 represent the condition code for new ports. If there is a specific
5907 condition code register in the machine, use a hard register. If the
5908 condition code or comparison result can be placed in any general register,
5909 or if there are multiple condition registers, use a pseudo register.
5910 Registers used to store the condition code value will usually have a mode
5911 that is in class @code{MODE_CC}.
5913 Alternatively, you can use @code{BImode} if the comparison operator is
5914 specified already in the compare instruction. In this case, you are not
5915 interested in most macros in this section.
5918 * CC0 Condition Codes:: Old style representation of condition codes.
5919 * MODE_CC Condition Codes:: Modern representation of condition codes.
5920 * Cond Exec Macros:: Macros to control conditional execution.
5923 @node CC0 Condition Codes
5924 @subsection Representation of condition codes using @code{(cc0)}
5928 The file @file{conditions.h} defines a variable @code{cc_status} to
5929 describe how the condition code was computed (in case the interpretation of
5930 the condition code depends on the instruction that it was set by). This
5931 variable contains the RTL expressions on which the condition code is
5932 currently based, and several standard flags.
5934 Sometimes additional machine-specific flags must be defined in the machine
5935 description header file. It can also add additional machine-specific
5936 information by defining @code{CC_STATUS_MDEP}.
5938 @defmac CC_STATUS_MDEP
5939 C code for a data type which is used for declaring the @code{mdep}
5940 component of @code{cc_status}. It defaults to @code{int}.
5942 This macro is not used on machines that do not use @code{cc0}.
5945 @defmac CC_STATUS_MDEP_INIT
5946 A C expression to initialize the @code{mdep} field to ``empty''.
5947 The default definition does nothing, since most machines don't use
5948 the field anyway. If you want to use the field, you should probably
5949 define this macro to initialize it.
5951 This macro is not used on machines that do not use @code{cc0}.
5954 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5955 A C compound statement to set the components of @code{cc_status}
5956 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5957 this macro's responsibility to recognize insns that set the condition
5958 code as a byproduct of other activity as well as those that explicitly
5961 This macro is not used on machines that do not use @code{cc0}.
5963 If there are insns that do not set the condition code but do alter
5964 other machine registers, this macro must check to see whether they
5965 invalidate the expressions that the condition code is recorded as
5966 reflecting. For example, on the 68000, insns that store in address
5967 registers do not set the condition code, which means that usually
5968 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5969 insns. But suppose that the previous insn set the condition code
5970 based on location @samp{a4@@(102)} and the current insn stores a new
5971 value in @samp{a4}. Although the condition code is not changed by
5972 this, it will no longer be true that it reflects the contents of
5973 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5974 @code{cc_status} in this case to say that nothing is known about the
5975 condition code value.
5977 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5978 with the results of peephole optimization: insns whose patterns are
5979 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5980 constants which are just the operands. The RTL structure of these
5981 insns is not sufficient to indicate what the insns actually do. What
5982 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5983 @code{CC_STATUS_INIT}.
5985 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5986 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5987 @samp{cc}. This avoids having detailed information about patterns in
5988 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5991 @node MODE_CC Condition Codes
5992 @subsection Representation of condition codes using registers
5996 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5997 On many machines, the condition code may be produced by other instructions
5998 than compares, for example the branch can use directly the condition
5999 code set by a subtract instruction. However, on some machines
6000 when the condition code is set this way some bits (such as the overflow
6001 bit) are not set in the same way as a test instruction, so that a different
6002 branch instruction must be used for some conditional branches. When
6003 this happens, use the machine mode of the condition code register to
6004 record different formats of the condition code register. Modes can
6005 also be used to record which compare instruction (e.g. a signed or an
6006 unsigned comparison) produced the condition codes.
6008 If other modes than @code{CCmode} are required, add them to
6009 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
6010 a mode given an operand of a compare. This is needed because the modes
6011 have to be chosen not only during RTL generation but also, for example,
6012 by instruction combination. The result of @code{SELECT_CC_MODE} should
6013 be consistent with the mode used in the patterns; for example to support
6014 the case of the add on the SPARC discussed above, we have the pattern
6018 [(set (reg:CC_NOOV 0)
6020 (plus:SI (match_operand:SI 0 "register_operand" "%r")
6021 (match_operand:SI 1 "arith_operand" "rI"))
6028 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
6029 for comparisons whose argument is a @code{plus}:
6032 #define SELECT_CC_MODE(OP,X,Y) \
6033 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
6034 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
6035 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
6036 || GET_CODE (X) == NEG) \
6037 ? CC_NOOVmode : CCmode))
6040 Another reason to use modes is to retain information on which operands
6041 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
6044 You should define this macro if and only if you define extra CC modes
6045 in @file{@var{machine}-modes.def}.
6048 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
6049 On some machines not all possible comparisons are defined, but you can
6050 convert an invalid comparison into a valid one. For example, the Alpha
6051 does not have a @code{GT} comparison, but you can use an @code{LT}
6052 comparison instead and swap the order of the operands.
6054 On such machines, define this macro to be a C statement to do any
6055 required conversions. @var{code} is the initial comparison code
6056 and @var{op0} and @var{op1} are the left and right operands of the
6057 comparison, respectively. You should modify @var{code}, @var{op0}, and
6058 @var{op1} as required.
6060 GCC will not assume that the comparison resulting from this macro is
6061 valid but will see if the resulting insn matches a pattern in the
6064 You need not define this macro if it would never change the comparison
6068 @defmac REVERSIBLE_CC_MODE (@var{mode})
6069 A C expression whose value is one if it is always safe to reverse a
6070 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6071 can ever return @var{mode} for a floating-point inequality comparison,
6072 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6074 You need not define this macro if it would always returns zero or if the
6075 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6076 For example, here is the definition used on the SPARC, where floating-point
6077 inequality comparisons are always given @code{CCFPEmode}:
6080 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
6084 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6085 A C expression whose value is reversed condition code of the @var{code} for
6086 comparison done in CC_MODE @var{mode}. The macro is used only in case
6087 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6088 machine has some non-standard way how to reverse certain conditionals. For
6089 instance in case all floating point conditions are non-trapping, compiler may
6090 freely convert unordered compares to ordered one. Then definition may look
6094 #define REVERSE_CONDITION(CODE, MODE) \
6095 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6096 : reverse_condition_maybe_unordered (CODE))
6100 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6101 On targets which do not use @code{(cc0)}, and which use a hard
6102 register rather than a pseudo-register to hold condition codes, the
6103 regular CSE passes are often not able to identify cases in which the
6104 hard register is set to a common value. Use this hook to enable a
6105 small pass which optimizes such cases. This hook should return true
6106 to enable this pass, and it should set the integers to which its
6107 arguments point to the hard register numbers used for condition codes.
6108 When there is only one such register, as is true on most systems, the
6109 integer pointed to by @var{p2} should be set to
6110 @code{INVALID_REGNUM}.
6112 The default version of this hook returns false.
6115 @deftypefn {Target Hook} {enum machine_mode} TARGET_CC_MODES_COMPATIBLE (enum machine_mode @var{m1}, enum machine_mode @var{m2})
6116 On targets which use multiple condition code modes in class
6117 @code{MODE_CC}, it is sometimes the case that a comparison can be
6118 validly done in more than one mode. On such a system, define this
6119 target hook to take two mode arguments and to return a mode in which
6120 both comparisons may be validly done. If there is no such mode,
6121 return @code{VOIDmode}.
6123 The default version of this hook checks whether the modes are the
6124 same. If they are, it returns that mode. If they are different, it
6125 returns @code{VOIDmode}.
6128 @node Cond Exec Macros
6129 @subsection Macros to control conditional execution
6130 @findex conditional execution
6133 There is one macro that may need to be defined for targets
6134 supporting conditional execution, independent of how they
6135 represent conditional branches.
6137 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6138 A C expression that returns true if the conditional execution predicate
6139 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6140 versa. Define this to return 0 if the target has conditional execution
6141 predicates that cannot be reversed safely. There is no need to validate
6142 that the arguments of op1 and op2 are the same, this is done separately.
6143 If no expansion is specified, this macro is defined as follows:
6146 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6147 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6152 @section Describing Relative Costs of Operations
6153 @cindex costs of instructions
6154 @cindex relative costs
6155 @cindex speed of instructions
6157 These macros let you describe the relative speed of various operations
6158 on the target machine.
6160 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6161 A C expression for the cost of moving data of mode @var{mode} from a
6162 register in class @var{from} to one in class @var{to}. The classes are
6163 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6164 value of 2 is the default; other values are interpreted relative to
6167 It is not required that the cost always equal 2 when @var{from} is the
6168 same as @var{to}; on some machines it is expensive to move between
6169 registers if they are not general registers.
6171 If reload sees an insn consisting of a single @code{set} between two
6172 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6173 classes returns a value of 2, reload does not check to ensure that the
6174 constraints of the insn are met. Setting a cost of other than 2 will
6175 allow reload to verify that the constraints are met. You should do this
6176 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6178 These macros are obsolete, new ports should use the target hook
6179 @code{TARGET_REGISTER_MOVE_COST} instead.
6182 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6183 This target hook should return the cost of moving data of mode @var{mode}
6184 from a register in class @var{from} to one in class @var{to}. The classes
6185 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6186 A value of 2 is the default; other values are interpreted relative to
6189 It is not required that the cost always equal 2 when @var{from} is the
6190 same as @var{to}; on some machines it is expensive to move between
6191 registers if they are not general registers.
6193 If reload sees an insn consisting of a single @code{set} between two
6194 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6195 classes returns a value of 2, reload does not check to ensure that the
6196 constraints of the insn are met. Setting a cost of other than 2 will
6197 allow reload to verify that the constraints are met. You should do this
6198 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6200 The default version of this function returns 2.
6203 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6204 A C expression for the cost of moving data of mode @var{mode} between a
6205 register of class @var{class} and memory; @var{in} is zero if the value
6206 is to be written to memory, nonzero if it is to be read in. This cost
6207 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6208 registers and memory is more expensive than between two registers, you
6209 should define this macro to express the relative cost.
6211 If you do not define this macro, GCC uses a default cost of 4 plus
6212 the cost of copying via a secondary reload register, if one is
6213 needed. If your machine requires a secondary reload register to copy
6214 between memory and a register of @var{class} but the reload mechanism is
6215 more complex than copying via an intermediate, define this macro to
6216 reflect the actual cost of the move.
6218 GCC defines the function @code{memory_move_secondary_cost} if
6219 secondary reloads are needed. It computes the costs due to copying via
6220 a secondary register. If your machine copies from memory using a
6221 secondary register in the conventional way but the default base value of
6222 4 is not correct for your machine, define this macro to add some other
6223 value to the result of that function. The arguments to that function
6224 are the same as to this macro.
6226 These macros are obsolete, new ports should use the target hook
6227 @code{TARGET_MEMORY_MOVE_COST} instead.
6230 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6231 This target hook should return the cost of moving data of mode @var{mode}
6232 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6233 if the value is to be written to memory, @code{true} if it is to be read in.
6234 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6235 If moving between registers and memory is more expensive than between two
6236 registers, you should add this target hook to express the relative cost.
6238 If you do not add this target hook, GCC uses a default cost of 4 plus
6239 the cost of copying via a secondary reload register, if one is
6240 needed. If your machine requires a secondary reload register to copy
6241 between memory and a register of @var{rclass} but the reload mechanism is
6242 more complex than copying via an intermediate, use this target hook to
6243 reflect the actual cost of the move.
6245 GCC defines the function @code{memory_move_secondary_cost} if
6246 secondary reloads are needed. It computes the costs due to copying via
6247 a secondary register. If your machine copies from memory using a
6248 secondary register in the conventional way but the default base value of
6249 4 is not correct for your machine, use this target hook to add some other
6250 value to the result of that function. The arguments to that function
6251 are the same as to this target hook.
6254 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6255 A C expression for the cost of a branch instruction. A value of 1 is
6256 the default; other values are interpreted relative to that. Parameter
6257 @var{speed_p} is true when the branch in question should be optimized
6258 for speed. When it is false, @code{BRANCH_COST} should return a value
6259 optimal for code size rather than performance. @var{predictable_p} is
6260 true for well-predicted branches. On many architectures the
6261 @code{BRANCH_COST} can be reduced then.
6264 Here are additional macros which do not specify precise relative costs,
6265 but only that certain actions are more expensive than GCC would
6268 @defmac SLOW_BYTE_ACCESS
6269 Define this macro as a C expression which is nonzero if accessing less
6270 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6271 faster than accessing a word of memory, i.e., if such access
6272 require more than one instruction or if there is no difference in cost
6273 between byte and (aligned) word loads.
6275 When this macro is not defined, the compiler will access a field by
6276 finding the smallest containing object; when it is defined, a fullword
6277 load will be used if alignment permits. Unless bytes accesses are
6278 faster than word accesses, using word accesses is preferable since it
6279 may eliminate subsequent memory access if subsequent accesses occur to
6280 other fields in the same word of the structure, but to different bytes.
6283 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6284 Define this macro to be the value 1 if memory accesses described by the
6285 @var{mode} and @var{alignment} parameters have a cost many times greater
6286 than aligned accesses, for example if they are emulated in a trap
6289 When this macro is nonzero, the compiler will act as if
6290 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6291 moves. This can cause significantly more instructions to be produced.
6292 Therefore, do not set this macro nonzero if unaligned accesses only add a
6293 cycle or two to the time for a memory access.
6295 If the value of this macro is always zero, it need not be defined. If
6296 this macro is defined, it should produce a nonzero value when
6297 @code{STRICT_ALIGNMENT} is nonzero.
6300 @defmac MOVE_RATIO (@var{speed})
6301 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6302 which a sequence of insns should be generated instead of a
6303 string move insn or a library call. Increasing the value will always
6304 make code faster, but eventually incurs high cost in increased code size.
6306 Note that on machines where the corresponding move insn is a
6307 @code{define_expand} that emits a sequence of insns, this macro counts
6308 the number of such sequences.
6310 The parameter @var{speed} is true if the code is currently being
6311 optimized for speed rather than size.
6313 If you don't define this, a reasonable default is used.
6316 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6317 A C expression used to determine whether @code{move_by_pieces} will be used to
6318 copy a chunk of memory, or whether some other block move mechanism
6319 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6320 than @code{MOVE_RATIO}.
6323 @defmac MOVE_MAX_PIECES
6324 A C expression used by @code{move_by_pieces} to determine the largest unit
6325 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6328 @defmac CLEAR_RATIO (@var{speed})
6329 The threshold of number of scalar move insns, @emph{below} which a sequence
6330 of insns should be generated to clear memory instead of a string clear insn
6331 or a library call. Increasing the value will always make code faster, but
6332 eventually incurs high cost in increased code size.
6334 The parameter @var{speed} is true if the code is currently being
6335 optimized for speed rather than size.
6337 If you don't define this, a reasonable default is used.
6340 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6341 A C expression used to determine whether @code{clear_by_pieces} will be used
6342 to clear a chunk of memory, or whether some other block clear mechanism
6343 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6344 than @code{CLEAR_RATIO}.
6347 @defmac SET_RATIO (@var{speed})
6348 The threshold of number of scalar move insns, @emph{below} which a sequence
6349 of insns should be generated to set memory to a constant value, instead of
6350 a block set insn or a library call.
6351 Increasing the value will always make code faster, but
6352 eventually incurs high cost in increased code size.
6354 The parameter @var{speed} is true if the code is currently being
6355 optimized for speed rather than size.
6357 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6360 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6361 A C expression used to determine whether @code{store_by_pieces} will be
6362 used to set a chunk of memory to a constant value, or whether some
6363 other mechanism will be used. Used by @code{__builtin_memset} when
6364 storing values other than constant zero.
6365 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6366 than @code{SET_RATIO}.
6369 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6370 A C expression used to determine whether @code{store_by_pieces} will be
6371 used to set a chunk of memory to a constant string value, or whether some
6372 other mechanism will be used. Used by @code{__builtin_strcpy} when
6373 called with a constant source string.
6374 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6375 than @code{MOVE_RATIO}.
6378 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6379 A C expression used to determine whether a load postincrement is a good
6380 thing to use for a given mode. Defaults to the value of
6381 @code{HAVE_POST_INCREMENT}.
6384 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6385 A C expression used to determine whether a load postdecrement is a good
6386 thing to use for a given mode. Defaults to the value of
6387 @code{HAVE_POST_DECREMENT}.
6390 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6391 A C expression used to determine whether a load preincrement is a good
6392 thing to use for a given mode. Defaults to the value of
6393 @code{HAVE_PRE_INCREMENT}.
6396 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6397 A C expression used to determine whether a load predecrement is a good
6398 thing to use for a given mode. Defaults to the value of
6399 @code{HAVE_PRE_DECREMENT}.
6402 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6403 A C expression used to determine whether a store postincrement is a good
6404 thing to use for a given mode. Defaults to the value of
6405 @code{HAVE_POST_INCREMENT}.
6408 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6409 A C expression used to determine whether a store postdecrement is a good
6410 thing to use for a given mode. Defaults to the value of
6411 @code{HAVE_POST_DECREMENT}.
6414 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6415 This macro is used to determine whether a store preincrement is a good
6416 thing to use for a given mode. Defaults to the value of
6417 @code{HAVE_PRE_INCREMENT}.
6420 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6421 This macro is used to determine whether a store predecrement is a good
6422 thing to use for a given mode. Defaults to the value of
6423 @code{HAVE_PRE_DECREMENT}.
6426 @defmac NO_FUNCTION_CSE
6427 Define this macro if it is as good or better to call a constant
6428 function address than to call an address kept in a register.
6431 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6432 Define this macro if a non-short-circuit operation produced by
6433 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6434 @code{BRANCH_COST} is greater than or equal to the value 2.
6437 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int @var{opno}, int *@var{total}, bool @var{speed})
6438 This target hook describes the relative costs of RTL expressions.
6440 The cost may depend on the precise form of the expression, which is
6441 available for examination in @var{x}, and the fact that @var{x} appears
6442 as operand @var{opno} of an expression with rtx code @var{outer_code}.
6443 That is, the hook can assume that there is some rtx @var{y} such
6444 that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that
6445 either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or
6446 (b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}.
6448 @var{code} is @var{x}'s expression code---redundant, since it can be
6449 obtained with @code{GET_CODE (@var{x})}.
6451 In implementing this hook, you can use the construct
6452 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6455 On entry to the hook, @code{*@var{total}} contains a default estimate
6456 for the cost of the expression. The hook should modify this value as
6457 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6458 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6459 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6461 When optimizing for code size, i.e.@: when @code{speed} is
6462 false, this target hook should be used to estimate the relative
6463 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6465 The hook returns true when all subexpressions of @var{x} have been
6466 processed, and false when @code{rtx_cost} should recurse.
6469 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, bool @var{speed})
6470 This hook computes the cost of an addressing mode that contains
6471 @var{address}. If not defined, the cost is computed from
6472 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6474 For most CISC machines, the default cost is a good approximation of the
6475 true cost of the addressing mode. However, on RISC machines, all
6476 instructions normally have the same length and execution time. Hence
6477 all addresses will have equal costs.
6479 In cases where more than one form of an address is known, the form with
6480 the lowest cost will be used. If multiple forms have the same, lowest,
6481 cost, the one that is the most complex will be used.
6483 For example, suppose an address that is equal to the sum of a register
6484 and a constant is used twice in the same basic block. When this macro
6485 is not defined, the address will be computed in a register and memory
6486 references will be indirect through that register. On machines where
6487 the cost of the addressing mode containing the sum is no higher than
6488 that of a simple indirect reference, this will produce an additional
6489 instruction and possibly require an additional register. Proper
6490 specification of this macro eliminates this overhead for such machines.
6492 This hook is never called with an invalid address.
6494 On machines where an address involving more than one register is as
6495 cheap as an address computation involving only one register, defining
6496 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6497 be live over a region of code where only one would have been if
6498 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6499 should be considered in the definition of this macro. Equivalent costs
6500 should probably only be given to addresses with different numbers of
6501 registers on machines with lots of registers.
6505 @section Adjusting the Instruction Scheduler
6507 The instruction scheduler may need a fair amount of machine-specific
6508 adjustment in order to produce good code. GCC provides several target
6509 hooks for this purpose. It is usually enough to define just a few of
6510 them: try the first ones in this list first.
6512 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6513 This hook returns the maximum number of instructions that can ever
6514 issue at the same time on the target machine. The default is one.
6515 Although the insn scheduler can define itself the possibility of issue
6516 an insn on the same cycle, the value can serve as an additional
6517 constraint to issue insns on the same simulated processor cycle (see
6518 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6519 This value must be constant over the entire compilation. If you need
6520 it to vary depending on what the instructions are, you must use
6521 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6524 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6525 This hook is executed by the scheduler after it has scheduled an insn
6526 from the ready list. It should return the number of insns which can
6527 still be issued in the current cycle. The default is
6528 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6529 @code{USE}, which normally are not counted against the issue rate.
6530 You should define this hook if some insns take more machine resources
6531 than others, so that fewer insns can follow them in the same cycle.
