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} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref})
704 If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object.
707 @deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list})
708 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.
711 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
712 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
713 but is called when the optimize level is changed via an attribute or
714 pragma or when it is reset at the end of the code affected by the
715 attribute or pragma. It is not called at the beginning of compilation
716 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
717 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
718 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
721 @defmac C_COMMON_OVERRIDE_OPTIONS
722 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
723 but is only used in the C
724 language frontends (C, Objective-C, C++, Objective-C++) and so can be
725 used to alter option flag variables which only exist in those
729 @deftypevr {Common Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
730 Some machines may desire to change what optimizations are performed for
731 various optimization levels. This variable, if defined, describes
732 options to enable at particular sets of optimization levels. These
733 options are processed once
734 just after the optimization level is determined and before the remainder
735 of the command options have been parsed, so may be overridden by other
736 options passed explicitly.
738 This processing is run once at program startup and when the optimization
739 options are changed via @code{#pragma GCC optimize} or by using the
740 @code{optimize} attribute.
743 @deftypefn {Common Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
744 Set target-dependent initial values of fields in @var{opts}.
747 @deftypefn {Common Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
748 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
751 @defmac SWITCHABLE_TARGET
752 Some targets need to switch between substantially different subtargets
753 during compilation. For example, the MIPS target has one subtarget for
754 the traditional MIPS architecture and another for MIPS16. Source code
755 can switch between these two subarchitectures using the @code{mips16}
756 and @code{nomips16} attributes.
758 Such subtargets can differ in things like the set of available
759 registers, the set of available instructions, the costs of various
760 operations, and so on. GCC caches a lot of this type of information
761 in global variables, and recomputing them for each subtarget takes a
762 significant amount of time. The compiler therefore provides a facility
763 for maintaining several versions of the global variables and quickly
764 switching between them; see @file{target-globals.h} for details.
766 Define this macro to 1 if your target needs this facility. The default
770 @node Per-Function Data
771 @section Defining data structures for per-function information.
772 @cindex per-function data
773 @cindex data structures
775 If the target needs to store information on a per-function basis, GCC
776 provides a macro and a couple of variables to allow this. Note, just
777 using statics to store the information is a bad idea, since GCC supports
778 nested functions, so you can be halfway through encoding one function
779 when another one comes along.
781 GCC defines a data structure called @code{struct function} which
782 contains all of the data specific to an individual function. This
783 structure contains a field called @code{machine} whose type is
784 @code{struct machine_function *}, which can be used by targets to point
785 to their own specific data.
787 If a target needs per-function specific data it should define the type
788 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
789 This macro should be used to initialize the function pointer
790 @code{init_machine_status}. This pointer is explained below.
792 One typical use of per-function, target specific data is to create an
793 RTX to hold the register containing the function's return address. This
794 RTX can then be used to implement the @code{__builtin_return_address}
795 function, for level 0.
797 Note---earlier implementations of GCC used a single data area to hold
798 all of the per-function information. Thus when processing of a nested
799 function began the old per-function data had to be pushed onto a
800 stack, and when the processing was finished, it had to be popped off the
801 stack. GCC used to provide function pointers called
802 @code{save_machine_status} and @code{restore_machine_status} to handle
803 the saving and restoring of the target specific information. Since the
804 single data area approach is no longer used, these pointers are no
807 @defmac INIT_EXPANDERS
808 Macro called to initialize any target specific information. This macro
809 is called once per function, before generation of any RTL has begun.
810 The intention of this macro is to allow the initialization of the
811 function pointer @code{init_machine_status}.
814 @deftypevar {void (*)(struct function *)} init_machine_status
815 If this function pointer is non-@code{NULL} it will be called once per
816 function, before function compilation starts, in order to allow the
817 target to perform any target specific initialization of the
818 @code{struct function} structure. It is intended that this would be
819 used to initialize the @code{machine} of that structure.
821 @code{struct machine_function} structures are expected to be freed by GC@.
822 Generally, any memory that they reference must be allocated by using
823 GC allocation, including the structure itself.
827 @section Storage Layout
828 @cindex storage layout
830 Note that the definitions of the macros in this table which are sizes or
831 alignments measured in bits do not need to be constant. They can be C
832 expressions that refer to static variables, such as the @code{target_flags}.
833 @xref{Run-time Target}.