6532 @var{file} is either a null pointer, or a stdio stream to write any
6533 debug output to. @var{verbose} is the verbose level provided by
6534 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6538 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6539 This function corrects the value of @var{cost} based on the
6540 relationship between @var{insn} and @var{dep_insn} through the
6541 dependence @var{link}. It should return the new value. The default
6542 is to make no adjustment to @var{cost}. This can be used for example
6543 to specify to the scheduler using the traditional pipeline description
6544 that an output- or anti-dependence does not incur the same cost as a
6545 data-dependence. If the scheduler using the automaton based pipeline
6546 description, the cost of anti-dependence is zero and the cost of
6547 output-dependence is maximum of one and the difference of latency
6548 times of the first and the second insns. If these values are not
6549 acceptable, you could use the hook to modify them too. See also
6550 @pxref{Processor pipeline description}.
6553 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6554 This hook adjusts the integer scheduling priority @var{priority} of
6555 @var{insn}. It should return the new priority. Increase the priority to
6556 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6557 later. Do not define this hook if you do not need to adjust the
6558 scheduling priorities of insns.
6561 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6562 This hook is executed by the scheduler after it has scheduled the ready
6563 list, to allow the machine description to reorder it (for example to
6564 combine two small instructions together on @samp{VLIW} machines).
6565 @var{file} is either a null pointer, or a stdio stream to write any
6566 debug output to. @var{verbose} is the verbose level provided by
6567 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6568 list of instructions that are ready to be scheduled. @var{n_readyp} is
6569 a pointer to the number of elements in the ready list. The scheduler
6570 reads the ready list in reverse order, starting with
6571 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6572 is the timer tick of the scheduler. You may modify the ready list and
6573 the number of ready insns. The return value is the number of insns that
6574 can issue this cycle; normally this is just @code{issue_rate}. See also
6575 @samp{TARGET_SCHED_REORDER2}.
6578 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6579 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6580 function is called whenever the scheduler starts a new cycle. This one
6581 is called once per iteration over a cycle, immediately after
6582 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6583 return the number of insns to be scheduled in the same cycle. Defining
6584 this hook can be useful if there are frequent situations where
6585 scheduling one insn causes other insns to become ready in the same
6586 cycle. These other insns can then be taken into account properly.
6589 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6590 This hook is called after evaluation forward dependencies of insns in
6591 chain given by two parameter values (@var{head} and @var{tail}
6592 correspondingly) but before insns scheduling of the insn chain. For
6593 example, it can be used for better insn classification if it requires
6594 analysis of dependencies. This hook can use backward and forward
6595 dependencies of the insn scheduler because they are already
6599 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6600 This hook is executed by the scheduler at the beginning of each block of
6601 instructions that are to be scheduled. @var{file} is either a null
6602 pointer, or a stdio stream to write any debug output to. @var{verbose}
6603 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6604 @var{max_ready} is the maximum number of insns in the current scheduling
6605 region that can be live at the same time. This can be used to allocate
6606 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6609 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6610 This hook is executed by the scheduler at the end of each block of
6611 instructions that are to be scheduled. It can be used to perform
6612 cleanup of any actions done by the other scheduling hooks. @var{file}
6613 is either a null pointer, or a stdio stream to write any debug output
6614 to. @var{verbose} is the verbose level provided by
6615 @option{-fsched-verbose-@var{n}}.
6618 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6619 This hook is executed by the scheduler after function level initializations.
6620 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6621 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6622 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6625 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6626 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6627 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6628 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6631 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6632 The hook returns an RTL insn. The automaton state used in the
6633 pipeline hazard recognizer is changed as if the insn were scheduled
6634 when the new simulated processor cycle starts. Usage of the hook may
6635 simplify the automaton pipeline description for some @acronym{VLIW}
6636 processors. If the hook is defined, it is used only for the automaton
6637 based pipeline description. The default is not to change the state
6638 when the new simulated processor cycle starts.
6641 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6642 The hook can be used to initialize data used by the previous hook.
6645 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6646 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6647 to changed the state as if the insn were scheduled when the new
6648 simulated processor cycle finishes.
6651 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6652 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6653 used to initialize data used by the previous hook.
6656 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
6657 The hook to notify target that the current simulated cycle is about to finish.
6658 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6659 to change the state in more complicated situations - e.g., when advancing
6660 state on a single insn is not enough.
6663 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
6664 The hook to notify target that new simulated cycle has just started.
6665 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6666 to change the state in more complicated situations - e.g., when advancing
6667 state on a single insn is not enough.
6670 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6671 This hook controls better choosing an insn from the ready insn queue
6672 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6673 chooses the first insn from the queue. If the hook returns a positive
6674 value, an additional scheduler code tries all permutations of
6675 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6676 subsequent ready insns to choose an insn whose issue will result in
6677 maximal number of issued insns on the same cycle. For the
6678 @acronym{VLIW} processor, the code could actually solve the problem of
6679 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6680 rules of @acronym{VLIW} packing are described in the automaton.
6682 This code also could be used for superscalar @acronym{RISC}
6683 processors. Let us consider a superscalar @acronym{RISC} processor
6684 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6685 @var{B}, some insns can be executed only in pipelines @var{B} or
6686 @var{C}, and one insn can be executed in pipeline @var{B}. The
6687 processor may issue the 1st insn into @var{A} and the 2nd one into
6688 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6689 until the next cycle. If the scheduler issues the 3rd insn the first,
6690 the processor could issue all 3 insns per cycle.
6692 Actually this code demonstrates advantages of the automaton based
6693 pipeline hazard recognizer. We try quickly and easy many insn
6694 schedules to choose the best one.
6696 The default is no multipass scheduling.
6699 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx @var{insn})
6701 This hook controls what insns from the ready insn queue will be
6702 considered for the multipass insn scheduling. If the hook returns
6703 zero for @var{insn}, the insn will be not chosen to
6706 The default is that any ready insns can be chosen to be issued.
6709 @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})
6710 This hook prepares the target backend for a new round of multipass
6714 @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})
6715 This hook is called when multipass scheduling evaluates instruction INSN.
6718 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, char *@var{ready_try}, int @var{n_ready})
6719 This is called when multipass scheduling backtracks from evaluation of
6723 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
6724 This hook notifies the target about the result of the concluded current
6725 round of multipass scheduling.
6728 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
6729 This hook initializes target-specific data used in multipass scheduling.
6732 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
6733 This hook finalizes target-specific data used in multipass scheduling.
6736 @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})
6737 This hook is called by the insn scheduler before issuing @var{insn}
6738 on cycle @var{clock}. If the hook returns nonzero,
6739 @var{insn} is not issued on this processor cycle. Instead,
6740 the processor cycle is advanced. If *@var{sort_p}
6741 is zero, the insn ready queue is not sorted on the new cycle
6742 start as usually. @var{dump} and @var{verbose} specify the file and
6743 verbosity level to use for debugging output.
6744 @var{last_clock} and @var{clock} are, respectively, the
6745 processor cycle on which the previous insn has been issued,
6746 and the current processor cycle.
6749 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
6750 This hook is used to define which dependences are considered costly by
6751 the target, so costly that it is not advisable to schedule the insns that
6752 are involved in the dependence too close to one another. The parameters
6753 to this hook are as follows: The first parameter @var{_dep} is the dependence
6754 being evaluated. The second parameter @var{cost} is the cost of the
6755 dependence as estimated by the scheduler, and the third
6756 parameter @var{distance} is the distance in cycles between the two insns.
6757 The hook returns @code{true} if considering the distance between the two
6758 insns the dependence between them is considered costly by the target,
6759 and @code{false} otherwise.
6761 Defining this hook can be useful in multiple-issue out-of-order machines,
6762 where (a) it's practically hopeless to predict the actual data/resource
6763 delays, however: (b) there's a better chance to predict the actual grouping
6764 that will be formed, and (c) correctly emulating the grouping can be very
6765 important. In such targets one may want to allow issuing dependent insns
6766 closer to one another---i.e., closer than the dependence distance; however,
6767 not in cases of ``costly dependences'', which this hooks allows to define.
6770 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6771 This hook is called by the insn scheduler after emitting a new instruction to
6772 the instruction stream. The hook notifies a target backend to extend its
6773 per instruction data structures.
6776 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6777 Return a pointer to a store large enough to hold target scheduling context.
6780 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6781 Initialize store pointed to by @var{tc} to hold target scheduling context.
6782 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6783 beginning of the block. Otherwise, copy the current context into @var{tc}.
6786 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6787 Copy target scheduling context pointed to by @var{tc} to the current context.
6790 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6791 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6794 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6795 Deallocate a store for target scheduling context pointed to by @var{tc}.
6798 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6799 This hook is called by the insn scheduler when @var{insn} has only
6800 speculative dependencies and therefore can be scheduled speculatively.
6801 The hook is used to check if the pattern of @var{insn} has a speculative
6802 version and, in case of successful check, to generate that speculative
6803 pattern. The hook should return 1, if the instruction has a speculative form,
6804 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6805 speculation. If the return value equals 1 then @var{new_pat} is assigned
6806 the generated speculative pattern.
6809 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (int @var{dep_status})
6810 This hook is called by the insn scheduler during generation of recovery code
6811 for @var{insn}. It should return @code{true}, if the corresponding check
6812 instruction should branch to recovery code, or @code{false} otherwise.
6815 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6816 This hook is called by the insn scheduler to generate a pattern for recovery
6817 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6818 speculative instruction for which the check should be generated.
6819 @var{label} is either a label of a basic block, where recovery code should
6820 be emitted, or a null pointer, when requested check doesn't branch to
6821 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6822 a pattern for a branchy check corresponding to a simple check denoted by
6823 @var{insn} should be generated. In this case @var{label} can't be null.
6826 @deftypefn {Target Hook} bool TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (const_rtx @var{insn})
6827 This hook is used as a workaround for
6828 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6829 called on the first instruction of the ready list. The hook is used to
6830 discard speculative instructions that stand first in the ready list from
6831 being scheduled on the current cycle. If the hook returns @code{false},
6832 @var{insn} will not be chosen to be issued.
6833 For non-speculative instructions,
6834 the hook should always return @code{true}. For example, in the ia64 backend
6835 the hook is used to cancel data speculative insns when the ALAT table
6839 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
6840 This hook is used by the insn scheduler to find out what features should be
6842 The structure *@var{spec_info} should be filled in by the target.
6843 The structure describes speculation types that can be used in the scheduler.
6846 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6847 This hook is called by the swing modulo scheduler to calculate a
6848 resource-based lower bound which is based on the resources available in
6849 the machine and the resources required by each instruction. The target
6850 backend can use @var{g} to calculate such bound. A very simple lower
6851 bound will be used in case this hook is not implemented: the total number
6852 of instructions divided by the issue rate.
6855 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx @var{insn}, int @var{x})
6856 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6857 is supported in hardware and the condition specified in the parameter is true.
6860 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx @var{insn}, int @var{x})
6861 This hook is called by Haifa Scheduler. It performs the operation specified
6862 in its second parameter.
6865 @deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE
6866 True if the processor has an exposed pipeline, which means that not just
6867 the order of instructions is important for correctness when scheduling, but
6868 also the latencies of operations.
6871 @deftypefn {Target Hook} int TARGET_SCHED_REASSOCIATION_WIDTH (unsigned int @var{opc}, enum machine_mode @var{mode})
6872 This hook is called by tree reassociator to determine a level of
6873 parallelism required in output calculations chain.
6877 @section Dividing the Output into Sections (Texts, Data, @dots{})
6878 @c the above section title is WAY too long. maybe cut the part between
6879 @c the (...)? --mew 10feb93
6881 An object file is divided into sections containing different types of
6882 data. In the most common case, there are three sections: the @dfn{text
6883 section}, which holds instructions and read-only data; the @dfn{data
6884 section}, which holds initialized writable data; and the @dfn{bss
6885 section}, which holds uninitialized data. Some systems have other kinds
6888 @file{varasm.c} provides several well-known sections, such as
6889 @code{text_section}, @code{data_section} and @code{bss_section}.
6890 The normal way of controlling a @code{@var{foo}_section} variable
6891 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6892 as described below. The macros are only read once, when @file{varasm.c}
6893 initializes itself, so their values must be run-time constants.
6894 They may however depend on command-line flags.
6896 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6897 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6898 to be string literals.
6900 Some assemblers require a different string to be written every time a
6901 section is selected. If your assembler falls into this category, you
6902 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6903 @code{get_unnamed_section} to set up the sections.
6905 You must always create a @code{text_section}, either by defining
6906 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6907 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6908 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6909 create a distinct @code{readonly_data_section}, the default is to
6910 reuse @code{text_section}.
6912 All the other @file{varasm.c} sections are optional, and are null
6913 if the target does not provide them.
6915 @defmac TEXT_SECTION_ASM_OP
6916 A C expression whose value is a string, including spacing, containing the
6917 assembler operation that should precede instructions and read-only data.
6918 Normally @code{"\t.text"} is right.
6921 @defmac HOT_TEXT_SECTION_NAME
6922 If defined, a C string constant for the name of the section containing most
6923 frequently executed functions of the program. If not defined, GCC will provide
6924 a default definition if the target supports named sections.
6927 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6928 If defined, a C string constant for the name of the section containing unlikely
6929 executed functions in the program.
6932 @defmac DATA_SECTION_ASM_OP
6933 A C expression whose value is a string, including spacing, containing the
6934 assembler operation to identify the following data as writable initialized
6935 data. Normally @code{"\t.data"} is right.
6938 @defmac SDATA_SECTION_ASM_OP
6939 If defined, a C expression whose value is a string, including spacing,
6940 containing the assembler operation to identify the following data as
6941 initialized, writable small data.
6944 @defmac READONLY_DATA_SECTION_ASM_OP
6945 A C expression whose value is a string, including spacing, containing the
6946 assembler operation to identify the following data as read-only initialized
6950 @defmac BSS_SECTION_ASM_OP
6951 If defined, a C expression whose value is a string, including spacing,
6952 containing the assembler operation to identify the following data as
6953 uninitialized global data. If not defined, and
6954 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
6955 uninitialized global data will be output in the data section if
6956 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6960 @defmac SBSS_SECTION_ASM_OP
6961 If defined, a C expression whose value is a string, including spacing,
6962 containing the assembler operation to identify the following data as
6963 uninitialized, writable small data.
6966 @defmac TLS_COMMON_ASM_OP
6967 If defined, a C expression whose value is a string containing the
6968 assembler operation to identify the following data as thread-local
6969 common data. The default is @code{".tls_common"}.
6972 @defmac TLS_SECTION_ASM_FLAG
6973 If defined, a C expression whose value is a character constant
6974 containing the flag used to mark a section as a TLS section. The
6975 default is @code{'T'}.
6978 @defmac INIT_SECTION_ASM_OP
6979 If defined, a C expression whose value is a string, including spacing,
6980 containing the assembler operation to identify the following data as
6981 initialization code. If not defined, GCC will assume such a section does
6982 not exist. This section has no corresponding @code{init_section}
6983 variable; it is used entirely in runtime code.
6986 @defmac FINI_SECTION_ASM_OP
6987 If defined, a C expression whose value is a string, including spacing,
6988 containing the assembler operation to identify the following data as
6989 finalization code. If not defined, GCC will assume such a section does
6990 not exist. This section has no corresponding @code{fini_section}
6991 variable; it is used entirely in runtime code.
6994 @defmac INIT_ARRAY_SECTION_ASM_OP
6995 If defined, a C expression whose value is a string, including spacing,
6996 containing the assembler operation to identify the following data as
6997 part of the @code{.init_array} (or equivalent) section. If not
6998 defined, GCC will assume such a section does not exist. Do not define
6999 both this macro and @code{INIT_SECTION_ASM_OP}.
7002 @defmac FINI_ARRAY_SECTION_ASM_OP
7003 If defined, a C expression whose value is a string, including spacing,
7004 containing the assembler operation to identify the following data as
7005 part of the @code{.fini_array} (or equivalent) section. If not
7006 defined, GCC will assume such a section does not exist. Do not define
7007 both this macro and @code{FINI_SECTION_ASM_OP}.
7010 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
7011 If defined, an ASM statement that switches to a different section
7012 via @var{section_op}, calls @var{function}, and switches back to
7013 the text section. This is used in @file{crtstuff.c} if
7014 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
7015 to initialization and finalization functions from the init and fini
7016 sections. By default, this macro uses a simple function call. Some
7017 ports need hand-crafted assembly code to avoid dependencies on
7018 registers initialized in the function prologue or to ensure that
7019 constant pools don't end up too far way in the text section.
7022 @defmac TARGET_LIBGCC_SDATA_SECTION
7023 If defined, a string which names the section into which small
7024 variables defined in crtstuff and libgcc should go. This is useful
7025 when the target has options for optimizing access to small data, and
7026 you want the crtstuff and libgcc routines to be conservative in what
7027 they expect of your application yet liberal in what your application
7028 expects. For example, for targets with a @code{.sdata} section (like
7029 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
7030 require small data support from your application, but use this macro
7031 to put small data into @code{.sdata} so that your application can
7032 access these variables whether it uses small data or not.
7035 @defmac FORCE_CODE_SECTION_ALIGN
7036 If defined, an ASM statement that aligns a code section to some
7037 arbitrary boundary. This is used to force all fragments of the
7038 @code{.init} and @code{.fini} sections to have to same alignment
7039 and thus prevent the linker from having to add any padding.
7042 @defmac JUMP_TABLES_IN_TEXT_SECTION
7043 Define this macro to be an expression with a nonzero value if jump
7044 tables (for @code{tablejump} insns) should be output in the text
7045 section, along with the assembler instructions. Otherwise, the
7046 readonly data section is used.
7048 This macro is irrelevant if there is no separate readonly data section.
7051 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
7052 Define this hook if you need to do something special to set up the
7053 @file{varasm.c} sections, or if your target has some special sections
7054 of its own that you need to create.
7056 GCC calls this hook after processing the command line, but before writing
7057 any assembly code, and before calling any of the section-returning hooks
7061 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
7062 Return a mask describing how relocations should be treated when
7063 selecting sections. Bit 1 should be set if global relocations
7064 should be placed in a read-write section; bit 0 should be set if
7065 local relocations should be placed in a read-write section.
7067 The default version of this function returns 3 when @option{-fpic}
7068 is in effect, and 0 otherwise. The hook is typically redefined
7069 when the target cannot support (some kinds of) dynamic relocations
7070 in read-only sections even in executables.
7073 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
7074 Return the section into which @var{exp} should be placed. You can
7075 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
7076 some sort. @var{reloc} indicates whether the initial value of @var{exp}
7077 requires link-time relocations. Bit 0 is set when variable contains
7078 local relocations only, while bit 1 is set for global relocations.
7079 @var{align} is the constant alignment in bits.
7081 The default version of this function takes care of putting read-only
7082 variables in @code{readonly_data_section}.
7084 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7087 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7088 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7089 for @code{FUNCTION_DECL}s as well as for variables and constants.
7091 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7092 function has been determined to be likely to be called, and nonzero if
7093 it is unlikely to be called.
7096 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7097 Build up a unique section name, expressed as a @code{STRING_CST} node,
7098 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7099 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7100 the initial value of @var{exp} requires link-time relocations.
7102 The default version of this function appends the symbol name to the
7103 ELF section name that would normally be used for the symbol. For
7104 example, the function @code{foo} would be placed in @code{.text.foo}.
7105 Whatever the actual target object format, this is often good enough.
7108 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7109 Return the readonly data section associated with
7110 @samp{DECL_SECTION_NAME (@var{decl})}.
7111 The default version of this function selects @code{.gnu.linkonce.r.name} if
7112 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7113 if function is in @code{.text.name}, and the normal readonly-data section
7117 @deftypevr {Target Hook} {const char *} TARGET_ASM_MERGEABLE_RODATA_PREFIX
7118 Usually, the compiler uses the prefix @code{".rodata"} to construct
7119 section names for mergeable constant data. Define this macro to override
7120 the string if a different section name should be used.
7123 @deftypefn {Target Hook} {section *} TARGET_ASM_TM_CLONE_TABLE_SECTION (void)
7124 Return the section that should be used for transactional memory clone tables.
7127 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7128 Return the section into which a constant @var{x}, of mode @var{mode},
7129 should be placed. You can assume that @var{x} is some kind of
7130 constant in RTL@. The argument @var{mode} is redundant except in the
7131 case of a @code{const_int} rtx. @var{align} is the constant alignment
7134 The default version of this function takes care of putting symbolic
7135 constants in @code{flag_pic} mode in @code{data_section} and everything
7136 else in @code{readonly_data_section}.
7139 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7140 Define this hook if you need to postprocess the assembler name generated
7141 by target-independent code. The @var{id} provided to this hook will be
7142 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7143 or the mangled name of the @var{decl} in C++). The return value of the
7144 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7145 your target system. The default implementation of this hook just
7146 returns the @var{id} provided.
7149 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7150 Define this hook if references to a symbol or a constant must be
7151 treated differently depending on something about the variable or
7152 function named by the symbol (such as what section it is in).
7154 The hook is executed immediately after rtl has been created for
7155 @var{decl}, which may be a variable or function declaration or
7156 an entry in the constant pool. In either case, @var{rtl} is the
7157 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7158 in this hook; that field may not have been initialized yet.
7160 In the case of a constant, it is safe to assume that the rtl is
7161 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7162 will also have this form, but that is not guaranteed. Global
7163 register variables, for instance, will have a @code{reg} for their
7164 rtl. (Normally the right thing to do with such unusual rtl is
7167 The @var{new_decl_p} argument will be true if this is the first time
7168 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7169 be false for subsequent invocations, which will happen for duplicate
7170 declarations. Whether or not anything must be done for the duplicate
7171 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7172 @var{new_decl_p} is always true when the hook is called for a constant.
7174 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7175 The usual thing for this hook to do is to record flags in the
7176 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7177 Historically, the name string was modified if it was necessary to
7178 encode more than one bit of information, but this practice is now
7179 discouraged; use @code{SYMBOL_REF_FLAGS}.
7181 The default definition of this hook, @code{default_encode_section_info}
7182 in @file{varasm.c}, sets a number of commonly-useful bits in
7183 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7184 before overriding it.
7187 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7188 Decode @var{name} and return the real name part, sans
7189 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7193 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7194 Returns true if @var{exp} should be placed into a ``small data'' section.
7195 The default version of this hook always returns false.
7198 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7199 Contains the value true if the target places read-only
7200 ``small data'' into a separate section. The default value is false.
7203 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7204 It returns true if target wants profile code emitted before prologue.
7206 The default version of this hook use the target macro
7207 @code{PROFILE_BEFORE_PROLOGUE}.
7210 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7211 Returns true if @var{exp} names an object for which name resolution
7212 rules must resolve to the current ``module'' (dynamic shared library
7213 or executable image).
7215 The default version of this hook implements the name resolution rules
7216 for ELF, which has a looser model of global name binding than other
7217 currently supported object file formats.
7220 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7221 Contains the value true if the target supports thread-local storage.
7222 The default value is false.
7227 @section Position Independent Code
7228 @cindex position independent code
7231 This section describes macros that help implement generation of position
7232 independent code. Simply defining these macros is not enough to
7233 generate valid PIC; you must also add support to the hook
7234 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7235 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7236 must modify the definition of @samp{movsi} to do something appropriate
7237 when the source operand contains a symbolic address. You may also
7238 need to alter the handling of switch statements so that they use
7240 @c i rearranged the order of the macros above to try to force one of
7241 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7243 @defmac PIC_OFFSET_TABLE_REGNUM
7244 The register number of the register used to address a table of static
7245 data addresses in memory. In some cases this register is defined by a
7246 processor's ``application binary interface'' (ABI)@. When this macro
7247 is defined, RTL is generated for this register once, as with the stack
7248 pointer and frame pointer registers. If this macro is not defined, it
7249 is up to the machine-dependent files to allocate such a register (if
7250 necessary). Note that this register must be fixed when in use (e.g.@:
7251 when @code{flag_pic} is true).
7254 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7255 A C expression that is nonzero if the register defined by
7256 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7257 the default is zero. Do not define
7258 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7261 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7262 A C expression that is nonzero if @var{x} is a legitimate immediate
7263 operand on the target machine when generating position independent code.
7264 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7265 check this. You can also assume @var{flag_pic} is true, so you need not
7266 check it either. You need not define this macro if all constants
7267 (including @code{SYMBOL_REF}) can be immediate operands when generating
7268 position independent code.
7271 @node Assembler Format
7272 @section Defining the Output Assembler Language
7274 This section describes macros whose principal purpose is to describe how
7275 to write instructions in assembler language---rather than what the
7279 * File Framework:: Structural information for the assembler file.
7280 * Data Output:: Output of constants (numbers, strings, addresses).
7281 * Uninitialized Data:: Output of uninitialized variables.
7282 * Label Output:: Output and generation of labels.
7283 * Initialization:: General principles of initialization
7284 and termination routines.
7285 * Macros for Initialization::
7286 Specific macros that control the handling of
7287 initialization and termination routines.
7288 * Instruction Output:: Output of actual instructions.
7289 * Dispatch Tables:: Output of jump tables.
7290 * Exception Region Output:: Output of exception region code.
7291 * Alignment Output:: Pseudo ops for alignment and skipping data.
7294 @node File Framework
7295 @subsection The Overall Framework of an Assembler File
7296 @cindex assembler format
7297 @cindex output of assembler code
7299 @c prevent bad page break with this line
7300 This describes the overall framework of an assembly file.
7302 @findex default_file_start
7303 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7304 Output to @code{asm_out_file} any text which the assembler expects to
7305 find at the beginning of a file. The default behavior is controlled
7306 by two flags, documented below. Unless your target's assembler is
7307 quite unusual, if you override the default, you should call
7308 @code{default_file_start} at some point in your target hook. This
7309 lets other target files rely on these variables.
7312 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7313 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7314 printed as the very first line in the assembly file, unless
7315 @option{-fverbose-asm} is in effect. (If that macro has been defined
7316 to the empty string, this variable has no effect.) With the normal
7317 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7318 assembler that it need not bother stripping comments or extra
7319 whitespace from its input. This allows it to work a bit faster.
7321 The default is false. You should not set it to true unless you have
7322 verified that your port does not generate any extra whitespace or
7323 comments that will cause GAS to issue errors in NO_APP mode.
7326 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7327 If this flag is true, @code{output_file_directive} will be called
7328 for the primary source file, immediately after printing
7329 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7330 this to be done. The default is false.
7333 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7334 Output to @code{asm_out_file} any text which the assembler expects
7335 to find at the end of a file. The default is to output nothing.
7338 @deftypefun void file_end_indicate_exec_stack ()
7339 Some systems use a common convention, the @samp{.note.GNU-stack}
7340 special section, to indicate whether or not an object file relies on
7341 the stack being executable. If your system uses this convention, you
7342 should define @code{TARGET_ASM_FILE_END} to this function. If you
7343 need to do other things in that hook, have your hook function call
7347 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7348 Output to @code{asm_out_file} any text which the assembler expects
7349 to find at the start of an LTO section. The default is to output
7353 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7354 Output to @code{asm_out_file} any text which the assembler expects
7355 to find at the end of an LTO section. The default is to output
7359 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7360 Output to @code{asm_out_file} any text which is needed before emitting
7361 unwind info and debug info at the end of a file. Some targets emit
7362 here PIC setup thunks that cannot be emitted at the end of file,
7363 because they couldn't have unwind info then. The default is to output
7367 @defmac ASM_COMMENT_START
7368 A C string constant describing how to begin a comment in the target
7369 assembler language. The compiler assumes that the comment will end at
7370 the end of the line.
7374 A C string constant for text to be output before each @code{asm}
7375 statement or group of consecutive ones. Normally this is
7376 @code{"#APP"}, which is a comment that has no effect on most
7377 assemblers but tells the GNU assembler that it must check the lines
7378 that follow for all valid assembler constructs.
7382 A C string constant for text to be output after each @code{asm}
7383 statement or group of consecutive ones. Normally this is
7384 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7385 time-saving assumptions that are valid for ordinary compiler output.
7388 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7389 A C statement to output COFF information or DWARF debugging information
7390 which indicates that filename @var{name} is the current source file to
7391 the stdio stream @var{stream}.
7393 This macro need not be defined if the standard form of output
7394 for the file format in use is appropriate.
7397 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7398 Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7400 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7403 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7404 A C statement to output the string @var{string} to the stdio stream
7405 @var{stream}. If you do not call the function @code{output_quoted_string}
7406 in your config files, GCC will only call it to output filenames to
7407 the assembler source. So you can use it to canonicalize the format
7408 of the filename using this macro.
7411 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7412 A C statement to output something to the assembler file to handle a
7413 @samp{#ident} directive containing the text @var{string}. If this
7414 macro is not defined, nothing is output for a @samp{#ident} directive.
7417 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7418 Output assembly directives to switch to section @var{name}. The section
7419 should have attributes as specified by @var{flags}, which is a bit mask
7420 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7421 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7422 this section is associated.
7425 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
7426 Return preferred text (sub)section for function @var{decl}.
7427 Main purpose of this function is to separate cold, normal and hot
7428 functions. @var{startup} is true when function is known to be used only
7429 at startup (from static constructors or it is @code{main()}).
7430 @var{exit} is true when function is known to be used only at exit
7431 (from static destructors).
7432 Return NULL if function should go to default text section.
7435 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
7436 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}.
7439 @deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7440 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7441 It must not be modified by command-line option processing.
7444 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7445 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7446 This flag is true if we can create zeroed data by switching to a BSS
7447 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7448 This is true on most ELF targets.
7451 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7452 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7453 based on a variable or function decl, a section name, and whether or not the
7454 declaration's initializer may contain runtime relocations. @var{decl} may be
7455 null, in which case read-write data should be assumed.
7457 The default version of this function handles choosing code vs data,
7458 read-only vs read-write data, and @code{flag_pic}. You should only
7459 need to override this if your target has special flags that might be
7460 set via @code{__attribute__}.
7463 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
7464 Provides the target with the ability to record the gcc command line
7465 switches that have been passed to the compiler, and options that are
7466 enabled. The @var{type} argument specifies what is being recorded.
7467 It can take the following values:
7470 @item SWITCH_TYPE_PASSED
7471 @var{text} is a command line switch that has been set by the user.
7473 @item SWITCH_TYPE_ENABLED
7474 @var{text} is an option which has been enabled. This might be as a
7475 direct result of a command line switch, or because it is enabled by
7476 default or because it has been enabled as a side effect of a different
7477 command line switch. For example, the @option{-O2} switch enables
7478 various different individual optimization passes.
7480 @item SWITCH_TYPE_DESCRIPTIVE
7481 @var{text} is either NULL or some descriptive text which should be
7482 ignored. If @var{text} is NULL then it is being used to warn the
7483 target hook that either recording is starting or ending. The first
7484 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7485 warning is for start up and the second time the warning is for
7486 wind down. This feature is to allow the target hook to make any
7487 necessary preparations before it starts to record switches and to
7488 perform any necessary tidying up after it has finished recording
7491 @item SWITCH_TYPE_LINE_START
7492 This option can be ignored by this target hook.
7494 @item SWITCH_TYPE_LINE_END
7495 This option can be ignored by this target hook.
7498 The hook's return value must be zero. Other return values may be
7499 supported in the future.
7501 By default this hook is set to NULL, but an example implementation is
7502 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7503 it records the switches as ASCII text inside a new, string mergeable
7504 section in the assembler output file. The name of the new section is
7505 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7509 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7510 This is the name of the section that will be created by the example
7511 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7517 @subsection Output of Data
7520 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7521 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7522 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7523 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7524 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7525 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7526 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7527 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7528 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7529 These hooks specify assembly directives for creating certain kinds
7530 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7531 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7532 aligned two-byte object, and so on. Any of the hooks may be
7533 @code{NULL}, indicating that no suitable directive is available.
7535 The compiler will print these strings at the start of a new line,
7536 followed immediately by the object's initial value. In most cases,
7537 the string should contain a tab, a pseudo-op, and then another tab.
7540 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7541 The @code{assemble_integer} function uses this hook to output an
7542 integer object. @var{x} is the object's value, @var{size} is its size
7543 in bytes and @var{aligned_p} indicates whether it is aligned. The
7544 function should return @code{true} if it was able to output the
7545 object. If it returns false, @code{assemble_integer} will try to
7546 split the object into smaller parts.
7548 The default implementation of this hook will use the
7549 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7550 when the relevant string is @code{NULL}.
7553 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
7554 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7555 can't deal with, and output assembly code to @var{file} corresponding to
7556 the pattern @var{x}. This may be used to allow machine-dependent
7557 @code{UNSPEC}s to appear within constants.
7559 If target hook fails to recognize a pattern, it must return @code{false},
7560 so that a standard error message is printed. If it prints an error message
7561 itself, by calling, for example, @code{output_operand_lossage}, it may just
7565 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7566 A C statement to output to the stdio stream @var{stream} an assembler
7567 instruction to assemble a string constant containing the @var{len}
7568 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7569 @code{char *} and @var{len} a C expression of type @code{int}.
7571 If the assembler has a @code{.ascii} pseudo-op as found in the
7572 Berkeley Unix assembler, do not define the macro
7573 @code{ASM_OUTPUT_ASCII}.
7576 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7577 A C statement to output word @var{n} of a function descriptor for
7578 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7579 is defined, and is otherwise unused.
7582 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7583 You may define this macro as a C expression. You should define the
7584 expression to have a nonzero value if GCC should output the constant
7585 pool for a function before the code for the function, or a zero value if
7586 GCC should output the constant pool after the function. If you do
7587 not define this macro, the usual case, GCC will output the constant
7588 pool before the function.
7591 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7592 A C statement to output assembler commands to define the start of the
7593 constant pool for a function. @var{funname} is a string giving
7594 the name of the function. Should the return type of the function
7595 be required, it can be obtained via @var{fundecl}. @var{size}
7596 is the size, in bytes, of the constant pool that will be written
7597 immediately after this call.
7599 If no constant-pool prefix is required, the usual case, this macro need
7603 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7604 A C statement (with or without semicolon) to output a constant in the
7605 constant pool, if it needs special treatment. (This macro need not do
7606 anything for RTL expressions that can be output normally.)
7608 The argument @var{file} is the standard I/O stream to output the
7609 assembler code on. @var{x} is the RTL expression for the constant to
7610 output, and @var{mode} is the machine mode (in case @var{x} is a
7611 @samp{const_int}). @var{align} is the required alignment for the value
7612 @var{x}; you should output an assembler directive to force this much
7615 The argument @var{labelno} is a number to use in an internal label for
7616 the address of this pool entry. The definition of this macro is
7617 responsible for outputting the label definition at the proper place.
7618 Here is how to do this:
7621 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7624 When you output a pool entry specially, you should end with a
7625 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7626 entry from being output a second time in the usual manner.
7628 You need not define this macro if it would do nothing.
7631 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7632 A C statement to output assembler commands to at the end of the constant
7633 pool for a function. @var{funname} is a string giving the name of the
7634 function. Should the return type of the function be required, you can
7635 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7636 constant pool that GCC wrote immediately before this call.
7638 If no constant-pool epilogue is required, the usual case, you need not
7642 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7643 Define this macro as a C expression which is nonzero if @var{C} is
7644 used as a logical line separator by the assembler. @var{STR} points
7645 to the position in the string where @var{C} was found; this can be used if
7646 a line separator uses multiple characters.
7648 If you do not define this macro, the default is that only
7649 the character @samp{;} is treated as a logical line separator.
7652 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7653 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7654 These target hooks are C string constants, describing the syntax in the
7655 assembler for grouping arithmetic expressions. If not overridden, they
7656 default to normal parentheses, which is correct for most assemblers.
7659 These macros are provided by @file{real.h} for writing the definitions
7660 of @code{ASM_OUTPUT_DOUBLE} and the like:
7662 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7663 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7664 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7665 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7666 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7667 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7668 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7669 target's floating point representation, and store its bit pattern in
7670 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7671 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7672 simple @code{long int}. For the others, it should be an array of
7673 @code{long int}. The number of elements in this array is determined
7674 by the size of the desired target floating point data type: 32 bits of
7675 it go in each @code{long int} array element. Each array element holds
7676 32 bits of the result, even if @code{long int} is wider than 32 bits
7677 on the host machine.
7679 The array element values are designed so that you can print them out
7680 using @code{fprintf} in the order they should appear in the target
7684 @node Uninitialized Data
7685 @subsection Output of Uninitialized Variables
7687 Each of the macros in this section is used to do the whole job of
7688 outputting a single uninitialized variable.
7690 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7691 A C statement (sans semicolon) to output to the stdio stream
7692 @var{stream} the assembler definition of a common-label named
7693 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7694 is the size rounded up to whatever alignment the caller wants. It is
7695 possible that @var{size} may be zero, for instance if a struct with no
7696 other member than a zero-length array is defined. In this case, the
7697 backend must output a symbol definition that allocates at least one
7698 byte, both so that the address of the resulting object does not compare
7699 equal to any other, and because some object formats cannot even express
7700 the concept of a zero-sized common symbol, as that is how they represent
7701 an ordinary undefined external.
7703 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7704 output the name itself; before and after that, output the additional
7705 assembler syntax for defining the name, and a newline.
7707 This macro controls how the assembler definitions of uninitialized
7708 common global variables are output.
7711 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7712 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7713 separate, explicit argument. If you define this macro, it is used in
7714 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7715 handling the required alignment of the variable. The alignment is specified
7716 as the number of bits.
7719 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7720 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7721 variable to be output, if there is one, or @code{NULL_TREE} if there
7722 is no corresponding variable. If you define this macro, GCC will use it
7723 in place of both @code{ASM_OUTPUT_COMMON} and
7724 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7725 the variable's decl in order to chose what to output.
7728 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7729 A C statement (sans semicolon) to output to the stdio stream
7730 @var{stream} the assembler definition of uninitialized global @var{decl} named
7731 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
7732 is the alignment specified as the number of bits.
7734 Try to use function @code{asm_output_aligned_bss} defined in file
7735 @file{varasm.c} when defining this macro. If unable, use the expression
7736 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7737 before and after that, output the additional assembler syntax for defining
7738 the name, and a newline.
7740 There are two ways of handling global BSS@. One is to define this macro.
7741 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7742 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7743 You do not need to do both.
7745 Some languages do not have @code{common} data, and require a
7746 non-common form of global BSS in order to handle uninitialized globals
7747 efficiently. C++ is one example of this. However, if the target does
7748 not support global BSS, the front end may choose to make globals
7749 common in order to save space in the object file.
7752 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7753 A C statement (sans semicolon) to output to the stdio stream
7754 @var{stream} the assembler definition of a local-common-label named
7755 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7756 is the size rounded up to whatever alignment the caller wants.
7758 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7759 output the name itself; before and after that, output the additional
7760 assembler syntax for defining the name, and a newline.
7762 This macro controls how the assembler definitions of uninitialized
7763 static variables are output.
7766 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7767 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7768 separate, explicit argument. If you define this macro, it is used in
7769 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7770 handling the required alignment of the variable. The alignment is specified
7771 as the number of bits.
7774 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7775 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7776 variable to be output, if there is one, or @code{NULL_TREE} if there
7777 is no corresponding variable. If you define this macro, GCC will use it
7778 in place of both @code{ASM_OUTPUT_DECL} and
7779 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7780 the variable's decl in order to chose what to output.
7784 @subsection Output and Generation of Labels
7786 @c prevent bad page break with this line
7787 This is about outputting labels.
7789 @findex assemble_name
7790 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7791 A C statement (sans semicolon) to output to the stdio stream
7792 @var{stream} the assembler definition of a label named @var{name}.
7793 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7794 output the name itself; before and after that, output the additional
7795 assembler syntax for defining the name, and a newline. A default
7796 definition of this macro is provided which is correct for most systems.
7799 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7800 A C statement (sans semicolon) to output to the stdio stream
7801 @var{stream} the assembler definition of a label named @var{name} of
7803 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7804 output the name itself; before and after that, output the additional
7805 assembler syntax for defining the name, and a newline. A default
7806 definition of this macro is provided which is correct for most systems.
7808 If this macro is not defined, then the function name is defined in the
7809 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7812 @findex assemble_name_raw
7813 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7814 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7815 to refer to a compiler-generated label. The default definition uses
7816 @code{assemble_name_raw}, which is like @code{assemble_name} except
7817 that it is more efficient.
7821 A C string containing the appropriate assembler directive to specify the
7822 size of a symbol, without any arguments. On systems that use ELF, the
7823 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7824 systems, the default is not to define this macro.
7826 Define this macro only if it is correct to use the default definitions
7827 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7828 for your system. If you need your own custom definitions of those
7829 macros, or if you do not need explicit symbol sizes at all, do not
7833 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7834 A C statement (sans semicolon) to output to the stdio stream
7835 @var{stream} a directive telling the assembler that the size of the
7836 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7837 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7841 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7842 A C statement (sans semicolon) to output to the stdio stream
7843 @var{stream} a directive telling the assembler to calculate the size of
7844 the symbol @var{name} by subtracting its address from the current
7847 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7848 provided. The default assumes that the assembler recognizes a special
7849 @samp{.} symbol as referring to the current address, and can calculate
7850 the difference between this and another symbol. If your assembler does
7851 not recognize @samp{.} or cannot do calculations with it, you will need
7852 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7856 A C string containing the appropriate assembler directive to specify the
7857 type of a symbol, without any arguments. On systems that use ELF, the
7858 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7859 systems, the default is not to define this macro.
7861 Define this macro only if it is correct to use the default definition of
7862 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7863 custom definition of this macro, or if you do not need explicit symbol
7864 types at all, do not define this macro.
7867 @defmac TYPE_OPERAND_FMT
7868 A C string which specifies (using @code{printf} syntax) the format of
7869 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7870 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7871 the default is not to define this macro.
7873 Define this macro only if it is correct to use the default definition of
7874 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7875 custom definition of this macro, or if you do not need explicit symbol
7876 types at all, do not define this macro.
7879 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7880 A C statement (sans semicolon) to output to the stdio stream
7881 @var{stream} a directive telling the assembler that the type of the
7882 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7883 that string is always either @samp{"function"} or @samp{"object"}, but
7884 you should not count on this.
7886 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7887 definition of this macro is provided.
7890 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7891 A C statement (sans semicolon) to output to the stdio stream
7892 @var{stream} any text necessary for declaring the name @var{name} of a
7893 function which is being defined. This macro is responsible for
7894 outputting the label definition (perhaps using
7895 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7896 @code{FUNCTION_DECL} tree node representing the function.
7898 If this macro is not defined, then the function name is defined in the
7899 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7901 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7905 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7906 A C statement (sans semicolon) to output to the stdio stream
7907 @var{stream} any text necessary for declaring the size of a function
7908 which is being defined. The argument @var{name} is the name of the
7909 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7910 representing the function.