835 @defmac BITS_BIG_ENDIAN
836 Define this macro to have the value 1 if the most significant bit in a
837 byte has the lowest number; otherwise define it to have the value zero.
838 This means that bit-field instructions count from the most significant
839 bit. If the machine has no bit-field instructions, then this must still
840 be defined, but it doesn't matter which value it is defined to. This
841 macro need not be a constant.
843 This macro does not affect the way structure fields are packed into
844 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
847 @defmac BYTES_BIG_ENDIAN
848 Define this macro to have the value 1 if the most significant byte in a
849 word has the lowest number. This macro need not be a constant.
852 @defmac WORDS_BIG_ENDIAN
853 Define this macro to have the value 1 if, in a multiword object, the
854 most significant word has the lowest number. This applies to both
855 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
856 order of words in memory is not the same as the order in registers. This
857 macro need not be a constant.
860 @defmac REG_WORDS_BIG_ENDIAN
861 On some machines, the order of words in a multiword object differs between
862 registers in memory. In such a situation, define this macro to describe
863 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
864 the order of words in memory.
867 @defmac FLOAT_WORDS_BIG_ENDIAN
868 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
869 @code{TFmode} floating point numbers are stored in memory with the word
870 containing the sign bit at the lowest address; otherwise define it to
871 have the value 0. This macro need not be a constant.
873 You need not define this macro if the ordering is the same as for
877 @defmac BITS_PER_UNIT
878 Define this macro to be the number of bits in an addressable storage
879 unit (byte). If you do not define this macro the default is 8.
882 @defmac BITS_PER_WORD
883 Number of bits in a word. If you do not define this macro, the default
884 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
887 @defmac MAX_BITS_PER_WORD
888 Maximum number of bits in a word. If this is undefined, the default is
889 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
890 largest value that @code{BITS_PER_WORD} can have at run-time.
893 @defmac UNITS_PER_WORD
894 Number of storage units in a word; normally the size of a general-purpose
895 register, a power of two from 1 or 8.
898 @defmac MIN_UNITS_PER_WORD
899 Minimum number of units in a word. If this is undefined, the default is
900 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
901 smallest value that @code{UNITS_PER_WORD} can have at run-time.
905 Width of a pointer, in bits. You must specify a value no wider than the
906 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
907 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
908 a value the default is @code{BITS_PER_WORD}.
911 @defmac POINTERS_EXTEND_UNSIGNED
912 A C expression that determines how pointers should be extended from
913 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
914 greater than zero if pointers should be zero-extended, zero if they
915 should be sign-extended, and negative if some other sort of conversion
916 is needed. In the last case, the extension is done by the target's
917 @code{ptr_extend} instruction.
919 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
920 and @code{word_mode} are all the same width.
923 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
924 A macro to update @var{m} and @var{unsignedp} when an object whose type
925 is @var{type} and which has the specified mode and signedness is to be
926 stored in a register. This macro is only called when @var{type} is a
929 On most RISC machines, which only have operations that operate on a full
930 register, define this macro to set @var{m} to @code{word_mode} if
931 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
932 cases, only integer modes should be widened because wider-precision
933 floating-point operations are usually more expensive than their narrower
936 For most machines, the macro definition does not change @var{unsignedp}.
937 However, some machines, have instructions that preferentially handle
938 either signed or unsigned quantities of certain modes. For example, on
939 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
940 sign-extend the result to 64 bits. On such machines, set
941 @var{unsignedp} according to which kind of extension is more efficient.
943 Do not define this macro if it would never modify @var{m}.
946 @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})
947 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
948 function return values. The target hook should return the new mode
949 and possibly change @code{*@var{punsignedp}} if the promotion should
950 change signedness. This function is called only for scalar @emph{or
953 @var{for_return} allows to distinguish the promotion of arguments and
954 return values. If it is @code{1}, a return value is being promoted and
955 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
956 If it is @code{2}, the returned mode should be that of the register in
957 which an incoming parameter is copied, or the outgoing result is computed;
958 then the hook should return the same mode as @code{promote_mode}, though
959 the signedness may be different.
961 @var{type} can be NULL when promoting function arguments of libcalls.