7912 If this macro is not defined, then the function size is not defined.
7914 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7918 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7919 A C statement (sans semicolon) to output to the stdio stream
7920 @var{stream} any text necessary for declaring the name @var{name} of an
7921 initialized variable which is being defined. This macro must output the
7922 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7923 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7925 If this macro is not defined, then the variable name is defined in the
7926 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7928 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7929 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7932 @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})
7933 A target hook to output to the stdio stream @var{file} any text necessary
7934 for declaring the name @var{name} of a constant which is being defined. This
7935 target hook is responsible for outputting the label definition (perhaps using
7936 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7937 and @var{size} is the size of the constant in bytes. The @var{name}
7938 will be an internal label.
7940 The default version of this target hook, define the @var{name} in the
7941 usual manner as a label (by means of @code{assemble_label}).
7943 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7946 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7947 A C statement (sans semicolon) to output to the stdio stream
7948 @var{stream} any text necessary for claiming a register @var{regno}
7949 for a global variable @var{decl} with name @var{name}.
7951 If you don't define this macro, that is equivalent to defining it to do
7955 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7956 A C statement (sans semicolon) to finish up declaring a variable name
7957 once the compiler has processed its initializer fully and thus has had a
7958 chance to determine the size of an array when controlled by an
7959 initializer. This is used on systems where it's necessary to declare
7960 something about the size of the object.
7962 If you don't define this macro, that is equivalent to defining it to do
7965 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7966 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7969 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7970 This target hook is a function to output to the stdio stream
7971 @var{stream} some commands that will make the label @var{name} global;
7972 that is, available for reference from other files.
7974 The default implementation relies on a proper definition of
7975 @code{GLOBAL_ASM_OP}.
7978 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7979 This target hook is a function to output to the stdio stream
7980 @var{stream} some commands that will make the name associated with @var{decl}
7981 global; that is, available for reference from other files.
7983 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7986 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7987 A C statement (sans semicolon) to output to the stdio stream
7988 @var{stream} some commands that will make the label @var{name} weak;
7989 that is, available for reference from other files but only used if
7990 no other definition is available. Use the expression
7991 @code{assemble_name (@var{stream}, @var{name})} to output the name
7992 itself; before and after that, output the additional assembler syntax
7993 for making that name weak, and a newline.
7995 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7996 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
8000 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
8001 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
8002 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
8003 or variable decl. If @var{value} is not @code{NULL}, this C statement
8004 should output to the stdio stream @var{stream} assembler code which
8005 defines (equates) the weak symbol @var{name} to have the value
8006 @var{value}. If @var{value} is @code{NULL}, it should output commands
8007 to make @var{name} weak.
8010 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
8011 Outputs a directive that enables @var{name} to be used to refer to
8012 symbol @var{value} with weak-symbol semantics. @code{decl} is the
8013 declaration of @code{name}.
8016 @defmac SUPPORTS_WEAK
8017 A preprocessor constant expression which evaluates to true if the target
8018 supports weak symbols.
8020 If you don't define this macro, @file{defaults.h} provides a default
8021 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
8022 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
8025 @defmac TARGET_SUPPORTS_WEAK
8026 A C expression which evaluates to true if the target supports weak symbols.
8028 If you don't define this macro, @file{defaults.h} provides a default
8029 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
8030 this macro if you want to control weak symbol support with a compiler
8031 flag such as @option{-melf}.
8034 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
8035 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
8036 public symbol such that extra copies in multiple translation units will
8037 be discarded by the linker. Define this macro if your object file
8038 format provides support for this concept, such as the @samp{COMDAT}
8039 section flags in the Microsoft Windows PE/COFF format, and this support
8040 requires changes to @var{decl}, such as putting it in a separate section.
8043 @defmac SUPPORTS_ONE_ONLY
8044 A C expression which evaluates to true if the target supports one-only
8047 If you don't define this macro, @file{varasm.c} provides a default
8048 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
8049 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
8050 you want to control one-only symbol support with a compiler flag, or if
8051 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
8052 be emitted as one-only.
8055 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
8056 This target hook is a function to output to @var{asm_out_file} some
8057 commands that will make the symbol(s) associated with @var{decl} have
8058 hidden, protected or internal visibility as specified by @var{visibility}.
8061 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
8062 A C expression that evaluates to true if the target's linker expects
8063 that weak symbols do not appear in a static archive's table of contents.
8064 The default is @code{0}.
8066 Leaving weak symbols out of an archive's table of contents means that,
8067 if a symbol will only have a definition in one translation unit and
8068 will have undefined references from other translation units, that
8069 symbol should not be weak. Defining this macro to be nonzero will
8070 thus have the effect that certain symbols that would normally be weak
8071 (explicit template instantiations, and vtables for polymorphic classes
8072 with noninline key methods) will instead be nonweak.
8074 The C++ ABI requires this macro to be zero. Define this macro for
8075 targets where full C++ ABI compliance is impossible and where linker
8076 restrictions require weak symbols to be left out of a static archive's
8080 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8081 A C statement (sans semicolon) to output to the stdio stream
8082 @var{stream} any text necessary for declaring the name of an external
8083 symbol named @var{name} which is referenced in this compilation but
8084 not defined. The value of @var{decl} is the tree node for the
8087 This macro need not be defined if it does not need to output anything.
8088 The GNU assembler and most Unix assemblers don't require anything.
8091 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8092 This target hook is a function to output to @var{asm_out_file} an assembler
8093 pseudo-op to declare a library function name external. The name of the
8094 library function is given by @var{symref}, which is a @code{symbol_ref}.
8097 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8098 This target hook is a function to output to @var{asm_out_file} an assembler
8099 directive to annotate @var{symbol} as used. The Darwin target uses the
8100 .no_dead_code_strip directive.
8103 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8104 A C statement (sans semicolon) to output to the stdio stream
8105 @var{stream} a reference in assembler syntax to a label named
8106 @var{name}. This should add @samp{_} to the front of the name, if that
8107 is customary on your operating system, as it is in most Berkeley Unix
8108 systems. This macro is used in @code{assemble_name}.
8111 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8112 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.
8115 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8116 A C statement (sans semicolon) to output a reference to
8117 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8118 will be used to output the name of the symbol. This macro may be used
8119 to modify the way a symbol is referenced depending on information
8120 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8123 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8124 A C statement (sans semicolon) to output a reference to @var{buf}, the
8125 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8126 @code{assemble_name} will be used to output the name of the symbol.
8127 This macro is not used by @code{output_asm_label}, or the @code{%l}
8128 specifier that calls it; the intention is that this macro should be set
8129 when it is necessary to output a label differently when its address is
8133 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8134 A function to output to the stdio stream @var{stream} a label whose
8135 name is made from the string @var{prefix} and the number @var{labelno}.
8137 It is absolutely essential that these labels be distinct from the labels
8138 used for user-level functions and variables. Otherwise, certain programs
8139 will have name conflicts with internal labels.
8141 It is desirable to exclude internal labels from the symbol table of the
8142 object file. Most assemblers have a naming convention for labels that
8143 should be excluded; on many systems, the letter @samp{L} at the
8144 beginning of a label has this effect. You should find out what
8145 convention your system uses, and follow it.
8147 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8150 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8151 A C statement to output to the stdio stream @var{stream} a debug info
8152 label whose name is made from the string @var{prefix} and the number
8153 @var{num}. This is useful for VLIW targets, where debug info labels
8154 may need to be treated differently than branch target labels. On some
8155 systems, branch target labels must be at the beginning of instruction
8156 bundles, but debug info labels can occur in the middle of instruction
8159 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8163 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8164 A C statement to store into the string @var{string} a label whose name
8165 is made from the string @var{prefix} and the number @var{num}.
8167 This string, when output subsequently by @code{assemble_name}, should
8168 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8169 with the same @var{prefix} and @var{num}.
8171 If the string begins with @samp{*}, then @code{assemble_name} will
8172 output the rest of the string unchanged. It is often convenient for
8173 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8174 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8175 to output the string, and may change it. (Of course,
8176 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8177 you should know what it does on your machine.)
8180 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8181 A C expression to assign to @var{outvar} (which is a variable of type
8182 @code{char *}) a newly allocated string made from the string
8183 @var{name} and the number @var{number}, with some suitable punctuation
8184 added. Use @code{alloca} to get space for the string.
8186 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8187 produce an assembler label for an internal static variable whose name is
8188 @var{name}. Therefore, the string must be such as to result in valid
8189 assembler code. The argument @var{number} is different each time this
8190 macro is executed; it prevents conflicts between similarly-named
8191 internal static variables in different scopes.
8193 Ideally this string should not be a valid C identifier, to prevent any
8194 conflict with the user's own symbols. Most assemblers allow periods
8195 or percent signs in assembler symbols; putting at least one of these
8196 between the name and the number will suffice.
8198 If this macro is not defined, a default definition will be provided
8199 which is correct for most systems.
8202 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8203 A C statement to output to the stdio stream @var{stream} assembler code
8204 which defines (equates) the symbol @var{name} to have the value @var{value}.
8207 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8208 correct for most systems.
8211 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8212 A C statement to output to the stdio stream @var{stream} assembler code
8213 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8214 to have the value of the tree node @var{decl_of_value}. This macro will
8215 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8216 the tree nodes are available.
8219 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8220 correct for most systems.
8223 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8224 A C statement that evaluates to true if the assembler code which defines
8225 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8226 of the tree node @var{decl_of_value} should be emitted near the end of the
8227 current compilation unit. The default is to not defer output of defines.
8228 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8229 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8232 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8233 A C statement to output to the stdio stream @var{stream} assembler code
8234 which defines (equates) the weak symbol @var{name} to have the value
8235 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8236 an undefined weak symbol.
8238 Define this macro if the target only supports weak aliases; define
8239 @code{ASM_OUTPUT_DEF} instead if possible.
8242 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8243 Define this macro to override the default assembler names used for
8244 Objective-C methods.
8246 The default name is a unique method number followed by the name of the
8247 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8248 the category is also included in the assembler name (e.g.@:
8251 These names are safe on most systems, but make debugging difficult since
8252 the method's selector is not present in the name. Therefore, particular
8253 systems define other ways of computing names.
8255 @var{buf} is an expression of type @code{char *} which gives you a
8256 buffer in which to store the name; its length is as long as
8257 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8258 50 characters extra.
8260 The argument @var{is_inst} specifies whether the method is an instance
8261 method or a class method; @var{class_name} is the name of the class;
8262 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8263 in a category); and @var{sel_name} is the name of the selector.
8265 On systems where the assembler can handle quoted names, you can use this
8266 macro to provide more human-readable names.
8269 @node Initialization
8270 @subsection How Initialization Functions Are Handled
8271 @cindex initialization routines
8272 @cindex termination routines
8273 @cindex constructors, output of
8274 @cindex destructors, output of
8276 The compiled code for certain languages includes @dfn{constructors}
8277 (also called @dfn{initialization routines})---functions to initialize
8278 data in the program when the program is started. These functions need
8279 to be called before the program is ``started''---that is to say, before
8280 @code{main} is called.
8282 Compiling some languages generates @dfn{destructors} (also called
8283 @dfn{termination routines}) that should be called when the program
8286 To make the initialization and termination functions work, the compiler
8287 must output something in the assembler code to cause those functions to
8288 be called at the appropriate time. When you port the compiler to a new
8289 system, you need to specify how to do this.
8291 There are two major ways that GCC currently supports the execution of
8292 initialization and termination functions. Each way has two variants.
8293 Much of the structure is common to all four variations.
8295 @findex __CTOR_LIST__
8296 @findex __DTOR_LIST__
8297 The linker must build two lists of these functions---a list of
8298 initialization functions, called @code{__CTOR_LIST__}, and a list of
8299 termination functions, called @code{__DTOR_LIST__}.
8301 Each list always begins with an ignored function pointer (which may hold
8302 0, @minus{}1, or a count of the function pointers after it, depending on
8303 the environment). This is followed by a series of zero or more function
8304 pointers to constructors (or destructors), followed by a function
8305 pointer containing zero.
8307 Depending on the operating system and its executable file format, either
8308 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8309 time and exit time. Constructors are called in reverse order of the
8310 list; destructors in forward order.
8312 The best way to handle static constructors works only for object file
8313 formats which provide arbitrarily-named sections. A section is set
8314 aside for a list of constructors, and another for a list of destructors.
8315 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8316 object file that defines an initialization function also puts a word in
8317 the constructor section to point to that function. The linker
8318 accumulates all these words into one contiguous @samp{.ctors} section.
8319 Termination functions are handled similarly.
8321 This method will be chosen as the default by @file{target-def.h} if
8322 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8323 support arbitrary sections, but does support special designated
8324 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8325 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8327 When arbitrary sections are available, there are two variants, depending
8328 upon how the code in @file{crtstuff.c} is called. On systems that
8329 support a @dfn{.init} section which is executed at program startup,
8330 parts of @file{crtstuff.c} are compiled into that section. The
8331 program is linked by the @command{gcc} driver like this:
8334 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8337 The prologue of a function (@code{__init}) appears in the @code{.init}
8338 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8339 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8340 files are provided by the operating system or by the GNU C library, but
8341 are provided by GCC for a few targets.
8343 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8344 compiled from @file{crtstuff.c}. They contain, among other things, code
8345 fragments within the @code{.init} and @code{.fini} sections that branch
8346 to routines in the @code{.text} section. The linker will pull all parts
8347 of a section together, which results in a complete @code{__init} function
8348 that invokes the routines we need at startup.
8350 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8353 If no init section is available, when GCC compiles any function called
8354 @code{main} (or more accurately, any function designated as a program
8355 entry point by the language front end calling @code{expand_main_function}),
8356 it inserts a procedure call to @code{__main} as the first executable code
8357 after the function prologue. The @code{__main} function is defined
8358 in @file{libgcc2.c} and runs the global constructors.
8360 In file formats that don't support arbitrary sections, there are again
8361 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8362 and an `a.out' format must be used. In this case,
8363 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8364 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8365 and with the address of the void function containing the initialization
8366 code as its value. The GNU linker recognizes this as a request to add
8367 the value to a @dfn{set}; the values are accumulated, and are eventually
8368 placed in the executable as a vector in the format described above, with
8369 a leading (ignored) count and a trailing zero element.
8370 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8371 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8372 the compilation of @code{main} to call @code{__main} as above, starting
8373 the initialization process.
8375 The last variant uses neither arbitrary sections nor the GNU linker.
8376 This is preferable when you want to do dynamic linking and when using
8377 file formats which the GNU linker does not support, such as `ECOFF'@. In
8378 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8379 termination functions are recognized simply by their names. This requires
8380 an extra program in the linkage step, called @command{collect2}. This program
8381 pretends to be the linker, for use with GCC; it does its job by running
8382 the ordinary linker, but also arranges to include the vectors of
8383 initialization and termination functions. These functions are called
8384 via @code{__main} as described above. In order to use this method,
8385 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8388 The following section describes the specific macros that control and
8389 customize the handling of initialization and termination functions.
8392 @node Macros for Initialization
8393 @subsection Macros Controlling Initialization Routines
8395 Here are the macros that control how the compiler handles initialization
8396 and termination functions:
8398 @defmac INIT_SECTION_ASM_OP
8399 If defined, a C string constant, including spacing, for the assembler
8400 operation to identify the following data as initialization code. If not
8401 defined, GCC will assume such a section does not exist. When you are
8402 using special sections for initialization and termination functions, this
8403 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8404 run the initialization functions.
8407 @defmac HAS_INIT_SECTION
8408 If defined, @code{main} will not call @code{__main} as described above.
8409 This macro should be defined for systems that control start-up code
8410 on a symbol-by-symbol basis, such as OSF/1, and should not
8411 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8414 @defmac LD_INIT_SWITCH
8415 If defined, a C string constant for a switch that tells the linker that
8416 the following symbol is an initialization routine.
8419 @defmac LD_FINI_SWITCH
8420 If defined, a C string constant for a switch that tells the linker that
8421 the following symbol is a finalization routine.
8424 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8425 If defined, a C statement that will write a function that can be
8426 automatically called when a shared library is loaded. The function
8427 should call @var{func}, which takes no arguments. If not defined, and
8428 the object format requires an explicit initialization function, then a
8429 function called @code{_GLOBAL__DI} will be generated.
8431 This function and the following one are used by collect2 when linking a
8432 shared library that needs constructors or destructors, or has DWARF2
8433 exception tables embedded in the code.
8436 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8437 If defined, a C statement that will write a function that can be
8438 automatically called when a shared library is unloaded. The function
8439 should call @var{func}, which takes no arguments. If not defined, and
8440 the object format requires an explicit finalization function, then a
8441 function called @code{_GLOBAL__DD} will be generated.
8444 @defmac INVOKE__main
8445 If defined, @code{main} will call @code{__main} despite the presence of
8446 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8447 where the init section is not actually run automatically, but is still
8448 useful for collecting the lists of constructors and destructors.
8451 @defmac SUPPORTS_INIT_PRIORITY
8452 If nonzero, the C++ @code{init_priority} attribute is supported and the
8453 compiler should emit instructions to control the order of initialization
8454 of objects. If zero, the compiler will issue an error message upon
8455 encountering an @code{init_priority} attribute.
8458 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8459 This value is true if the target supports some ``native'' method of
8460 collecting constructors and destructors to be run at startup and exit.
8461 It is false if we must use @command{collect2}.
8464 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8465 If defined, a function that outputs assembler code to arrange to call
8466 the function referenced by @var{symbol} at initialization time.
8468 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8469 no arguments and with no return value. If the target supports initialization
8470 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8471 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8473 If this macro is not defined by the target, a suitable default will
8474 be chosen if (1) the target supports arbitrary section names, (2) the
8475 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8479 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8480 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8481 functions rather than initialization functions.
8484 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8485 generated for the generated object file will have static linkage.
8487 If your system uses @command{collect2} as the means of processing
8488 constructors, then that program normally uses @command{nm} to scan
8489 an object file for constructor functions to be called.
8491 On certain kinds of systems, you can define this macro to make
8492 @command{collect2} work faster (and, in some cases, make it work at all):
8494 @defmac OBJECT_FORMAT_COFF
8495 Define this macro if the system uses COFF (Common Object File Format)
8496 object files, so that @command{collect2} can assume this format and scan
8497 object files directly for dynamic constructor/destructor functions.
8499 This macro is effective only in a native compiler; @command{collect2} as
8500 part of a cross compiler always uses @command{nm} for the target machine.
8503 @defmac REAL_NM_FILE_NAME
8504 Define this macro as a C string constant containing the file name to use
8505 to execute @command{nm}. The default is to search the path normally for
8510 @command{collect2} calls @command{nm} to scan object files for static
8511 constructors and destructors and LTO info. By default, @option{-n} is
8512 passed. Define @code{NM_FLAGS} to a C string constant if other options
8513 are needed to get the same output format as GNU @command{nm -n}
8517 If your system supports shared libraries and has a program to list the
8518 dynamic dependencies of a given library or executable, you can define
8519 these macros to enable support for running initialization and
8520 termination functions in shared libraries:
8523 Define this macro to a C string constant containing the name of the program
8524 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8527 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8528 Define this macro to be C code that extracts filenames from the output
8529 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8530 of type @code{char *} that points to the beginning of a line of output
8531 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8532 code must advance @var{ptr} to the beginning of the filename on that
8533 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8536 @defmac SHLIB_SUFFIX
8537 Define this macro to a C string constant containing the default shared
8538 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8539 strips version information after this suffix when generating global
8540 constructor and destructor names. This define is only needed on targets
8541 that use @command{collect2} to process constructors and destructors.
8544 @node Instruction Output
8545 @subsection Output of Assembler Instructions
8547 @c prevent bad page break with this line
8548 This describes assembler instruction output.
8550 @defmac REGISTER_NAMES
8551 A C initializer containing the assembler's names for the machine
8552 registers, each one as a C string constant. This is what translates
8553 register numbers in the compiler into assembler language.
8556 @defmac ADDITIONAL_REGISTER_NAMES
8557 If defined, a C initializer for an array of structures containing a name
8558 and a register number. This macro defines additional names for hard
8559 registers, thus allowing the @code{asm} option in declarations to refer
8560 to registers using alternate names.
8563 @defmac OVERLAPPING_REGISTER_NAMES
8564 If defined, a C initializer for an array of structures containing a
8565 name, a register number and a count of the number of consecutive
8566 machine registers the name overlaps. This macro defines additional
8567 names for hard registers, thus allowing the @code{asm} option in
8568 declarations to refer to registers using alternate names. Unlike
8569 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8570 register name implies multiple underlying registers.
8572 This macro should be used when it is important that a clobber in an
8573 @code{asm} statement clobbers all the underlying values implied by the
8574 register name. For example, on ARM, clobbering the double-precision
8575 VFP register ``d0'' implies clobbering both single-precision registers
8579 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8580 Define this macro if you are using an unusual assembler that
8581 requires different names for the machine instructions.
8583 The definition is a C statement or statements which output an
8584 assembler instruction opcode to the stdio stream @var{stream}. The
8585 macro-operand @var{ptr} is a variable of type @code{char *} which
8586 points to the opcode name in its ``internal'' form---the form that is
8587 written in the machine description. The definition should output the
8588 opcode name to @var{stream}, performing any translation you desire, and
8589 increment the variable @var{ptr} to point at the end of the opcode
8590 so that it will not be output twice.
8592 In fact, your macro definition may process less than the entire opcode
8593 name, or more than the opcode name; but if you want to process text
8594 that includes @samp{%}-sequences to substitute operands, you must take
8595 care of the substitution yourself. Just be sure to increment
8596 @var{ptr} over whatever text should not be output normally.
8598 @findex recog_data.operand
8599 If you need to look at the operand values, they can be found as the
8600 elements of @code{recog_data.operand}.
8602 If the macro definition does nothing, the instruction is output
8606 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8607 If defined, a C statement to be executed just prior to the output of
8608 assembler code for @var{insn}, to modify the extracted operands so
8609 they will be output differently.
8611 Here the argument @var{opvec} is the vector containing the operands
8612 extracted from @var{insn}, and @var{noperands} is the number of
8613 elements of the vector which contain meaningful data for this insn.
8614 The contents of this vector are what will be used to convert the insn
8615 template into assembler code, so you can change the assembler output
8616 by changing the contents of the vector.
8618 This macro is useful when various assembler syntaxes share a single
8619 file of instruction patterns; by defining this macro differently, you
8620 can cause a large class of instructions to be output differently (such
8621 as with rearranged operands). Naturally, variations in assembler
8622 syntax affecting individual insn patterns ought to be handled by
8623 writing conditional output routines in those patterns.
8625 If this macro is not defined, it is equivalent to a null statement.
8628 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx @var{insn}, rtx *@var{opvec}, int @var{noperands})
8629 If defined, this target hook is a function which is executed just after the
8630 output of assembler code for @var{insn}, to change the mode of the assembler
8633 Here the argument @var{opvec} is the vector containing the operands
8634 extracted from @var{insn}, and @var{noperands} is the number of
8635 elements of the vector which contain meaningful data for this insn.
8636 The contents of this vector are what was used to convert the insn
8637 template into assembler code, so you can change the assembler mode
8638 by checking the contents of the vector.
8641 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8642 A C compound statement to output to stdio stream @var{stream} the
8643 assembler syntax for an instruction operand @var{x}. @var{x} is an
8646 @var{code} is a value that can be used to specify one of several ways
8647 of printing the operand. It is used when identical operands must be
8648 printed differently depending on the context. @var{code} comes from
8649 the @samp{%} specification that was used to request printing of the
8650 operand. If the specification was just @samp{%@var{digit}} then
8651 @var{code} is 0; if the specification was @samp{%@var{ltr}
8652 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8655 If @var{x} is a register, this macro should print the register's name.
8656 The names can be found in an array @code{reg_names} whose type is
8657 @code{char *[]}. @code{reg_names} is initialized from
8658 @code{REGISTER_NAMES}.
8660 When the machine description has a specification @samp{%@var{punct}}
8661 (a @samp{%} followed by a punctuation character), this macro is called
8662 with a null pointer for @var{x} and the punctuation character for
8666 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8667 A C expression which evaluates to true if @var{code} is a valid
8668 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8669 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8670 punctuation characters (except for the standard one, @samp{%}) are used
8674 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8675 A C compound statement to output to stdio stream @var{stream} the
8676 assembler syntax for an instruction operand that is a memory reference
8677 whose address is @var{x}. @var{x} is an RTL expression.
8679 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8680 On some machines, the syntax for a symbolic address depends on the
8681 section that the address refers to. On these machines, define the hook
8682 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8683 @code{symbol_ref}, and then check for it here. @xref{Assembler
8687 @findex dbr_sequence_length
8688 @defmac DBR_OUTPUT_SEQEND (@var{file})
8689 A C statement, to be executed after all slot-filler instructions have
8690 been output. If necessary, call @code{dbr_sequence_length} to
8691 determine the number of slots filled in a sequence (zero if not
8692 currently outputting a sequence), to decide how many no-ops to output,
8695 Don't define this macro if it has nothing to do, but it is helpful in
8696 reading assembly output if the extent of the delay sequence is made
8697 explicit (e.g.@: with white space).
8700 @findex final_sequence
8701 Note that output routines for instructions with delay slots must be
8702 prepared to deal with not being output as part of a sequence
8703 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8704 found.) The variable @code{final_sequence} is null when not
8705 processing a sequence, otherwise it contains the @code{sequence} rtx
8709 @defmac REGISTER_PREFIX
8710 @defmacx LOCAL_LABEL_PREFIX
8711 @defmacx USER_LABEL_PREFIX
8712 @defmacx IMMEDIATE_PREFIX
8713 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8714 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8715 @file{final.c}). These are useful when a single @file{md} file must
8716 support multiple assembler formats. In that case, the various @file{tm.h}
8717 files can define these macros differently.
8720 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8721 If defined this macro should expand to a series of @code{case}
8722 statements which will be parsed inside the @code{switch} statement of
8723 the @code{asm_fprintf} function. This allows targets to define extra
8724 printf formats which may useful when generating their assembler
8725 statements. Note that uppercase letters are reserved for future
8726 generic extensions to asm_fprintf, and so are not available to target
8727 specific code. The output file is given by the parameter @var{file}.
8728 The varargs input pointer is @var{argptr} and the rest of the format
8729 string, starting the character after the one that is being switched
8730 upon, is pointed to by @var{format}.
8733 @defmac ASSEMBLER_DIALECT
8734 If your target supports multiple dialects of assembler language (such as
8735 different opcodes), define this macro as a C expression that gives the
8736 numeric index of the assembler language dialect to use, with zero as the
8739 If this macro is defined, you may use constructs of the form
8741 @samp{@{option0|option1|option2@dots{}@}}
8744 in the output templates of patterns (@pxref{Output Template}) or in the
8745 first argument of @code{asm_fprintf}. This construct outputs
8746 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8747 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8748 within these strings retain their usual meaning. If there are fewer
8749 alternatives within the braces than the value of
8750 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8752 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8753 @samp{@}} do not have any special meaning when used in templates or
8754 operands to @code{asm_fprintf}.
8756 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8757 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8758 the variations in assembler language syntax with that mechanism. Define
8759 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8760 if the syntax variant are larger and involve such things as different
8761 opcodes or operand order.
8764 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8765 A C expression to output to @var{stream} some assembler code
8766 which will push hard register number @var{regno} onto the stack.
8767 The code need not be optimal, since this macro is used only when
8771 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8772 A C expression to output to @var{stream} some assembler code
8773 which will pop hard register number @var{regno} off of the stack.
8774 The code need not be optimal, since this macro is used only when
8778 @node Dispatch Tables
8779 @subsection Output of Dispatch Tables
8781 @c prevent bad page break with this line
8782 This concerns dispatch tables.
8784 @cindex dispatch table
8785 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8786 A C statement to output to the stdio stream @var{stream} an assembler
8787 pseudo-instruction to generate a difference between two labels.
8788 @var{value} and @var{rel} are the numbers of two internal labels. The
8789 definitions of these labels are output using
8790 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8791 way here. For example,
8794 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8795 @var{value}, @var{rel})
8798 You must provide this macro on machines where the addresses in a
8799 dispatch table are relative to the table's own address. If defined, GCC
8800 will also use this macro on all machines when producing PIC@.
8801 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8802 mode and flags can be read.
8805 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8806 This macro should be provided on machines where the addresses
8807 in a dispatch table are absolute.
8809 The definition should be a C statement to output to the stdio stream
8810 @var{stream} an assembler pseudo-instruction to generate a reference to
8811 a label. @var{value} is the number of an internal label whose
8812 definition is output using @code{(*targetm.asm_out.internal_label)}.
8816 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8820 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8821 Define this if the label before a jump-table needs to be output
8822 specially. The first three arguments are the same as for
8823 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8824 jump-table which follows (a @code{jump_insn} containing an
8825 @code{addr_vec} or @code{addr_diff_vec}).
8827 This feature is used on system V to output a @code{swbeg} statement
8830 If this macro is not defined, these labels are output with
8831 @code{(*targetm.asm_out.internal_label)}.
8834 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8835 Define this if something special must be output at the end of a
8836 jump-table. The definition should be a C statement to be executed
8837 after the assembler code for the table is written. It should write
8838 the appropriate code to stdio stream @var{stream}. The argument
8839 @var{table} is the jump-table insn, and @var{num} is the label-number
8840 of the preceding label.
8842 If this macro is not defined, nothing special is output at the end of
8846 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
8847 This target hook emits a label at the beginning of each FDE@. It
8848 should be defined on targets where FDEs need special labels, and it
8849 should write the appropriate label, for the FDE associated with the
8850 function declaration @var{decl}, to the stdio stream @var{stream}.
8851 The third argument, @var{for_eh}, is a boolean: true if this is for an
8852 exception table. The fourth argument, @var{empty}, is a boolean:
8853 true if this is a placeholder label for an omitted FDE@.
8855 The default is that FDEs are not given nonlocal labels.
8858 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
8859 This target hook emits a label at the beginning of the exception table.
8860 It should be defined on targets where it is desirable for the table
8861 to be broken up according to function.
8863 The default is that no label is emitted.
8866 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
8867 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.
8870 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx @var{insn})
8871 This target hook emits assembly directives required to unwind the
8872 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8873 returns @code{UI_TARGET}.
8876 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8877 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.
8880 @node Exception Region Output
8881 @subsection Assembler Commands for Exception Regions
8883 @c prevent bad page break with this line
8885 This describes commands marking the start and the end of an exception
8888 @defmac EH_FRAME_SECTION_NAME
8889 If defined, a C string constant for the name of the section containing
8890 exception handling frame unwind information. If not defined, GCC will
8891 provide a default definition if the target supports named sections.
8892 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8894 You should define this symbol if your target supports DWARF 2 frame
8895 unwind information and the default definition does not work.
8898 @defmac EH_FRAME_IN_DATA_SECTION
8899 If defined, DWARF 2 frame unwind information will be placed in the
8900 data section even though the target supports named sections. This
8901 might be necessary, for instance, if the system linker does garbage
8902 collection and sections cannot be marked as not to be collected.
8904 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8908 @defmac EH_TABLES_CAN_BE_READ_ONLY
8909 Define this macro to 1 if your target is such that no frame unwind
8910 information encoding used with non-PIC code will ever require a
8911 runtime relocation, but the linker may not support merging read-only
8912 and read-write sections into a single read-write section.
8915 @defmac MASK_RETURN_ADDR
8916 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8917 that it does not contain any extraneous set bits in it.
8920 @defmac DWARF2_UNWIND_INFO
8921 Define this macro to 0 if your target supports DWARF 2 frame unwind
8922 information, but it does not yet work with exception handling.
8923 Otherwise, if your target supports this information (if it defines
8924 @code{INCOMING_RETURN_ADDR_RTX} and either @code{UNALIGNED_INT_ASM_OP}
8925 or @code{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8928 @deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
8929 This hook defines the mechanism that will be used for exception handling
8930 by the target. If the target has ABI specified unwind tables, the hook
8931 should return @code{UI_TARGET}. If the target is to use the
8932 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8933 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8934 information, the hook should return @code{UI_DWARF2}.
8936 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8937 This may end up simplifying other parts of target-specific code. The
8938 default implementation of this hook never returns @code{UI_NONE}.
8940 Note that the value returned by this hook should be constant. It should
8941 not depend on anything except the command-line switches described by
8942 @var{opts}. In particular, the
8943 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8944 macros and builtin functions related to exception handling are set up
8945 depending on this setting.
8947 The default implementation of the hook first honors the
8948 @option{--enable-sjlj-exceptions} configure option, then
8949 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
8950 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
8951 must define this hook so that @var{opts} is used correctly.
8954 @deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8955 This variable should be set to @code{true} if the target ABI requires unwinding
8956 tables even when exceptions are not used. It must not be modified by
8957 command-line option processing.
8960 @defmac DONT_USE_BUILTIN_SETJMP
8961 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8962 should use the @code{setjmp}/@code{longjmp} functions from the C library
8963 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8966 @defmac DWARF_CIE_DATA_ALIGNMENT
8967 This macro need only be defined if the target might save registers in the
8968 function prologue at an offset to the stack pointer that is not aligned to
8969 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8970 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8971 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8972 the target supports DWARF 2 frame unwind information.
8975 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8976 Contains the value true if the target should add a zero word onto the
8977 end of a Dwarf-2 frame info section when used for exception handling.
8978 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8982 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8983 Given a register, this hook should return a parallel of registers to
8984 represent where to find the register pieces. Define this hook if the
8985 register and its mode are represented in Dwarf in non-contiguous
8986 locations, or if the register should be represented in more than one
8987 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8988 If not defined, the default is to return @code{NULL_RTX}.
8991 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8992 If some registers are represented in Dwarf-2 unwind information in
8993 multiple pieces, define this hook to fill in information about the
8994 sizes of those pieces in the table used by the unwinder at runtime.
8995 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8996 filling in a single size corresponding to each hard register;
8997 @var{address} is the address of the table.
9000 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
9001 This hook is used to output a reference from a frame unwinding table to
9002 the type_info object identified by @var{sym}. It should return @code{true}
9003 if the reference was output. Returning @code{false} will cause the
9004 reference to be output using the normal Dwarf2 routines.
9007 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
9008 This flag should be set to @code{true} on targets that use an ARM EABI
9009 based unwinding library, and @code{false} on other targets. This effects
9010 the format of unwinding tables, and how the unwinder in entered after
9011 running a cleanup. The default is @code{false}.
9014 @node Alignment Output
9015 @subsection Assembler Commands for Alignment
9017 @c prevent bad page break with this line
9018 This describes commands for alignment.
9020 @defmac JUMP_ALIGN (@var{label})
9021 The alignment (log base 2) to put in front of @var{label}, which is
9022 a common destination of jumps and has no fallthru incoming edge.
9024 This macro need not be defined if you don't want any special alignment
9025 to be done at such a time. Most machine descriptions do not currently
9028 Unless it's necessary to inspect the @var{label} parameter, it is better
9029 to set the variable @var{align_jumps} in the target's
9030 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9031 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
9034 @deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx @var{label})
9035 The maximum number of bytes to skip before @var{label} when applying
9036 @code{JUMP_ALIGN}. This works only if
9037 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9040 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
9041 The alignment (log base 2) to put in front of @var{label}, which follows
9044 This macro need not be defined if you don't want any special alignment
9045 to be done at such a time. Most machine descriptions do not currently
9049 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx @var{label})
9050 The maximum number of bytes to skip before @var{label} when applying
9051 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
9052 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9055 @defmac LOOP_ALIGN (@var{label})
9056 The alignment (log base 2) to put in front of @var{label}, which follows
9057 a @code{NOTE_INSN_LOOP_BEG} note.
9059 This macro need not be defined if you don't want any special alignment
9060 to be done at such a time. Most machine descriptions do not currently
9063 Unless it's necessary to inspect the @var{label} parameter, it is better
9064 to set the variable @code{align_loops} in the target's
9065 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9066 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
9069 @deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx @var{label})
9070 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
9071 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
9075 @defmac LABEL_ALIGN (@var{label})
9076 The alignment (log base 2) to put in front of @var{label}.
9077 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
9078 the maximum of the specified values is used.
9080 Unless it's necessary to inspect the @var{label} parameter, it is better
9081 to set the variable @code{align_labels} in the target's
9082 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9083 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9086 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx @var{label})
9087 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
9088 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
9092 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9093 A C statement to output to the stdio stream @var{stream} an assembler
9094 instruction to advance the location counter by @var{nbytes} bytes.
9095 Those bytes should be zero when loaded. @var{nbytes} will be a C
9096 expression of type @code{unsigned HOST_WIDE_INT}.
9099 @defmac ASM_NO_SKIP_IN_TEXT
9100 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9101 text section because it fails to put zeros in the bytes that are skipped.
9102 This is true on many Unix systems, where the pseudo--op to skip bytes
9103 produces no-op instructions rather than zeros when used in the text
9107 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9108 A C statement to output to the stdio stream @var{stream} an assembler
9109 command to advance the location counter to a multiple of 2 to the
9110 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9113 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9114 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9115 for padding, if necessary.
9118 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9119 A C statement to output to the stdio stream @var{stream} an assembler
9120 command to advance the location counter to a multiple of 2 to the
9121 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9122 satisfy the alignment request. @var{power} and @var{max_skip} will be
9123 a C expression of type @code{int}.
9127 @node Debugging Info
9128 @section Controlling Debugging Information Format
9130 @c prevent bad page break with this line
9131 This describes how to specify debugging information.
9134 * All Debuggers:: Macros that affect all debugging formats uniformly.
9135 * DBX Options:: Macros enabling specific options in DBX format.
9136 * DBX Hooks:: Hook macros for varying DBX format.
9137 * File Names and DBX:: Macros controlling output of file names in DBX format.
9138 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9139 * VMS Debug:: Macros for VMS debug format.
9143 @subsection Macros Affecting All Debugging Formats
9145 @c prevent bad page break with this line
9146 These macros affect all debugging formats.
9148 @defmac DBX_REGISTER_NUMBER (@var{regno})
9149 A C expression that returns the DBX register number for the compiler
9150 register number @var{regno}. In the default macro provided, the value
9151 of this expression will be @var{regno} itself. But sometimes there are
9152 some registers that the compiler knows about and DBX does not, or vice
9153 versa. In such cases, some register may need to have one number in the
9154 compiler and another for DBX@.
9156 If two registers have consecutive numbers inside GCC, and they can be
9157 used as a pair to hold a multiword value, then they @emph{must} have
9158 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9159 Otherwise, debuggers will be unable to access such a pair, because they
9160 expect register pairs to be consecutive in their own numbering scheme.
9162 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9163 does not preserve register pairs, then what you must do instead is
9164 redefine the actual register numbering scheme.
9167 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9168 A C expression that returns the integer offset value for an automatic
9169 variable having address @var{x} (an RTL expression). The default
9170 computation assumes that @var{x} is based on the frame-pointer and
9171 gives the offset from the frame-pointer. This is required for targets
9172 that produce debugging output for DBX or COFF-style debugging output
9173 for SDB and allow the frame-pointer to be eliminated when the
9174 @option{-g} options is used.
9177 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9178 A C expression that returns the integer offset value for an argument
9179 having address @var{x} (an RTL expression). The nominal offset is
9183 @defmac PREFERRED_DEBUGGING_TYPE
9184 A C expression that returns the type of debugging output GCC should
9185 produce when the user specifies just @option{-g}. Define
9186 this if you have arranged for GCC to support more than one format of
9187 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9188 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9189 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9191 When the user specifies @option{-ggdb}, GCC normally also uses the
9192 value of this macro to select the debugging output format, but with two
9193 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9194 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9195 defined, GCC uses @code{DBX_DEBUG}.
9197 The value of this macro only affects the default debugging output; the
9198 user can always get a specific type of output by using @option{-gstabs},
9199 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9203 @subsection Specific Options for DBX Output
9205 @c prevent bad page break with this line
9206 These are specific options for DBX output.
9208 @defmac DBX_DEBUGGING_INFO
9209 Define this macro if GCC should produce debugging output for DBX
9210 in response to the @option{-g} option.
9213 @defmac XCOFF_DEBUGGING_INFO
9214 Define this macro if GCC should produce XCOFF format debugging output
9215 in response to the @option{-g} option. This is a variant of DBX format.
9218 @defmac DEFAULT_GDB_EXTENSIONS
9219 Define this macro to control whether GCC should by default generate
9220 GDB's extended version of DBX debugging information (assuming DBX-format
9221 debugging information is enabled at all). If you don't define the
9222 macro, the default is 1: always generate the extended information
9223 if there is any occasion to.
9226 @defmac DEBUG_SYMS_TEXT
9227 Define this macro if all @code{.stabs} commands should be output while
9228 in the text section.
9231 @defmac ASM_STABS_OP
9232 A C string constant, including spacing, naming the assembler pseudo op to
9233 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9234 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9235 applies only to DBX debugging information format.
9238 @defmac ASM_STABD_OP
9239 A C string constant, including spacing, naming the assembler pseudo op to
9240 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9241 value is the current location. If you don't define this macro,
9242 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9246 @defmac ASM_STABN_OP
9247 A C string constant, including spacing, naming the assembler pseudo op to
9248 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9249 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9250 macro applies only to DBX debugging information format.
9253 @defmac DBX_NO_XREFS
9254 Define this macro if DBX on your system does not support the construct
9255 @samp{xs@var{tagname}}. On some systems, this construct is used to
9256 describe a forward reference to a structure named @var{tagname}.
9257 On other systems, this construct is not supported at all.
9260 @defmac DBX_CONTIN_LENGTH
9261 A symbol name in DBX-format debugging information is normally
9262 continued (split into two separate @code{.stabs} directives) when it
9263 exceeds a certain length (by default, 80 characters). On some
9264 operating systems, DBX requires this splitting; on others, splitting
9265 must not be done. You can inhibit splitting by defining this macro
9266 with the value zero. You can override the default splitting-length by
9267 defining this macro as an expression for the length you desire.
9270 @defmac DBX_CONTIN_CHAR
9271 Normally continuation is indicated by adding a @samp{\} character to
9272 the end of a @code{.stabs} string when a continuation follows. To use
9273 a different character instead, define this macro as a character
9274 constant for the character you want to use. Do not define this macro
9275 if backslash is correct for your system.
9278 @defmac DBX_STATIC_STAB_DATA_SECTION
9279 Define this macro if it is necessary to go to the data section before
9280 outputting the @samp{.stabs} pseudo-op for a non-global static
9284 @defmac DBX_TYPE_DECL_STABS_CODE
9285 The value to use in the ``code'' field of the @code{.stabs} directive
9286 for a typedef. The default is @code{N_LSYM}.