963 The default is to not promote arguments and return values. You can
964 also define the hook to @code{default_promote_function_mode_always_promote}
965 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
968 @defmac PARM_BOUNDARY
969 Normal alignment required for function parameters on the stack, in
970 bits. All stack parameters receive at least this much alignment
971 regardless of data type. On most machines, this is the same as the
975 @defmac STACK_BOUNDARY
976 Define this macro to the minimum alignment enforced by hardware for the
977 stack pointer on this machine. The definition is a C expression for the
978 desired alignment (measured in bits). This value is used as a default
979 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
980 this should be the same as @code{PARM_BOUNDARY}.
983 @defmac PREFERRED_STACK_BOUNDARY
984 Define this macro if you wish to preserve a certain alignment for the
985 stack pointer, greater than what the hardware enforces. The definition
986 is a C expression for the desired alignment (measured in bits). This
987 macro must evaluate to a value equal to or larger than
988 @code{STACK_BOUNDARY}.
991 @defmac INCOMING_STACK_BOUNDARY
992 Define this macro if the incoming stack boundary may be different
993 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
994 to a value equal to or larger than @code{STACK_BOUNDARY}.
997 @defmac FUNCTION_BOUNDARY
998 Alignment required for a function entry point, in bits.
1001 @defmac BIGGEST_ALIGNMENT
1002 Biggest alignment that any data type can require on this machine, in
1003 bits. Note that this is not the biggest alignment that is supported,
1004 just the biggest alignment that, when violated, may cause a fault.
1007 @defmac MALLOC_ABI_ALIGNMENT
1008 Alignment, in bits, a C conformant malloc implementation has to
1009 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1012 @defmac ATTRIBUTE_ALIGNED_VALUE
1013 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1014 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1017 @defmac MINIMUM_ATOMIC_ALIGNMENT
1018 If defined, the smallest alignment, in bits, that can be given to an
1019 object that can be referenced in one operation, without disturbing any
1020 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1021 on machines that don't have byte or half-word store operations.
1024 @defmac BIGGEST_FIELD_ALIGNMENT
1025 Biggest alignment that any structure or union field can require on this
1026 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1027 structure and union fields only, unless the field alignment has been set
1028 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1031 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1032 An expression for the alignment of a structure field @var{field} if the
1033 alignment computed in the usual way (including applying of
1034 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1035 alignment) is @var{computed}. It overrides alignment only if the
1036 field alignment has not been set by the
1037 @code{__attribute__ ((aligned (@var{n})))} construct.
1040 @defmac MAX_STACK_ALIGNMENT
1041 Biggest stack alignment guaranteed by the backend. Use this macro
1042 to specify the maximum alignment of a variable on stack.
1044 If not defined, the default value is @code{STACK_BOUNDARY}.
1046 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1047 @c But the fix for PR 32893 indicates that we can only guarantee
1048 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1049 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1052 @defmac MAX_OFILE_ALIGNMENT
1053 Biggest alignment supported by the object file format of this machine.
1054 Use this macro to limit the alignment which can be specified using the
1055 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1056 the default value is @code{BIGGEST_ALIGNMENT}.
1058 On systems that use ELF, the default (in @file{config/elfos.h}) is
1059 the largest supported 32-bit ELF section alignment representable on
1060 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1061 On 32-bit ELF the largest supported section alignment in bits is
1062 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1065 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1066 If defined, a C expression to compute the alignment for a variable in
1067 the static store. @var{type} is the data type, and @var{basic-align} is
1068 the alignment that the object would ordinarily have. The value of this
1069 macro is used instead of that alignment to align the object.
1071 If this macro is not defined, then @var{basic-align} is used.
1074 One use of this macro is to increase alignment of medium-size data to
1075 make it all fit in fewer cache lines. Another is to cause character
1076 arrays to be word-aligned so that @code{strcpy} calls that copy
1077 constants to character arrays can be done inline.
1080 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1081 If defined, a C expression to compute the alignment given to a constant
1082 that is being placed in memory. @var{constant} is the constant and
1083 @var{basic-align} is the alignment that the object would ordinarily
1084 have. The value of this macro is used instead of that alignment to
1087 If this macro is not defined, then @var{basic-align} is used.
1089 The typical use of this macro is to increase alignment for string
1090 constants to be word aligned so that @code{strcpy} calls that copy
1091 constants can be done inline.
1094 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1095 If defined, a C expression to compute the alignment for a variable in
1096 the local store. @var{type} is the data type, and @var{basic-align} is
1097 the alignment that the object would ordinarily have. The value of this
1098 macro is used instead of that alignment to align the object.