9289 @defmac DBX_STATIC_CONST_VAR_CODE
9290 The value to use in the ``code'' field of the @code{.stabs} directive
9291 for a static variable located in the text section. DBX format does not
9292 provide any ``right'' way to do this. The default is @code{N_FUN}.
9295 @defmac DBX_REGPARM_STABS_CODE
9296 The value to use in the ``code'' field of the @code{.stabs} directive
9297 for a parameter passed in registers. DBX format does not provide any
9298 ``right'' way to do this. The default is @code{N_RSYM}.
9301 @defmac DBX_REGPARM_STABS_LETTER
9302 The letter to use in DBX symbol data to identify a symbol as a parameter
9303 passed in registers. DBX format does not customarily provide any way to
9304 do this. The default is @code{'P'}.
9307 @defmac DBX_FUNCTION_FIRST
9308 Define this macro if the DBX information for a function and its
9309 arguments should precede the assembler code for the function. Normally,
9310 in DBX format, the debugging information entirely follows the assembler
9314 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9315 Define this macro, with value 1, if the value of a symbol describing
9316 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9317 relative to the start of the enclosing function. Normally, GCC uses
9318 an absolute address.
9321 @defmac DBX_LINES_FUNCTION_RELATIVE
9322 Define this macro, with value 1, if the value of a symbol indicating
9323 the current line number (@code{N_SLINE}) should be relative to the
9324 start of the enclosing function. Normally, GCC uses an absolute address.
9327 @defmac DBX_USE_BINCL
9328 Define this macro if GCC should generate @code{N_BINCL} and
9329 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9330 macro also directs GCC to output a type number as a pair of a file
9331 number and a type number within the file. Normally, GCC does not
9332 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9333 number for a type number.
9337 @subsection Open-Ended Hooks for DBX Format
9339 @c prevent bad page break with this line
9340 These are hooks for DBX format.
9342 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9343 Define this macro to say how to output to @var{stream} the debugging
9344 information for the start of a scope level for variable names. The
9345 argument @var{name} is the name of an assembler symbol (for use with
9346 @code{assemble_name}) whose value is the address where the scope begins.
9349 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9350 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9353 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9354 Define this macro if the target machine requires special handling to
9355 output an @code{N_FUN} entry for the function @var{decl}.
9358 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9359 A C statement to output DBX debugging information before code for line
9360 number @var{line} of the current source file to the stdio stream
9361 @var{stream}. @var{counter} is the number of time the macro was
9362 invoked, including the current invocation; it is intended to generate
9363 unique labels in the assembly output.
9365 This macro should not be defined if the default output is correct, or
9366 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9369 @defmac NO_DBX_FUNCTION_END
9370 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9371 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9372 On those machines, define this macro to turn this feature off without
9373 disturbing the rest of the gdb extensions.
9376 @defmac NO_DBX_BNSYM_ENSYM
9377 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9378 extension construct. On those machines, define this macro to turn this
9379 feature off without disturbing the rest of the gdb extensions.
9382 @node File Names and DBX
9383 @subsection File Names in DBX Format
9385 @c prevent bad page break with this line
9386 This describes file names in DBX format.
9388 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9389 A C statement to output DBX debugging information to the stdio stream
9390 @var{stream}, which indicates that file @var{name} is the main source
9391 file---the file specified as the input file for compilation.
9392 This macro is called only once, at the beginning of compilation.
9394 This macro need not be defined if the standard form of output
9395 for DBX debugging information is appropriate.
9397 It may be necessary to refer to a label equal to the beginning of the
9398 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9399 to do so. If you do this, you must also set the variable
9400 @var{used_ltext_label_name} to @code{true}.
9403 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9404 Define this macro, with value 1, if GCC should not emit an indication
9405 of the current directory for compilation and current source language at
9406 the beginning of the file.
9409 @defmac NO_DBX_GCC_MARKER
9410 Define this macro, with value 1, if GCC should not emit an indication
9411 that this object file was compiled by GCC@. The default is to emit
9412 an @code{N_OPT} stab at the beginning of every source file, with
9413 @samp{gcc2_compiled.} for the string and value 0.
9416 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9417 A C statement to output DBX debugging information at the end of
9418 compilation of the main source file @var{name}. Output should be
9419 written to the stdio stream @var{stream}.
9421 If you don't define this macro, nothing special is output at the end
9422 of compilation, which is correct for most machines.
9425 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9426 Define this macro @emph{instead of} defining
9427 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9428 the end of compilation is an @code{N_SO} stab with an empty string,
9429 whose value is the highest absolute text address in the file.
9434 @subsection Macros for SDB and DWARF Output
9436 @c prevent bad page break with this line
9437 Here are macros for SDB and DWARF output.
9439 @defmac SDB_DEBUGGING_INFO
9440 Define this macro if GCC should produce COFF-style debugging output
9441 for SDB in response to the @option{-g} option.
9444 @defmac DWARF2_DEBUGGING_INFO
9445 Define this macro if GCC should produce dwarf version 2 format
9446 debugging output in response to the @option{-g} option.
9448 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
9449 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9450 be emitted for each function. Instead of an integer return the enum
9451 value for the @code{DW_CC_} tag.
9454 To support optional call frame debugging information, you must also
9455 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9456 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9457 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9458 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9461 @defmac DWARF2_FRAME_INFO
9462 Define this macro to a nonzero value if GCC should always output
9463 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9464 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9465 exceptions are enabled, GCC will output this information not matter
9466 how you define @code{DWARF2_FRAME_INFO}.
9469 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
9470 This hook defines the mechanism that will be used for describing frame
9471 unwind information to the debugger. Normally the hook will return
9472 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9473 return @code{UI_NONE} otherwise.
9475 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9476 is disabled in order to always output DWARF 2 frame information.
9478 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9479 This will suppress generation of the normal debug frame unwind information.
9482 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9483 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9484 line debug info sections. This will result in much more compact line number
9485 tables, and hence is desirable if it works.
9488 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
9489 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.
9492 @deftypevr {Target Hook} bool TARGET_FORCE_AT_COMP_DIR
9493 True if the @code{DW_AT_comp_dir} attribute should be emitted for each compilation unit. This attribute is required for the darwin linker to emit debug information.
9496 @deftypevr {Target Hook} bool TARGET_DELAY_SCHED2
9497 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.
9500 @deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK
9501 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.
9504 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9505 A C statement to issue assembly directives that create a difference
9506 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9509 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9510 A C statement to issue assembly directives that create a difference
9511 between the two given labels in system defined units, e.g. instruction
9512 slots on IA64 VMS, using an integer of the given size.
9515 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9516 A C statement to issue assembly directives that create a
9517 section-relative reference to the given @var{label}, using an integer of the
9518 given @var{size}. The label is known to be defined in the given @var{section}.
9521 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9522 A C statement to issue assembly directives that create a self-relative
9523 reference to the given @var{label}, using an integer of the given @var{size}.
9526 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9527 A C statement to issue assembly directives that create a reference to
9528 the DWARF table identifier @var{label} from the current section. This
9529 is used on some systems to avoid garbage collecting a DWARF table which
9530 is referenced by a function.
9533 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
9534 If defined, this target hook is a function which outputs a DTP-relative
9535 reference to the given TLS symbol of the specified size.
9538 @defmac PUT_SDB_@dots{}
9539 Define these macros to override the assembler syntax for the special
9540 SDB assembler directives. See @file{sdbout.c} for a list of these
9541 macros and their arguments. If the standard syntax is used, you need
9542 not define them yourself.
9546 Some assemblers do not support a semicolon as a delimiter, even between
9547 SDB assembler directives. In that case, define this macro to be the
9548 delimiter to use (usually @samp{\n}). It is not necessary to define
9549 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9553 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9554 Define this macro to allow references to unknown structure,
9555 union, or enumeration tags to be emitted. Standard COFF does not
9556 allow handling of unknown references, MIPS ECOFF has support for
9560 @defmac SDB_ALLOW_FORWARD_REFERENCES
9561 Define this macro to allow references to structure, union, or
9562 enumeration tags that have not yet been seen to be handled. Some
9563 assemblers choke if forward tags are used, while some require it.
9566 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9567 A C statement to output SDB debugging information before code for line
9568 number @var{line} of the current source file to the stdio stream
9569 @var{stream}. The default is to emit an @code{.ln} directive.
9574 @subsection Macros for VMS Debug Format
9576 @c prevent bad page break with this line
9577 Here are macros for VMS debug format.
9579 @defmac VMS_DEBUGGING_INFO
9580 Define this macro if GCC should produce debugging output for VMS
9581 in response to the @option{-g} option. The default behavior for VMS
9582 is to generate minimal debug info for a traceback in the absence of
9583 @option{-g} unless explicitly overridden with @option{-g0}. This
9584 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9585 @code{TARGET_OPTION_OVERRIDE}.
9588 @node Floating Point
9589 @section Cross Compilation and Floating Point
9590 @cindex cross compilation and floating point
9591 @cindex floating point and cross compilation
9593 While all modern machines use twos-complement representation for integers,
9594 there are a variety of representations for floating point numbers. This
9595 means that in a cross-compiler the representation of floating point numbers
9596 in the compiled program may be different from that used in the machine
9597 doing the compilation.
9599 Because different representation systems may offer different amounts of
9600 range and precision, all floating point constants must be represented in
9601 the target machine's format. Therefore, the cross compiler cannot
9602 safely use the host machine's floating point arithmetic; it must emulate
9603 the target's arithmetic. To ensure consistency, GCC always uses
9604 emulation to work with floating point values, even when the host and
9605 target floating point formats are identical.
9607 The following macros are provided by @file{real.h} for the compiler to
9608 use. All parts of the compiler which generate or optimize
9609 floating-point calculations must use these macros. They may evaluate
9610 their operands more than once, so operands must not have side effects.
9612 @defmac REAL_VALUE_TYPE
9613 The C data type to be used to hold a floating point value in the target
9614 machine's format. Typically this is a @code{struct} containing an
9615 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9619 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9620 Compares for equality the two values, @var{x} and @var{y}. If the target
9621 floating point format supports negative zeroes and/or NaNs,
9622 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9623 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9626 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9627 Tests whether @var{x} is less than @var{y}.
9630 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9631 Truncates @var{x} to a signed integer, rounding toward zero.
9634 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9635 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9636 @var{x} is negative, returns zero.
9639 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9640 Converts @var{string} into a floating point number in the target machine's
9641 representation for mode @var{mode}. This routine can handle both
9642 decimal and hexadecimal floating point constants, using the syntax
9643 defined by the C language for both.
9646 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9647 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9650 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9651 Determines whether @var{x} represents infinity (positive or negative).
9654 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9655 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9658 @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})
9659 Calculates an arithmetic operation on the two floating point values
9660 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9663 The operation to be performed is specified by @var{code}. Only the
9664 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9665 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9667 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9668 target's floating point format cannot represent infinity, it will call
9669 @code{abort}. Callers should check for this situation first, using
9670 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9673 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9674 Returns the negative of the floating point value @var{x}.
9677 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9678 Returns the absolute value of @var{x}.
9681 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9682 Truncates the floating point value @var{x} to fit in @var{mode}. The
9683 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9684 appropriate bit pattern to be output as a floating constant whose
9685 precision accords with mode @var{mode}.
9688 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9689 Converts a floating point value @var{x} into a double-precision integer
9690 which is then stored into @var{low} and @var{high}. If the value is not
9691 integral, it is truncated.
9694 @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})
9695 Converts a double-precision integer found in @var{low} and @var{high},
9696 into a floating point value which is then stored into @var{x}. The
9697 value is truncated to fit in mode @var{mode}.
9700 @node Mode Switching
9701 @section Mode Switching Instructions
9702 @cindex mode switching
9703 The following macros control mode switching optimizations:
9705 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9706 Define this macro if the port needs extra instructions inserted for mode
9707 switching in an optimizing compilation.
9709 For an example, the SH4 can perform both single and double precision
9710 floating point operations, but to perform a single precision operation,
9711 the FPSCR PR bit has to be cleared, while for a double precision
9712 operation, this bit has to be set. Changing the PR bit requires a general
9713 purpose register as a scratch register, hence these FPSCR sets have to
9714 be inserted before reload, i.e.@: you can't put this into instruction emitting
9715 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9717 You can have multiple entities that are mode-switched, and select at run time
9718 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9719 return nonzero for any @var{entity} that needs mode-switching.
9720 If you define this macro, you also have to define
9721 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9722 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9723 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9727 @defmac NUM_MODES_FOR_MODE_SWITCHING
9728 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9729 initializer for an array of integers. Each initializer element
9730 N refers to an entity that needs mode switching, and specifies the number
9731 of different modes that might need to be set for this entity.
9732 The position of the initializer in the initializer---starting counting at
9733 zero---determines the integer that is used to refer to the mode-switched
9735 In macros that take mode arguments / yield a mode result, modes are
9736 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9737 switch is needed / supplied.
9740 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9741 @var{entity} is an integer specifying a mode-switched entity. If
9742 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9743 return an integer value not larger than the corresponding element in
9744 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9745 be switched into prior to the execution of @var{insn}.
9748 @defmac MODE_AFTER (@var{mode}, @var{insn})
9749 If this macro is defined, it is evaluated for every @var{insn} during
9750 mode switching. It determines the mode that an insn results in (if
9751 different from the incoming mode).
9754 @defmac MODE_ENTRY (@var{entity})
9755 If this macro is defined, it is evaluated for every @var{entity} that needs
9756 mode switching. It should evaluate to an integer, which is a mode that
9757 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9758 is defined then @code{MODE_EXIT} must be defined.
9761 @defmac MODE_EXIT (@var{entity})
9762 If this macro is defined, it is evaluated for every @var{entity} that needs
9763 mode switching. It should evaluate to an integer, which is a mode that
9764 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9765 is defined then @code{MODE_ENTRY} must be defined.
9768 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9769 This macro specifies the order in which modes for @var{entity} are processed.
9770 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9771 lowest. The value of the macro should be an integer designating a mode
9772 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9773 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9774 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9777 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9778 Generate one or more insns to set @var{entity} to @var{mode}.
9779 @var{hard_reg_live} is the set of hard registers live at the point where
9780 the insn(s) are to be inserted.
9783 @node Target Attributes
9784 @section Defining target-specific uses of @code{__attribute__}
9785 @cindex target attributes
9786 @cindex machine attributes
9787 @cindex attributes, target-specific
9789 Target-specific attributes may be defined for functions, data and types.
9790 These are described using the following target hooks; they also need to
9791 be documented in @file{extend.texi}.
9793 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9794 If defined, this target hook points to an array of @samp{struct
9795 attribute_spec} (defined in @file{tree.h}) specifying the machine
9796 specific attributes for this target and some of the restrictions on the
9797 entities to which these attributes are applied and the arguments they
9801 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
9802 If defined, this target hook is a function which returns true if the
9803 machine-specific attribute named @var{name} expects an identifier
9804 given as its first argument to be passed on as a plain identifier, not
9805 subjected to name lookup. If this is not defined, the default is
9806 false for all machine-specific attributes.
9809 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
9810 If defined, this target hook is a function which returns zero if the attributes on
9811 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9812 and two if they are nearly compatible (which causes a warning to be
9813 generated). If this is not defined, machine-specific attributes are
9814 supposed always to be compatible.
9817 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9818 If defined, this target hook is a function which assigns default attributes to
9819 the newly defined @var{type}.
9822 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9823 Define this target hook if the merging of type attributes needs special
9824 handling. If defined, the result is a list of the combined
9825 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9826 that @code{comptypes} has already been called and returned 1. This
9827 function may call @code{merge_attributes} to handle machine-independent
9831 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9832 Define this target hook if the merging of decl attributes needs special
9833 handling. If defined, the result is a list of the combined
9834 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9835 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9836 when this is needed are when one attribute overrides another, or when an
9837 attribute is nullified by a subsequent definition. This function may
9838 call @code{merge_attributes} to handle machine-independent merging.
9840 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9841 If the only target-specific handling you require is @samp{dllimport}
9842 for Microsoft Windows targets, you should define the macro
9843 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9844 will then define a function called
9845 @code{merge_dllimport_decl_attributes} which can then be defined as
9846 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9847 add @code{handle_dll_attribute} in the attribute table for your port
9848 to perform initial processing of the @samp{dllimport} and
9849 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9850 @file{i386/i386.c}, for example.
9853 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
9854 @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}.
9857 @defmac TARGET_DECLSPEC
9858 Define this macro to a nonzero value if you want to treat
9859 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9860 default, this behavior is enabled only for targets that define
9861 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9862 of @code{__declspec} is via a built-in macro, but you should not rely
9863 on this implementation detail.
9866 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9867 Define this target hook if you want to be able to add attributes to a decl
9868 when it is being created. This is normally useful for back ends which
9869 wish to implement a pragma by using the attributes which correspond to
9870 the pragma's effect. The @var{node} argument is the decl which is being
9871 created. The @var{attr_ptr} argument is a pointer to the attribute list
9872 for this decl. The list itself should not be modified, since it may be
9873 shared with other decls, but attributes may be chained on the head of
9874 the list and @code{*@var{attr_ptr}} modified to point to the new
9875 attributes, or a copy of the list may be made if further changes are
9879 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
9881 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9882 into the current function, despite its having target-specific
9883 attributes, @code{false} otherwise. By default, if a function has a
9884 target specific attribute attached to it, it will not be inlined.
9887 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9888 This hook is called to parse the @code{attribute(option("..."))}, and
9889 it allows the function to set different target machine compile time
9890 options for the current function that might be different than the
9891 options specified on the command line. The hook should return
9892 @code{true} if the options are valid.
9894 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9895 the function declaration to hold a pointer to a target specific
9896 @var{struct cl_target_option} structure.
9899 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr})
9900 This hook is called to save any additional target specific information
9901 in the @var{struct cl_target_option} structure for function specific
9903 @xref{Option file format}.
9906 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct cl_target_option *@var{ptr})
9907 This hook is called to restore any additional target specific
9908 information in the @var{struct cl_target_option} structure for
9909 function specific options.
9912 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
9913 This hook is called to print any additional target specific
9914 information in the @var{struct cl_target_option} structure for
9915 function specific options.
9918 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
9919 This target hook parses the options for @code{#pragma GCC option} to
9920 set the machine specific options for functions that occur later in the
9921 input stream. The options should be the same as handled by the
9922 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9925 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
9926 Sometimes certain combinations of command options do not make sense on
9927 a particular target machine. You can override the hook
9928 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9929 once just after all the command options have been parsed.
9931 Don't use this hook to turn on various extra optimizations for
9932 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9934 If you need to do something whenever the optimization level is
9935 changed via the optimize attribute or pragma, see
9936 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9939 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
9940 This target hook returns @code{false} if the @var{caller} function
9941 cannot inline @var{callee}, based on target specific information. By
9942 default, inlining is not allowed if the callee function has function
9943 specific target options and the caller does not use the same options.
9947 @section Emulating TLS
9948 @cindex Emulated TLS
9950 For targets whose psABI does not provide Thread Local Storage via
9951 specific relocations and instruction sequences, an emulation layer is
9952 used. A set of target hooks allows this emulation layer to be
9953 configured for the requirements of a particular target. For instance
9954 the psABI may in fact specify TLS support in terms of an emulation
9957 The emulation layer works by creating a control object for every TLS
9958 object. To access the TLS object, a lookup function is provided
9959 which, when given the address of the control object, will return the
9960 address of the current thread's instance of the TLS object.
9962 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9963 Contains the name of the helper function that uses a TLS control
9964 object to locate a TLS instance. The default causes libgcc's
9965 emulated TLS helper function to be used.
9968 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9969 Contains the name of the helper function that should be used at
9970 program startup to register TLS objects that are implicitly
9971 initialized to zero. If this is @code{NULL}, all TLS objects will
9972 have explicit initializers. The default causes libgcc's emulated TLS
9973 registration function to be used.
9976 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9977 Contains the name of the section in which TLS control variables should
9978 be placed. The default of @code{NULL} allows these to be placed in
9982 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9983 Contains the name of the section in which TLS initializers should be
9984 placed. The default of @code{NULL} allows these to be placed in any
9988 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9989 Contains the prefix to be prepended to TLS control variable names.
9990 The default of @code{NULL} uses a target-specific prefix.
9993 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9994 Contains the prefix to be prepended to TLS initializer objects. The
9995 default of @code{NULL} uses a target-specific prefix.
9998 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
9999 Specifies a function that generates the FIELD_DECLs for a TLS control
10000 object type. @var{type} is the RECORD_TYPE the fields are for and
10001 @var{name} should be filled with the structure tag, if the default of
10002 @code{__emutls_object} is unsuitable. The default creates a type suitable
10003 for libgcc's emulated TLS function.
10006 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
10007 Specifies a function that generates the CONSTRUCTOR to initialize a
10008 TLS control object. @var{var} is the TLS control object, @var{decl}
10009 is the TLS object and @var{tmpl_addr} is the address of the
10010 initializer. The default initializes libgcc's emulated TLS control object.
10013 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
10014 Specifies whether the alignment of TLS control variable objects is
10015 fixed and should not be increased as some backends may do to optimize
10016 single objects. The default is false.
10019 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
10020 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
10021 may be used to describe emulated TLS control objects.
10024 @node MIPS Coprocessors
10025 @section Defining coprocessor specifics for MIPS targets.
10026 @cindex MIPS coprocessor-definition macros
10028 The MIPS specification allows MIPS implementations to have as many as 4
10029 coprocessors, each with as many as 32 private registers. GCC supports
10030 accessing these registers and transferring values between the registers
10031 and memory using asm-ized variables. For example:
10034 register unsigned int cp0count asm ("c0r1");
10040 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
10041 names may be added as described below, or the default names may be
10042 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
10044 Coprocessor registers are assumed to be epilogue-used; sets to them will
10045 be preserved even if it does not appear that the register is used again
10046 later in the function.
10048 Another note: according to the MIPS spec, coprocessor 1 (if present) is
10049 the FPU@. One accesses COP1 registers through standard mips
10050 floating-point support; they are not included in this mechanism.
10052 There is one macro used in defining the MIPS coprocessor interface which
10053 you may want to override in subtargets; it is described below.
10055 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
10056 A comma-separated list (with leading comma) of pairs describing the
10057 alternate names of coprocessor registers. The format of each entry should be
10059 @{ @var{alternatename}, @var{register_number}@}
10065 @section Parameters for Precompiled Header Validity Checking
10066 @cindex parameters, precompiled headers
10068 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
10069 This hook returns a pointer to the data needed by
10070 @code{TARGET_PCH_VALID_P} and sets
10071 @samp{*@var{sz}} to the size of the data in bytes.
10074 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
10075 This hook checks whether the options used to create a PCH file are
10076 compatible with the current settings. It returns @code{NULL}
10077 if so and a suitable error message if not. Error messages will
10078 be presented to the user and must be localized using @samp{_(@var{msg})}.
10080 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
10081 when the PCH file was created and @var{sz} is the size of that data in bytes.
10082 It's safe to assume that the data was created by the same version of the
10083 compiler, so no format checking is needed.
10085 The default definition of @code{default_pch_valid_p} should be
10086 suitable for most targets.
10089 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
10090 If this hook is nonnull, the default implementation of
10091 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
10092 of @code{target_flags}. @var{pch_flags} specifies the value that
10093 @code{target_flags} had when the PCH file was created. The return
10094 value is the same as for @code{TARGET_PCH_VALID_P}.
10097 @deftypefn {Target Hook} void TARGET_PREPARE_PCH_SAVE (void)
10098 Called before writing out a PCH file. If the target has some
10099 garbage-collected data that needs to be in a particular state on PCH loads,
10100 it can use this hook to enforce that state. Very few targets need
10101 to do anything here.
10105 @section C++ ABI parameters
10106 @cindex parameters, c++ abi
10108 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10109 Define this hook to override the integer type used for guard variables.
10110 These are used to implement one-time construction of static objects. The
10111 default is long_long_integer_type_node.
10114 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10115 This hook determines how guard variables are used. It should return
10116 @code{false} (the default) if the first byte should be used. A return value of
10117 @code{true} indicates that only the least significant bit should be used.
10120 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10121 This hook returns the size of the cookie to use when allocating an array
10122 whose elements have the indicated @var{type}. Assumes that it is already
10123 known that a cookie is needed. The default is
10124 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10125 IA64/Generic C++ ABI@.
10128 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10129 This hook should return @code{true} if the element size should be stored in
10130 array cookies. The default is to return @code{false}.
10133 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10134 If defined by a backend this hook allows the decision made to export
10135 class @var{type} to be overruled. Upon entry @var{import_export}
10136 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10137 to be imported and 0 otherwise. This function should return the
10138 modified value and perform any other actions necessary to support the
10139 backend's targeted operating system.
10142 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10143 This hook should return @code{true} if constructors and destructors return
10144 the address of the object created/destroyed. The default is to return
10148 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10149 This hook returns true if the key method for a class (i.e., the method
10150 which, if defined in the current translation unit, causes the virtual
10151 table to be emitted) may be an inline function. Under the standard
10152 Itanium C++ ABI the key method may be an inline function so long as
10153 the function is not declared inline in the class definition. Under
10154 some variants of the ABI, an inline function can never be the key
10155 method. The default is to return @code{true}.
10158 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10159 @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}.
10162 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10163 This hook returns true (the default) if virtual tables and other
10164 similar implicit class data objects are always COMDAT if they have
10165 external linkage. If this hook returns false, then class data for
10166 classes whose virtual table will be emitted in only one translation
10167 unit will not be COMDAT.
10170 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10171 This hook returns true (the default) if the RTTI information for
10172 the basic types which is defined in the C++ runtime should always
10173 be COMDAT, false if it should not be COMDAT.
10176 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10177 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10178 should be used to register static destructors when @option{-fuse-cxa-atexit}
10179 is in effect. The default is to return false to use @code{__cxa_atexit}.
10182 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10183 This hook returns true if the target @code{atexit} function can be used
10184 in the same manner as @code{__cxa_atexit} to register C++ static
10185 destructors. This requires that @code{atexit}-registered functions in
10186 shared libraries are run in the correct order when the libraries are
10187 unloaded. The default is to return false.
10190 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10191 @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).
10194 @deftypefn {Target Hook} tree TARGET_CXX_DECL_MANGLING_CONTEXT (const_tree @var{decl})
10195 Return target-specific mangling context of @var{decl} or @code{NULL_TREE}.
10198 @node Named Address Spaces
10199 @section Adding support for named address spaces
10200 @cindex named address spaces
10202 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10203 standards committee, @cite{Programming Languages - C - Extensions to
10204 support embedded processors}, specifies a syntax for embedded
10205 processors to specify alternate address spaces. You can configure a
10206 GCC port to support section 5.1 of the draft report to add support for
10207 address spaces other than the default address space. These address
10208 spaces are new keywords that are similar to the @code{volatile} and
10209 @code{const} type attributes.
10211 Pointers to named address spaces can have a different size than
10212 pointers to the generic address space.
10214 For example, the SPU port uses the @code{__ea} address space to refer
10215 to memory in the host processor, rather than memory local to the SPU
10216 processor. Access to memory in the @code{__ea} address space involves
10217 issuing DMA operations to move data between the host processor and the
10218 local processor memory address space. Pointers in the @code{__ea}
10219 address space are either 32 bits or 64 bits based on the
10220 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10223 Internally, address spaces are represented as a small integer in the
10224 range 0 to 15 with address space 0 being reserved for the generic
10227 To register a named address space qualifier keyword with the C front end,
10228 the target may call the @code{c_register_addr_space} routine. For example,
10229 the SPU port uses the following to declare @code{__ea} as the keyword for
10230 named address space #1:
10232 #define ADDR_SPACE_EA 1
10233 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10236 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10237 Define this to return the machine mode to use for pointers to
10238 @var{address_space} if the target supports named address spaces.
10239 The default version of this hook returns @code{ptr_mode} for the
10240 generic address space only.
10243 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10244 Define this to return the machine mode to use for addresses in
10245 @var{address_space} if the target supports named address spaces.
10246 The default version of this hook returns @code{Pmode} for the
10247 generic address space only.
10250 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (enum machine_mode @var{mode}, addr_space_t @var{as})
10251 Define this to return nonzero if the port can handle pointers
10252 with machine mode @var{mode} to address space @var{as}. This target
10253 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10254 except that it includes explicit named address space support. The default
10255 version of this hook returns true for the modes returned by either the
10256 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10257 target hooks for the given address space.
10260 @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})
10261 Define this to return true if @var{exp} is a valid address for mode
10262 @var{mode} in the named address space @var{as}. The @var{strict}
10263 parameter says whether strict addressing is in effect after reload has
10264 finished. This target hook is the same as the
10265 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10266 explicit named address space support.
10269 @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})
10270 Define this to modify an invalid address @var{x} to be a valid address
10271 with mode @var{mode} in the named address space @var{as}. This target
10272 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10273 except that it includes explicit named address space support.
10276 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{subset}, addr_space_t @var{superset})
10277 Define this to return whether the @var{subset} named address space is
10278 contained within the @var{superset} named address space. Pointers to
10279 a named address space that is a subset of another named address space
10280 will be converted automatically without a cast if used together in
10281 arithmetic operations. Pointers to a superset address space can be
10282 converted to pointers to a subset address space via explicit casts.
10285 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10286 Define this to convert the pointer expression represented by the RTL
10287 @var{op} with type @var{from_type} that points to a named address
10288 space to a new pointer expression with type @var{to_type} that points
10289 to a different named address space. When this hook it called, it is
10290 guaranteed that one of the two address spaces is a subset of the other,
10291 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10295 @section Miscellaneous Parameters
10296 @cindex parameters, miscellaneous
10298 @c prevent bad page break with this line
10299 Here are several miscellaneous parameters.
10301 @defmac HAS_LONG_COND_BRANCH
10302 Define this boolean macro to indicate whether or not your architecture
10303 has conditional branches that can span all of memory. It is used in
10304 conjunction with an optimization that partitions hot and cold basic
10305 blocks into separate sections of the executable. If this macro is
10306 set to false, gcc will convert any conditional branches that attempt
10307 to cross between sections into unconditional branches or indirect jumps.
10310 @defmac HAS_LONG_UNCOND_BRANCH
10311 Define this boolean macro to indicate whether or not your architecture
10312 has unconditional branches that can span all of memory. It is used in
10313 conjunction with an optimization that partitions hot and cold basic
10314 blocks into separate sections of the executable. If this macro is
10315 set to false, gcc will convert any unconditional branches that attempt
10316 to cross between sections into indirect jumps.
10319 @defmac CASE_VECTOR_MODE
10320 An alias for a machine mode name. This is the machine mode that
10321 elements of a jump-table should have.
10324 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10325 Optional: return the preferred mode for an @code{addr_diff_vec}
10326 when the minimum and maximum offset are known. If you define this,
10327 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10328 To make this work, you also have to define @code{INSN_ALIGN} and
10329 make the alignment for @code{addr_diff_vec} explicit.
10330 The @var{body} argument is provided so that the offset_unsigned and scale
10331 flags can be updated.
10334 @defmac CASE_VECTOR_PC_RELATIVE
10335 Define this macro to be a C expression to indicate when jump-tables
10336 should contain relative addresses. You need not define this macro if
10337 jump-tables never contain relative addresses, or jump-tables should
10338 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10342 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10343 This function return the smallest number of different values for which it
10344 is best to use a jump-table instead of a tree of conditional branches.
10345 The default is four for machines with a @code{casesi} instruction and
10346 five otherwise. This is best for most machines.
10349 @defmac CASE_USE_BIT_TESTS
10350 Define this macro to be a C expression to indicate whether C switch
10351 statements may be implemented by a sequence of bit tests. This is
10352 advantageous on processors that can efficiently implement left shift
10353 of 1 by the number of bits held in a register, but inappropriate on
10354 targets that would require a loop. By default, this macro returns
10355 @code{true} if the target defines an @code{ashlsi3} pattern, and
10356 @code{false} otherwise.
10359 @defmac WORD_REGISTER_OPERATIONS
10360 Define this macro if operations between registers with integral mode
10361 smaller than a word are always performed on the entire register.
10362 Most RISC machines have this property and most CISC machines do not.
10365 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10366 Define this macro to be a C expression indicating when insns that read
10367 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10368 bits outside of @var{mem_mode} to be either the sign-extension or the
10369 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10370 of @var{mem_mode} for which the
10371 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10372 @code{UNKNOWN} for other modes.
10374 This macro is not called with @var{mem_mode} non-integral or with a width
10375 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10376 value in this case. Do not define this macro if it would always return
10377 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10378 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10380 You may return a non-@code{UNKNOWN} value even if for some hard registers
10381 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10382 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10383 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10384 integral mode larger than this but not larger than @code{word_mode}.
10386 You must return @code{UNKNOWN} if for some hard registers that allow this
10387 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10388 @code{word_mode}, but that they can change to another integral mode that
10389 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10392 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10393 Define this macro if loading short immediate values into registers sign
10397 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
10398 Define this macro if the same instructions that convert a floating
10399 point number to a signed fixed point number also convert validly to an
10403 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
10404 When @option{-ffast-math} is in effect, GCC tries to optimize
10405 divisions by the same divisor, by turning them into multiplications by
10406 the reciprocal. This target hook specifies the minimum number of divisions
10407 that should be there for GCC to perform the optimization for a variable
10408 of mode @var{mode}. The default implementation returns 3 if the machine
10409 has an instruction for the division, and 2 if it does not.
10413 The maximum number of bytes that a single instruction can move quickly
10414 between memory and registers or between two memory locations.
10417 @defmac MAX_MOVE_MAX
10418 The maximum number of bytes that a single instruction can move quickly
10419 between memory and registers or between two memory locations. If this
10420 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10421 constant value that is the largest value that @code{MOVE_MAX} can have
10425 @defmac SHIFT_COUNT_TRUNCATED
10426 A C expression that is nonzero if on this machine the number of bits
10427 actually used for the count of a shift operation is equal to the number
10428 of bits needed to represent the size of the object being shifted. When
10429 this macro is nonzero, the compiler will assume that it is safe to omit
10430 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10431 truncates the count of a shift operation. On machines that have
10432 instructions that act on bit-fields at variable positions, which may
10433 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10434 also enables deletion of truncations of the values that serve as
10435 arguments to bit-field instructions.
10437 If both types of instructions truncate the count (for shifts) and
10438 position (for bit-field operations), or if no variable-position bit-field
10439 instructions exist, you should define this macro.
10441 However, on some machines, such as the 80386 and the 680x0, truncation
10442 only applies to shift operations and not the (real or pretended)
10443 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10444 such machines. Instead, add patterns to the @file{md} file that include
10445 the implied truncation of the shift instructions.
10447 You need not define this macro if it would always have the value of zero.
10450 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10451 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
10452 This function describes how the standard shift patterns for @var{mode}
10453 deal with shifts by negative amounts or by more than the width of the mode.
10454 @xref{shift patterns}.
10456 On many machines, the shift patterns will apply a mask @var{m} to the
10457 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10458 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10459 this is true for mode @var{mode}, the function should return @var{m},
10460 otherwise it should return 0. A return value of 0 indicates that no
10461 particular behavior is guaranteed.
10463 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10464 @emph{not} apply to general shift rtxes; it applies only to instructions
10465 that are generated by the named shift patterns.
10467 The default implementation of this function returns
10468 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10469 and 0 otherwise. This definition is always safe, but if
10470 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10471 nevertheless truncate the shift count, you may get better code
10475 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10476 A C expression which is nonzero if on this machine it is safe to
10477 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10478 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10479 operating on it as if it had only @var{outprec} bits.
10481 On many machines, this expression can be 1.
10483 @c rearranged this, removed the phrase "it is reported that". this was
10484 @c to fix an overfull hbox. --mew 10feb93
10485 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10486 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10487 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10488 such cases may improve things.
10491 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
10492 The representation of an integral mode can be such that the values
10493 are always extended to a wider integral mode. Return
10494 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10495 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10496 otherwise. (Currently, none of the targets use zero-extended
10497 representation this way so unlike @code{LOAD_EXTEND_OP},
10498 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10499 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10500 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10501 widest integral mode and currently we take advantage of this fact.)
10503 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10504 value even if the extension is not performed on certain hard registers
10505 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10506 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10508 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10509 describe two related properties. If you define
10510 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10511 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10514 In order to enforce the representation of @code{mode},
10515 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10519 @defmac STORE_FLAG_VALUE
10520 A C expression describing the value returned by a comparison operator
10521 with an integral mode and stored by a store-flag instruction
10522 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10523 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10524 comparison operators whose results have a @code{MODE_INT} mode.
10526 A value of 1 or @minus{}1 means that the instruction implementing the
10527 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10528 and 0 when the comparison is false. Otherwise, the value indicates
10529 which bits of the result are guaranteed to be 1 when the comparison is
10530 true. This value is interpreted in the mode of the comparison
10531 operation, which is given by the mode of the first operand in the
10532 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10533 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10536 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10537 generate code that depends only on the specified bits. It can also
10538 replace comparison operators with equivalent operations if they cause
10539 the required bits to be set, even if the remaining bits are undefined.
10540 For example, on a machine whose comparison operators return an
10541 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10542 @samp{0x80000000}, saying that just the sign bit is relevant, the
10546 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10550 can be converted to
10553 (ashift:SI @var{x} (const_int @var{n}))
10557 where @var{n} is the appropriate shift count to move the bit being
10558 tested into the sign bit.
10560 There is no way to describe a machine that always sets the low-order bit
10561 for a true value, but does not guarantee the value of any other bits,
10562 but we do not know of any machine that has such an instruction. If you
10563 are trying to port GCC to such a machine, include an instruction to
10564 perform a logical-and of the result with 1 in the pattern for the
10565 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10567 Often, a machine will have multiple instructions that obtain a value
10568 from a comparison (or the condition codes). Here are rules to guide the
10569 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10574 Use the shortest sequence that yields a valid definition for
10575 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10576 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10577 comparison operators to do so because there may be opportunities to
10578 combine the normalization with other operations.
10581 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10582 slightly preferred on machines with expensive jumps and 1 preferred on
10586 As a second choice, choose a value of @samp{0x80000001} if instructions
10587 exist that set both the sign and low-order bits but do not define the
10591 Otherwise, use a value of @samp{0x80000000}.
10594 Many machines can produce both the value chosen for
10595 @code{STORE_FLAG_VALUE} and its negation in the same number of
10596 instructions. On those machines, you should also define a pattern for
10597 those cases, e.g., one matching
10600 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10603 Some machines can also perform @code{and} or @code{plus} operations on
10604 condition code values with less instructions than the corresponding
10605 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10606 machines, define the appropriate patterns. Use the names @code{incscc}
10607 and @code{decscc}, respectively, for the patterns which perform
10608 @code{plus} or @code{minus} operations on condition code values. See
10609 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10610 find such instruction sequences on other machines.
10612 If this macro is not defined, the default value, 1, is used. You need
10613 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10614 instructions, or if the value generated by these instructions is 1.
10617 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10618 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10619 returned when comparison operators with floating-point results are true.
10620 Define this macro on machines that have comparison operations that return
10621 floating-point values. If there are no such operations, do not define
10625 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10626 A C expression that gives a rtx representing the nonzero true element
10627 for vector comparisons. The returned rtx should be valid for the inner
10628 mode of @var{mode} which is guaranteed to be a vector mode. Define
10629 this macro on machines that have vector comparison operations that
10630 return a vector result. If there are no such operations, do not define
10631 this macro. Typically, this macro is defined as @code{const1_rtx} or
10632 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10633 the compiler optimizing such vector comparison operations for the
10637 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10638 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10639 A C expression that indicates whether the architecture defines a value
10640 for @code{clz} or @code{ctz} with a zero operand.
10641 A result of @code{0} indicates the value is undefined.
10642 If the value is defined for only the RTL expression, the macro should
10643 evaluate to @code{1}; if the value applies also to the corresponding optab
10644 entry (which is normally the case if it expands directly into
10645 the corresponding RTL), then the macro should evaluate to @code{2}.
10646 In the cases where the value is defined, @var{value} should be set to
10649 If this macro is not defined, the value of @code{clz} or
10650 @code{ctz} at zero is assumed to be undefined.
10652 This macro must be defined if the target's expansion for @code{ffs}
10653 relies on a particular value to get correct results. Otherwise it
10654 is not necessary, though it may be used to optimize some corner cases, and
10655 to provide a default expansion for the @code{ffs} optab.
10657 Note that regardless of this macro the ``definedness'' of @code{clz}
10658 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10659 visible to the user. Thus one may be free to adjust the value at will
10660 to match the target expansion of these operations without fear of
10665 An alias for the machine mode for pointers. On most machines, define
10666 this to be the integer mode corresponding to the width of a hardware
10667 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10668 On some machines you must define this to be one of the partial integer
10669 modes, such as @code{PSImode}.
10671 The width of @code{Pmode} must be at least as large as the value of
10672 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10673 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10677 @defmac FUNCTION_MODE
10678 An alias for the machine mode used for memory references to functions
10679 being called, in @code{call} RTL expressions. On most CISC machines,
10680 where an instruction can begin at any byte address, this should be
10681 @code{QImode}. On most RISC machines, where all instructions have fixed
10682 size and alignment, this should be a mode with the same size and alignment
10683 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10686 @defmac STDC_0_IN_SYSTEM_HEADERS
10687 In normal operation, the preprocessor expands @code{__STDC__} to the
10688 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10689 hosts, like Solaris, the system compiler uses a different convention,
10690 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10691 strict conformance to the C Standard.
10693 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10694 convention when processing system header files, but when processing user
10695 files @code{__STDC__} will always expand to 1.
10698 @defmac NO_IMPLICIT_EXTERN_C
10699 Define this macro if the system header files support C++ as well as C@.
10700 This macro inhibits the usual method of using system header files in
10701 C++, which is to pretend that the file's contents are enclosed in
10702 @samp{extern "C" @{@dots{}@}}.
10707 @defmac REGISTER_TARGET_PRAGMAS ()
10708 Define this macro if you want to implement any target-specific pragmas.
10709 If defined, it is a C expression which makes a series of calls to
10710 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10711 for each pragma. The macro may also do any
10712 setup required for the pragmas.
10714 The primary reason to define this macro is to provide compatibility with
10715 other compilers for the same target. In general, we discourage
10716 definition of target-specific pragmas for GCC@.