1100 If this macro is not defined, then @var{basic-align} is used.
1102 One use of this macro is to increase alignment of medium-size data to
1103 make it all fit in fewer cache lines.
1105 If the value of this macro has a type, it should be an unsigned type.
1108 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_VECTOR_ALIGNMENT (const_tree @var{type})
1109 This hook can be used to define the alignment for a vector of type
1110 @var{type}, in order to comply with a platform ABI. The default is to
1111 require natural alignment for vector types. The alignment returned by
1112 this hook must be a power-of-two multiple of the default alignment of
1113 the vector element type.
1116 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1117 If defined, a C expression to compute the alignment for stack slot.
1118 @var{type} is the data type, @var{mode} is the widest mode available,
1119 and @var{basic-align} is the alignment that the slot would ordinarily
1120 have. The value of this macro is used instead of that alignment to
1123 If this macro is not defined, then @var{basic-align} is used when
1124 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1127 This macro is to set alignment of stack slot to the maximum alignment
1128 of all possible modes which the slot may have.
1130 If the value of this macro has a type, it should be an unsigned type.
1133 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1134 If defined, a C expression to compute the alignment for a local
1135 variable @var{decl}.
1137 If this macro is not defined, then
1138 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1141 One use of this macro is to increase alignment of medium-size data to
1142 make it all fit in fewer cache lines.
1144 If the value of this macro has a type, it should be an unsigned type.
1147 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1148 If defined, a C expression to compute the minimum required alignment
1149 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1150 @var{mode}, assuming normal alignment @var{align}.
1152 If this macro is not defined, then @var{align} will be used.
1155 @defmac EMPTY_FIELD_BOUNDARY
1156 Alignment in bits to be given to a structure bit-field that follows an
1157 empty field such as @code{int : 0;}.
1159 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1162 @defmac STRUCTURE_SIZE_BOUNDARY
1163 Number of bits which any structure or union's size must be a multiple of.
1164 Each structure or union's size is rounded up to a multiple of this.
1166 If you do not define this macro, the default is the same as
1167 @code{BITS_PER_UNIT}.
1170 @defmac STRICT_ALIGNMENT
1171 Define this macro to be the value 1 if instructions will fail to work
1172 if given data not on the nominal alignment. If instructions will merely
1173 go slower in that case, define this macro as 0.
1176 @defmac PCC_BITFIELD_TYPE_MATTERS
1177 Define this if you wish to imitate the way many other C compilers handle
1178 alignment of bit-fields and the structures that contain them.
1180 The behavior is that the type written for a named bit-field (@code{int},
1181 @code{short}, or other integer type) imposes an alignment for the entire
1182 structure, as if the structure really did contain an ordinary field of
1183 that type. In addition, the bit-field is placed within the structure so
1184 that it would fit within such a field, not crossing a boundary for it.
1186 Thus, on most machines, a named bit-field whose type is written as
1187 @code{int} would not cross a four-byte boundary, and would force
1188 four-byte alignment for the whole structure. (The alignment used may
1189 not be four bytes; it is controlled by the other alignment parameters.)
1191 An unnamed bit-field will not affect the alignment of the containing
1194 If the macro is defined, its definition should be a C expression;
1195 a nonzero value for the expression enables this behavior.
1197 Note that if this macro is not defined, or its value is zero, some
1198 bit-fields may cross more than one alignment boundary. The compiler can
1199 support such references if there are @samp{insv}, @samp{extv}, and
1200 @samp{extzv} insns that can directly reference memory.
1202 The other known way of making bit-fields work is to define
1203 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1204 Then every structure can be accessed with fullwords.
1206 Unless the machine has bit-field instructions or you define
1207 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1208 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1210 If your aim is to make GCC use the same conventions for laying out
1211 bit-fields as are used by another compiler, here is how to investigate
1212 what the other compiler does. Compile and run this program:
1231 printf ("Size of foo1 is %d\n",
1232 sizeof (struct foo1));
1233 printf ("Size of foo2 is %d\n",
1234 sizeof (struct foo2));
1239 If this prints 2 and 5, then the compiler's behavior is what you would
1240 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1243 @defmac BITFIELD_NBYTES_LIMITED
1244 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1245 to aligning a bit-field within the structure.
1248 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1249 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1250 whether unnamed bitfields affect the alignment of the containing
1251 structure. The hook should return true if the structure should inherit
1252 the alignment requirements of an unnamed bitfield's type.
1255 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1256 This target hook should return @code{true} if accesses to volatile bitfields
1257 should use the narrowest mode possible. It should return @code{false} if
1258 these accesses should use the bitfield container type.
1260 The default is @code{!TARGET_STRICT_ALIGN}.
1263 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1264 Return 1 if a structure or array containing @var{field} should be accessed using
1267 If @var{field} is the only field in the structure, @var{mode} is its
1268 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1269 case where structures of one field would require the structure's mode to
1270 retain the field's mode.
1272 Normally, this is not needed.
1275 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1276 Define this macro as an expression for the alignment of a type (given
1277 by @var{type} as a tree node) if the alignment computed in the usual
1278 way is @var{computed} and the alignment explicitly specified was
1281 The default is to use @var{specified} if it is larger; otherwise, use
1282 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1285 @defmac MAX_FIXED_MODE_SIZE
1286 An integer expression for the size in bits of the largest integer
1287 machine mode that should actually be used. All integer machine modes of
1288 this size or smaller can be used for structures and unions with the
1289 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1290 (DImode)} is assumed.
1293 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1294 If defined, an expression of type @code{enum machine_mode} that
1295 specifies the mode of the save area operand of a
1296 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1297 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1298 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1299 having its mode specified.
1301 You need not define this macro if it always returns @code{Pmode}. You
1302 would most commonly define this macro if the
1303 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1307 @defmac STACK_SIZE_MODE
1308 If defined, an expression of type @code{enum machine_mode} that
1309 specifies the mode of the size increment operand of an
1310 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1312 You need not define this macro if it always returns @code{word_mode}.
1313 You would most commonly define this macro if the @code{allocate_stack}
1314 pattern needs to support both a 32- and a 64-bit mode.
1317 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE (void)
1318 This target hook should return the mode to be used for the return value
1319 of compare instructions expanded to libgcc calls. If not defined
1320 @code{word_mode} is returned which is the right choice for a majority of
1324 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1325 This target hook should return the mode to be used for the shift count operand
1326 of shift instructions expanded to libgcc calls. If not defined
1327 @code{word_mode} is returned which is the right choice for a majority of
1331 @deftypefn {Target Hook} {enum machine_mode} TARGET_UNWIND_WORD_MODE (void)
1332 Return machine mode to be used for @code{_Unwind_Word} type.
1333 The default is to use @code{word_mode}.
1336 @defmac ROUND_TOWARDS_ZERO
1337 If defined, this macro should be true if the prevailing rounding
1338 mode is towards zero.
1340 Defining this macro only affects the way @file{libgcc.a} emulates
1341 floating-point arithmetic.
1343 Not defining this macro is equivalent to returning zero.
1346 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1347 This macro should return true if floats with @var{size}
1348 bits do not have a NaN or infinity representation, but use the largest
1349 exponent for normal numbers instead.
1351 Defining this macro only affects the way @file{libgcc.a} emulates
1352 floating-point arithmetic.
1354 The default definition of this macro returns false for all sizes.
1357 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1358 This target hook returns @code{true} if bit-fields in the given
1359 @var{record_type} are to be laid out following the rules of Microsoft
1360 Visual C/C++, namely: (i) a bit-field won't share the same storage
1361 unit with the previous bit-field if their underlying types have
1362 different sizes, and the bit-field will be aligned to the highest
1363 alignment of the underlying types of itself and of the previous
1364 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1365 the whole enclosing structure, even if it is unnamed; except that
1366 (iii) a zero-sized bit-field will be disregarded unless it follows
1367 another bit-field of nonzero size. If this hook returns @code{true},
1368 other macros that control bit-field layout are ignored.
1370 When a bit-field is inserted into a packed record, the whole size
1371 of the underlying type is used by one or more same-size adjacent
1372 bit-fields (that is, if its long:3, 32 bits is used in the record,
1373 and any additional adjacent long bit-fields are packed into the same
1374 chunk of 32 bits. However, if the size changes, a new field of that
1375 size is allocated). In an unpacked record, this is the same as using
1376 alignment, but not equivalent when packing.