10718 If the pragma can be implemented by attributes then you should consider
10719 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10721 Preprocessor macros that appear on pragma lines are not expanded. All
10722 @samp{#pragma} directives that do not match any registered pragma are
10723 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10726 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10727 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10729 Each call to @code{c_register_pragma} or
10730 @code{c_register_pragma_with_expansion} establishes one pragma. The
10731 @var{callback} routine will be called when the preprocessor encounters a
10735 #pragma [@var{space}] @var{name} @dots{}
10738 @var{space} is the case-sensitive namespace of the pragma, or
10739 @code{NULL} to put the pragma in the global namespace. The callback
10740 routine receives @var{pfile} as its first argument, which can be passed
10741 on to cpplib's functions if necessary. You can lex tokens after the
10742 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10743 callback will be silently ignored. The end of the line is indicated by
10744 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10745 arguments of pragmas registered with
10746 @code{c_register_pragma_with_expansion} but not on the arguments of
10747 pragmas registered with @code{c_register_pragma}.
10749 Note that the use of @code{pragma_lex} is specific to the C and C++
10750 compilers. It will not work in the Java or Fortran compilers, or any
10751 other language compilers for that matter. Thus if @code{pragma_lex} is going
10752 to be called from target-specific code, it must only be done so when
10753 building the C and C++ compilers. This can be done by defining the
10754 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10755 target entry in the @file{config.gcc} file. These variables should name
10756 the target-specific, language-specific object file which contains the
10757 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10758 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10759 how to build this object file.
10762 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10763 Define this macro if macros should be expanded in the
10764 arguments of @samp{#pragma pack}.
10767 @deftypevr {Target Hook} bool TARGET_HANDLE_PRAGMA_EXTERN_PREFIX
10768 True if @code{#pragma extern_prefix} is to be supported.
10771 @defmac TARGET_DEFAULT_PACK_STRUCT
10772 If your target requires a structure packing default other than 0 (meaning
10773 the machine default), define this macro to the necessary value (in bytes).
10774 This must be a value that would also be valid to use with
10775 @samp{#pragma pack()} (that is, a small power of two).
10778 @defmac DOLLARS_IN_IDENTIFIERS
10779 Define this macro to control use of the character @samp{$} in
10780 identifier names for the C family of languages. 0 means @samp{$} is
10781 not allowed by default; 1 means it is allowed. 1 is the default;
10782 there is no need to define this macro in that case.
10785 @defmac NO_DOLLAR_IN_LABEL
10786 Define this macro if the assembler does not accept the character
10787 @samp{$} in label names. By default constructors and destructors in
10788 G++ have @samp{$} in the identifiers. If this macro is defined,
10789 @samp{.} is used instead.
10792 @defmac NO_DOT_IN_LABEL
10793 Define this macro if the assembler does not accept the character
10794 @samp{.} in label names. By default constructors and destructors in G++
10795 have names that use @samp{.}. If this macro is defined, these names
10796 are rewritten to avoid @samp{.}.
10799 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10800 Define this macro as a C expression that is nonzero if it is safe for the
10801 delay slot scheduler to place instructions in the delay slot of @var{insn},
10802 even if they appear to use a resource set or clobbered in @var{insn}.
10803 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10804 every @code{call_insn} has this behavior. On machines where some @code{insn}
10805 or @code{jump_insn} is really a function call and hence has this behavior,
10806 you should define this macro.
10808 You need not define this macro if it would always return zero.
10811 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10812 Define this macro as a C expression that is nonzero if it is safe for the
10813 delay slot scheduler to place instructions in the delay slot of @var{insn},
10814 even if they appear to set or clobber a resource referenced in @var{insn}.
10815 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10816 some @code{insn} or @code{jump_insn} is really a function call and its operands
10817 are registers whose use is actually in the subroutine it calls, you should
10818 define this macro. Doing so allows the delay slot scheduler to move
10819 instructions which copy arguments into the argument registers into the delay
10820 slot of @var{insn}.
10822 You need not define this macro if it would always return zero.
10825 @defmac MULTIPLE_SYMBOL_SPACES
10826 Define this macro as a C expression that is nonzero if, in some cases,
10827 global symbols from one translation unit may not be bound to undefined
10828 symbols in another translation unit without user intervention. For
10829 instance, under Microsoft Windows symbols must be explicitly imported
10830 from shared libraries (DLLs).
10832 You need not define this macro if it would always evaluate to zero.
10835 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10836 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10837 any hard regs the port wishes to automatically clobber for an asm.
10838 It should return the result of the last @code{tree_cons} used to add a
10839 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10840 corresponding parameters to the asm and may be inspected to avoid
10841 clobbering a register that is an input or output of the asm. You can use
10842 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10843 for overlap with regards to asm-declared registers.
10846 @defmac MATH_LIBRARY
10847 Define this macro as a C string constant for the linker argument to link
10848 in the system math library, minus the initial @samp{"-l"}, or
10849 @samp{""} if the target does not have a
10850 separate math library.
10852 You need only define this macro if the default of @samp{"m"} is wrong.
10855 @defmac LIBRARY_PATH_ENV
10856 Define this macro as a C string constant for the environment variable that
10857 specifies where the linker should look for libraries.
10859 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10863 @defmac TARGET_POSIX_IO
10864 Define this macro if the target supports the following POSIX@ file
10865 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10866 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10867 to use file locking when exiting a program, which avoids race conditions
10868 if the program has forked. It will also create directories at run-time
10869 for cross-profiling.
10872 @defmac MAX_CONDITIONAL_EXECUTE
10874 A C expression for the maximum number of instructions to execute via
10875 conditional execution instructions instead of a branch. A value of
10876 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10877 1 if it does use cc0.
10880 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10881 Used if the target needs to perform machine-dependent modifications on the
10882 conditionals used for turning basic blocks into conditionally executed code.
10883 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10884 contains information about the currently processed blocks. @var{true_expr}
10885 and @var{false_expr} are the tests that are used for converting the
10886 then-block and the else-block, respectively. Set either @var{true_expr} or
10887 @var{false_expr} to a null pointer if the tests cannot be converted.
10890 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10891 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10892 if-statements into conditions combined by @code{and} and @code{or} operations.
10893 @var{bb} contains the basic block that contains the test that is currently
10894 being processed and about to be turned into a condition.
10897 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10898 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10899 be converted to conditional execution format. @var{ce_info} points to
10900 a data structure, @code{struct ce_if_block}, which contains information
10901 about the currently processed blocks.
10904 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10905 A C expression to perform any final machine dependent modifications in
10906 converting code to conditional execution. The involved basic blocks
10907 can be found in the @code{struct ce_if_block} structure that is pointed
10908 to by @var{ce_info}.
10911 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10912 A C expression to cancel any machine dependent modifications in
10913 converting code to conditional execution. The involved basic blocks
10914 can be found in the @code{struct ce_if_block} structure that is pointed
10915 to by @var{ce_info}.
10918 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10919 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10920 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10923 @defmac IFCVT_EXTRA_FIELDS
10924 If defined, it should expand to a set of field declarations that will be
10925 added to the @code{struct ce_if_block} structure. These should be initialized
10926 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10929 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
10930 If non-null, this hook performs a target-specific pass over the
10931 instruction stream. The compiler will run it at all optimization levels,
10932 just before the point at which it normally does delayed-branch scheduling.
10934 The exact purpose of the hook varies from target to target. Some use
10935 it to do transformations that are necessary for correctness, such as
10936 laying out in-function constant pools or avoiding hardware hazards.
10937 Others use it as an opportunity to do some machine-dependent optimizations.
10939 You need not implement the hook if it has nothing to do. The default
10940 definition is null.
10943 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
10944 Define this hook if you have any machine-specific built-in functions
10945 that need to be defined. It should be a function that performs the
10948 Machine specific built-in functions can be useful to expand special machine
10949 instructions that would otherwise not normally be generated because
10950 they have no equivalent in the source language (for example, SIMD vector
10951 instructions or prefetch instructions).
10953 To create a built-in function, call the function
10954 @code{lang_hooks.builtin_function}
10955 which is defined by the language front end. You can use any type nodes set
10956 up by @code{build_common_tree_nodes};
10957 only language front ends that use those two functions will call
10958 @samp{TARGET_INIT_BUILTINS}.
10961 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
10962 Define this hook if you have any machine-specific built-in functions
10963 that need to be defined. It should be a function that returns the
10964 builtin function declaration for the builtin function code @var{code}.
10965 If there is no such builtin and it cannot be initialized at this time
10966 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10967 If @var{code} is out of range the function should return
10968 @code{error_mark_node}.
10971 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10973 Expand a call to a machine specific built-in function that was set up by
10974 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10975 function call; the result should go to @var{target} if that is
10976 convenient, and have mode @var{mode} if that is convenient.
10977 @var{subtarget} may be used as the target for computing one of
10978 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10979 ignored. This function should return the result of the call to the
10983 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
10984 Select a replacement for a machine specific built-in function that
10985 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10986 @emph{before} regular type checking, and so allows the target to
10987 implement a crude form of function overloading. @var{fndecl} is the
10988 declaration of the built-in function. @var{arglist} is the list of
10989 arguments passed to the built-in function. The result is a
10990 complete expression that implements the operation, usually
10991 another @code{CALL_EXPR}.
10992 @var{arglist} really has type @samp{VEC(tree,gc)*}
10995 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
10996 Fold a call to a machine specific built-in function that was set up by
10997 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10998 built-in function. @var{n_args} is the number of arguments passed to
10999 the function; the arguments themselves are pointed to by @var{argp}.
11000 The result is another tree containing a simplified expression for the
11001 call's result. If @var{ignore} is true the value will be ignored.
11004 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const_rtx @var{insn})
11006 Take an instruction in @var{insn} and return NULL if it is valid within a
11007 low-overhead loop, otherwise return a string explaining why doloop
11008 could not be applied.
11010 Many targets use special registers for low-overhead looping. For any
11011 instruction that clobbers these this function should return a string indicating
11012 the reason why the doloop could not be applied.
11013 By default, the RTL loop optimizer does not use a present doloop pattern for
11014 loops containing function calls or branch on table instructions.
11017 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
11019 Take a branch insn in @var{branch1} and another in @var{branch2}.
11020 Return true if redirecting @var{branch1} to the destination of
11021 @var{branch2} is possible.
11023 On some targets, branches may have a limited range. Optimizing the
11024 filling of delay slots can result in branches being redirected, and this
11025 may in turn cause a branch offset to overflow.
11028 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
11029 This target hook returns @code{true} if @var{x} is considered to be commutative.
11030 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
11031 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
11032 of the enclosing rtl, if known, otherwise it is UNKNOWN.
11035 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
11037 When the initial value of a hard register has been copied in a pseudo
11038 register, it is often not necessary to actually allocate another register
11039 to this pseudo register, because the original hard register or a stack slot
11040 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
11041 is called at the start of register allocation once for each hard register
11042 that had its initial value copied by using
11043 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
11044 Possible values are @code{NULL_RTX}, if you don't want
11045 to do any special allocation, a @code{REG} rtx---that would typically be
11046 the hard register itself, if it is known not to be clobbered---or a
11048 If you are returning a @code{MEM}, this is only a hint for the allocator;
11049 it might decide to use another register anyways.
11050 You may use @code{current_function_leaf_function} in the hook, functions
11051 that use @code{REG_N_SETS}, to determine if the hard
11052 register in question will not be clobbered.
11053 The default value of this hook is @code{NULL}, which disables any special
11057 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
11058 This target hook returns nonzero if @var{x}, an @code{unspec} or
11059 @code{unspec_volatile} operation, might cause a trap. Targets can use
11060 this hook to enhance precision of analysis for @code{unspec} and
11061 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
11062 to analyze inner elements of @var{x} in which case @var{flags} should be
11066 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
11067 The compiler invokes this hook whenever it changes its current function
11068 context (@code{cfun}). You can define this function if
11069 the back end needs to perform any initialization or reset actions on a
11070 per-function basis. For example, it may be used to implement function
11071 attributes that affect register usage or code generation patterns.
11072 The argument @var{decl} is the declaration for the new function context,
11073 and may be null to indicate that the compiler has left a function context
11074 and is returning to processing at the top level.
11075 The default hook function does nothing.
11077 GCC sets @code{cfun} to a dummy function context during initialization of
11078 some parts of the back end. The hook function is not invoked in this
11079 situation; you need not worry about the hook being invoked recursively,
11080 or when the back end is in a partially-initialized state.
11081 @code{cfun} might be @code{NULL} to indicate processing at top level,
11082 outside of any function scope.
11085 @defmac TARGET_OBJECT_SUFFIX
11086 Define this macro to be a C string representing the suffix for object
11087 files on your target machine. If you do not define this macro, GCC will
11088 use @samp{.o} as the suffix for object files.
11091 @defmac TARGET_EXECUTABLE_SUFFIX
11092 Define this macro to be a C string representing the suffix to be
11093 automatically added to executable files on your target machine. If you
11094 do not define this macro, GCC will use the null string as the suffix for
11098 @defmac COLLECT_EXPORT_LIST
11099 If defined, @code{collect2} will scan the individual object files
11100 specified on its command line and create an export list for the linker.
11101 Define this macro for systems like AIX, where the linker discards
11102 object files that are not referenced from @code{main} and uses export
11106 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11107 Define this macro to a C expression representing a variant of the
11108 method call @var{mdecl}, if Java Native Interface (JNI) methods
11109 must be invoked differently from other methods on your target.
11110 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11111 the @code{stdcall} calling convention and this macro is then
11112 defined as this expression:
11115 build_type_attribute_variant (@var{mdecl},
11117 (get_identifier ("stdcall"),
11122 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11123 This target hook returns @code{true} past the point in which new jump
11124 instructions could be created. On machines that require a register for
11125 every jump such as the SHmedia ISA of SH5, this point would typically be
11126 reload, so this target hook should be defined to a function such as:
11130 cannot_modify_jumps_past_reload_p ()
11132 return (reload_completed || reload_in_progress);
11137 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11138 This target hook returns a register class for which branch target register
11139 optimizations should be applied. All registers in this class should be
11140 usable interchangeably. After reload, registers in this class will be
11141 re-allocated and loads will be hoisted out of loops and be subjected
11142 to inter-block scheduling.
11145 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11146 Branch target register optimization will by default exclude callee-saved
11148 that are not already live during the current function; if this target hook
11149 returns true, they will be included. The target code must than make sure
11150 that all target registers in the class returned by
11151 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11152 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11153 epilogues have already been generated. Note, even if you only return
11154 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11155 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11156 to reserve space for caller-saved target registers.
11159 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11160 This target hook returns true if the target supports conditional execution.
11161 This target hook is required only when the target has several different
11162 modes and they have different conditional execution capability, such as ARM.
11165 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop})
11166 This target hook returns a new value for the number of times @var{loop}
11167 should be unrolled. The parameter @var{nunroll} is the number of times
11168 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11169 the loop, which is going to be checked for unrolling. This target hook
11170 is required only when the target has special constraints like maximum
11171 number of memory accesses.
11174 @defmac POWI_MAX_MULTS
11175 If defined, this macro is interpreted as a signed integer C expression
11176 that specifies the maximum number of floating point multiplications
11177 that should be emitted when expanding exponentiation by an integer
11178 constant inline. When this value is defined, exponentiation requiring
11179 more than this number of multiplications is implemented by calling the
11180 system library's @code{pow}, @code{powf} or @code{powl} routines.
11181 The default value places no upper bound on the multiplication count.
11184 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11185 This target hook should register any extra include files for the
11186 target. The parameter @var{stdinc} indicates if normal include files
11187 are present. The parameter @var{sysroot} is the system root directory.
11188 The parameter @var{iprefix} is the prefix for the gcc directory.
11191 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11192 This target hook should register any extra include files for the
11193 target before any standard headers. The parameter @var{stdinc}
11194 indicates if normal include files are present. The parameter
11195 @var{sysroot} is the system root directory. The parameter
11196 @var{iprefix} is the prefix for the gcc directory.
11199 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11200 This target hook should register special include paths for the target.
11201 The parameter @var{path} is the include to register. On Darwin
11202 systems, this is used for Framework includes, which have semantics
11203 that are different from @option{-I}.
11206 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11207 This target macro returns @code{true} if it is safe to use a local alias
11208 for a virtual function @var{fndecl} when constructing thunks,
11209 @code{false} otherwise. By default, the macro returns @code{true} for all
11210 functions, if a target supports aliases (i.e.@: defines
11211 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11214 @defmac TARGET_FORMAT_TYPES
11215 If defined, this macro is the name of a global variable containing
11216 target-specific format checking information for the @option{-Wformat}
11217 option. The default is to have no target-specific format checks.
11220 @defmac TARGET_N_FORMAT_TYPES
11221 If defined, this macro is the number of entries in
11222 @code{TARGET_FORMAT_TYPES}.
11225 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11226 If defined, this macro is the name of a global variable containing
11227 target-specific format overrides for the @option{-Wformat} option. The
11228 default is to have no target-specific format overrides. If defined,
11229 @code{TARGET_FORMAT_TYPES} must be defined, too.
11232 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11233 If defined, this macro specifies the number of entries in
11234 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11237 @defmac TARGET_OVERRIDES_FORMAT_INIT
11238 If defined, this macro specifies the optional initialization
11239 routine for target specific customizations of the system printf
11240 and scanf formatter settings.
11243 @deftypevr {Target Hook} bool TARGET_RELAXED_ORDERING
11244 If set to @code{true}, means that the target's memory model does not
11245 guarantee that loads which do not depend on one another will access
11246 main memory in the order of the instruction stream; if ordering is
11247 important, an explicit memory barrier must be used. This is true of
11248 many recent processors which implement a policy of ``relaxed,''
11249 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11250 and ia64. The default is @code{false}.
11253 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11254 If defined, this macro returns the diagnostic message when it is
11255 illegal to pass argument @var{val} to function @var{funcdecl}
11256 with prototype @var{typelist}.
11259 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11260 If defined, this macro returns the diagnostic message when it is
11261 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11262 if validity should be determined by the front end.
11265 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11266 If defined, this macro returns the diagnostic message when it is
11267 invalid to apply operation @var{op} (where unary plus is denoted by
11268 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11269 if validity should be determined by the front end.
11272 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11273 If defined, this macro returns the diagnostic message when it is
11274 invalid to apply operation @var{op} to operands of types @var{type1}
11275 and @var{type2}, or @code{NULL} if validity should be determined by
11279 @deftypefn {Target Hook} {const char *} TARGET_INVALID_PARAMETER_TYPE (const_tree @var{type})
11280 If defined, this macro returns the diagnostic message when it is
11281 invalid for functions to include parameters of type @var{type},
11282 or @code{NULL} if validity should be determined by
11283 the front end. This is currently used only by the C and C++ front ends.
11286 @deftypefn {Target Hook} {const char *} TARGET_INVALID_RETURN_TYPE (const_tree @var{type})
11287 If defined, this macro returns the diagnostic message when it is
11288 invalid for functions to have return type @var{type},
11289 or @code{NULL} if validity should be determined by
11290 the front end. This is currently used only by the C and C++ front ends.
11293 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11294 If defined, this target hook returns the type to which values of
11295 @var{type} should be promoted when they appear in expressions,
11296 analogous to the integer promotions, or @code{NULL_TREE} to use the
11297 front end's normal promotion rules. This hook is useful when there are
11298 target-specific types with special promotion rules.
11299 This is currently used only by the C and C++ front ends.
11302 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11303 If defined, this hook returns the result of converting @var{expr} to
11304 @var{type}. It should return the converted expression,
11305 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11306 This hook is useful when there are target-specific types with special
11308 This is currently used only by the C and C++ front ends.
11311 @defmac TARGET_USE_JCR_SECTION
11312 This macro determines whether to use the JCR section to register Java
11313 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11314 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11318 This macro determines the size of the objective C jump buffer for the
11319 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11322 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11323 Define this macro if any target-specific attributes need to be attached
11324 to the functions in @file{libgcc} that provide low-level support for
11325 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11326 and the associated definitions of those functions.
11329 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11330 Define this macro to update the current function stack boundary if
11334 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11335 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11336 different argument pointer register is needed to access the function's
11337 argument list due to stack realignment. Return @code{NULL} if no DRAP
11341 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11342 When optimization is disabled, this hook indicates whether or not
11343 arguments should be allocated to stack slots. Normally, GCC allocates
11344 stacks slots for arguments when not optimizing in order to make
11345 debugging easier. However, when a function is declared with
11346 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11347 cannot safely move arguments from the registers in which they are passed
11348 to the stack. Therefore, this hook should return true in general, but
11349 false for naked functions. The default implementation always returns true.
11352 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11353 On some architectures it can take multiple instructions to synthesize
11354 a constant. If there is another constant already in a register that
11355 is close enough in value then it is preferable that the new constant
11356 is computed from this register using immediate addition or
11357 subtraction. We accomplish this through CSE. Besides the value of
11358 the constant we also add a lower and an upper constant anchor to the
11359 available expressions. These are then queried when encountering new
11360 constants. The anchors are computed by rounding the constant up and
11361 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11362 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11363 accepted by immediate-add plus one. We currently assume that the
11364 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11365 MIPS, where add-immediate takes a 16-bit signed value,
11366 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11367 is zero, which disables this optimization. @end deftypevr
11369 @deftypevr {Target Hook} {unsigned char} TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
11370 This value should be set if the result written by @code{atomic_test_and_set} is not exactly 1, i.e. the @code{bool} @code{true}.