1378 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1379 the latter will take precedence. If @samp{__attribute__((packed))} is
1380 used on a single field when MS bit-fields are in use, it will take
1381 precedence for that field, but the alignment of the rest of the structure
1382 may affect its placement.
1385 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1386 Returns true if the target supports decimal floating point.
1389 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1390 Returns true if the target supports fixed-point arithmetic.
1393 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1394 This hook is called just before expansion into rtl, allowing the target
1395 to perform additional initializations or analysis before the expansion.
1396 For example, the rs6000 port uses it to allocate a scratch stack slot
1397 for use in copying SDmode values between memory and floating point
1398 registers whenever the function being expanded has any SDmode
1402 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1403 This hook allows the backend to perform additional instantiations on rtl
1404 that are not actually in any insns yet, but will be later.
1407 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1408 If your target defines any fundamental types, or any types your target
1409 uses should be mangled differently from the default, define this hook
1410 to return the appropriate encoding for these types as part of a C++
1411 mangled name. The @var{type} argument is the tree structure representing
1412 the type to be mangled. The hook may be applied to trees which are
1413 not target-specific fundamental types; it should return @code{NULL}
1414 for all such types, as well as arguments it does not recognize. If the
1415 return value is not @code{NULL}, it must point to a statically-allocated
1418 Target-specific fundamental types might be new fundamental types or
1419 qualified versions of ordinary fundamental types. Encode new
1420 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1421 is the name used for the type in source code, and @var{n} is the
1422 length of @var{name} in decimal. Encode qualified versions of
1423 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1424 @var{name} is the name used for the type qualifier in source code,
1425 @var{n} is the length of @var{name} as above, and @var{code} is the
1426 code used to represent the unqualified version of this type. (See
1427 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1428 codes.) In both cases the spaces are for clarity; do not include any
1429 spaces in your string.
1431 This hook is applied to types prior to typedef resolution. If the mangled
1432 name for a particular type depends only on that type's main variant, you
1433 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1436 The default version of this hook always returns @code{NULL}, which is
1437 appropriate for a target that does not define any new fundamental
1442 @section Layout of Source Language Data Types
1444 These macros define the sizes and other characteristics of the standard
1445 basic data types used in programs being compiled. Unlike the macros in
1446 the previous section, these apply to specific features of C and related
1447 languages, rather than to fundamental aspects of storage layout.
1449 @defmac INT_TYPE_SIZE
1450 A C expression for the size in bits of the type @code{int} on the
1451 target machine. If you don't define this, the default is one word.
1454 @defmac SHORT_TYPE_SIZE
1455 A C expression for the size in bits of the type @code{short} on the
1456 target machine. If you don't define this, the default is half a word.
1457 (If this would be less than one storage unit, it is rounded up to one
1461 @defmac LONG_TYPE_SIZE
1462 A C expression for the size in bits of the type @code{long} on the
1463 target machine. If you don't define this, the default is one word.
1466 @defmac ADA_LONG_TYPE_SIZE
1467 On some machines, the size used for the Ada equivalent of the type
1468 @code{long} by a native Ada compiler differs from that used by C@. In
1469 that situation, define this macro to be a C expression to be used for
1470 the size of that type. If you don't define this, the default is the
1471 value of @code{LONG_TYPE_SIZE}.
1474 @defmac LONG_LONG_TYPE_SIZE
1475 A C expression for the size in bits of the type @code{long long} on the
1476 target machine. If you don't define this, the default is two
1477 words. If you want to support GNU Ada on your machine, the value of this
1478 macro must be at least 64.
1481 @defmac CHAR_TYPE_SIZE
1482 A C expression for the size in bits of the type @code{char} on the
1483 target machine. If you don't define this, the default is
1484 @code{BITS_PER_UNIT}.
1487 @defmac BOOL_TYPE_SIZE
1488 A C expression for the size in bits of the C++ type @code{bool} and
1489 C99 type @code{_Bool} on the target machine. If you don't define
1490 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1493 @defmac FLOAT_TYPE_SIZE
1494 A C expression for the size in bits of the type @code{float} on the
1495 target machine. If you don't define this, the default is one word.
1498 @defmac DOUBLE_TYPE_SIZE
1499 A C expression for the size in bits of the type @code{double} on the
1500 target machine. If you don't define this, the default is two
1504 @defmac LONG_DOUBLE_TYPE_SIZE
1505 A C expression for the size in bits of the type @code{long double} on
1506 the target machine. If you don't define this, the default is two
1510 @defmac SHORT_FRACT_TYPE_SIZE
1511 A C expression for the size in bits of the type @code{short _Fract} on
1512 the target machine. If you don't define this, the default is
1513 @code{BITS_PER_UNIT}.
1516 @defmac FRACT_TYPE_SIZE
1517 A C expression for the size in bits of the type @code{_Fract} on
1518 the target machine. If you don't define this, the default is
1519 @code{BITS_PER_UNIT * 2}.
1522 @defmac LONG_FRACT_TYPE_SIZE
1523 A C expression for the size in bits of the type @code{long _Fract} on
1524 the target machine. If you don't define this, the default is
1525 @code{BITS_PER_UNIT * 4}.
1528 @defmac LONG_LONG_FRACT_TYPE_SIZE
1529 A C expression for the size in bits of the type @code{long long _Fract} on
1530 the target machine. If you don't define this, the default is
1531 @code{BITS_PER_UNIT * 8}.
1534 @defmac SHORT_ACCUM_TYPE_SIZE
1535 A C expression for the size in bits of the type @code{short _Accum} on
1536 the target machine. If you don't define this, the default is
1537 @code{BITS_PER_UNIT * 2}.
1540 @defmac ACCUM_TYPE_SIZE
1541 A C expression for the size in bits of the type @code{_Accum} on
1542 the target machine. If you don't define this, the default is
1543 @code{BITS_PER_UNIT * 4}.
1546 @defmac LONG_ACCUM_TYPE_SIZE
1547 A C expression for the size in bits of the type @code{long _Accum} on
1548 the target machine. If you don't define this, the default is
1549 @code{BITS_PER_UNIT * 8}.
1552 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1553 A C expression for the size in bits of the type @code{long long _Accum} on
1554 the target machine. If you don't define this, the default is
1555 @code{BITS_PER_UNIT * 16}.
1558 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1559 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1560 if you want routines in @file{libgcc2.a} for a size other than
1561 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1562 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1565 @defmac LIBGCC2_HAS_DF_MODE
1566 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1567 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1568 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1569 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1570 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1574 @defmac LIBGCC2_HAS_XF_MODE
1575 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1576 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1577 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1578 is 80 then the default is 1, otherwise it is 0.
1581 @defmac LIBGCC2_HAS_TF_MODE
1582 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1583 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1584 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1585 is 128 then the default is 1, otherwise it is 0.
1588 @defmac LIBGCC2_GNU_PREFIX
1589 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1590 hook and should be defined if that hook is overriden to be true. It
1591 causes function names in libgcc to be changed to use a @code{__gnu_}
1592 prefix for their name rather than the default @code{__}. A port which
1593 uses this macro should also arrange to use @file{t-gnu-prefix} in
1594 the libgcc @file{config.host}.
1601 Define these macros to be the size in bits of the mantissa of
1602 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1603 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1604 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1605 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1606 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1607 @code{DOUBLE_TYPE_SIZE} or
1608 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1611 @defmac TARGET_FLT_EVAL_METHOD
1612 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1613 assuming, if applicable, that the floating-point control word is in its
1614 default state. If you do not define this macro the value of
1615 @code{FLT_EVAL_METHOD} will be zero.
1618 @defmac WIDEST_HARDWARE_FP_SIZE
1619 A C expression for the size in bits of the widest floating-point format
1620 supported by the hardware. If you define this macro, you must specify a
1621 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1622 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1626 @defmac DEFAULT_SIGNED_CHAR
1627 An expression whose value is 1 or 0, according to whether the type
1628 @code{char} should be signed or unsigned by default. The user can
1629 always override this default with the options @option{-fsigned-char}
1630 and @option{-funsigned-char}.
1633 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1634 This target hook should return true if the compiler should give an
1635 @code{enum} type only as many bytes as it takes to represent the range
1636 of possible values of that type. It should return false if all
1637 @code{enum} types should be allocated like @code{int}.
1639 The default is to return false.
1643 A C expression for a string describing the name of the data type to use
1644 for size values. The typedef name @code{size_t} is defined using the
1645 contents of the string.
1647 The string can contain more than one keyword. If so, separate them with
1648 spaces, and write first any length keyword, then @code{unsigned} if
1649 appropriate, and finally @code{int}. The string must exactly match one
1650 of the data type names defined in the function
1651 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1652 omit @code{int} or change the order---that would cause the compiler to
1655 If you don't define this macro, the default is @code{"long unsigned
1659 @defmac PTRDIFF_TYPE
1660 A C expression for a string describing the name of the data type to use
1661 for the result of subtracting two pointers. The typedef name
1662 @code{ptrdiff_t} is defined using the contents of the string. See
1663 @code{SIZE_TYPE} above for more information.
1665 If you don't define this macro, the default is @code{"long int"}.
1669 A C expression for a string describing the name of the data type to use
1670 for wide characters. The typedef name @code{wchar_t} is defined using
1671 the contents of the string. See @code{SIZE_TYPE} above for more
1674 If you don't define this macro, the default is @code{"int"}.
1677 @defmac WCHAR_TYPE_SIZE
1678 A C expression for the size in bits of the data type for wide
1679 characters. This is used in @code{cpp}, which cannot make use of
1684 A C expression for a string describing the name of the data type to
1685 use for wide characters passed to @code{printf} and returned from
1686 @code{getwc}. The typedef name @code{wint_t} is defined using the
1687 contents of the string. See @code{SIZE_TYPE} above for more
1690 If you don't define this macro, the default is @code{"unsigned int"}.
1694 A C expression for a string describing the name of the data type that
1695 can represent any value of any standard or extended signed integer type.
1696 The typedef name @code{intmax_t} is defined using the contents of the
1697 string. See @code{SIZE_TYPE} above for more information.
1699 If you don't define this macro, the default is the first of
1700 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1701 much precision as @code{long long int}.
1704 @defmac UINTMAX_TYPE
1705 A C expression for a string describing the name of the data type that
1706 can represent any value of any standard or extended unsigned integer
1707 type. The typedef name @code{uintmax_t} is defined using the contents
1708 of the string. See @code{SIZE_TYPE} above for more information.
1710 If you don't define this macro, the default is the first of
1711 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1712 unsigned int"} that has as much precision as @code{long long unsigned
1716 @defmac SIG_ATOMIC_TYPE
1722 @defmacx UINT16_TYPE
1723 @defmacx UINT32_TYPE
1724 @defmacx UINT64_TYPE
1725 @defmacx INT_LEAST8_TYPE
1726 @defmacx INT_LEAST16_TYPE
1727 @defmacx INT_LEAST32_TYPE
1728 @defmacx INT_LEAST64_TYPE
1729 @defmacx UINT_LEAST8_TYPE
1730 @defmacx UINT_LEAST16_TYPE
1731 @defmacx UINT_LEAST32_TYPE
1732 @defmacx UINT_LEAST64_TYPE
1733 @defmacx INT_FAST8_TYPE
1734 @defmacx INT_FAST16_TYPE
1735 @defmacx INT_FAST32_TYPE
1736 @defmacx INT_FAST64_TYPE
1737 @defmacx UINT_FAST8_TYPE
1738 @defmacx UINT_FAST16_TYPE
1739 @defmacx UINT_FAST32_TYPE
1740 @defmacx UINT_FAST64_TYPE
1741 @defmacx INTPTR_TYPE
1742 @defmacx UINTPTR_TYPE
1743 C expressions for the standard types @code{sig_atomic_t},
1744 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1745 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1746 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1747 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1748 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1749 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1750 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1751 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1752 @code{SIZE_TYPE} above for more information.
1754 If any of these macros evaluates to a null pointer, the corresponding
1755 type is not supported; if GCC is configured to provide
1756 @code{<stdint.h>} in such a case, the header provided may not conform
1757 to C99, depending on the type in question. The defaults for all of
1758 these macros are null pointers.
1761 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1762 The C++ compiler represents a pointer-to-member-function with a struct
1769 ptrdiff_t vtable_index;
1776 The C++ compiler must use one bit to indicate whether the function that
1777 will be called through a pointer-to-member-function is virtual.
1778 Normally, we assume that the low-order bit of a function pointer must
1779 always be zero. Then, by ensuring that the vtable_index is odd, we can
1780 distinguish which variant of the union is in use. But, on some
1781 platforms function pointers can be odd, and so this doesn't work. In
1782 that case, we use the low-order bit of the @code{delta} field, and shift
1783 the remainder of the @code{delta} field to the left.
1785 GCC will automatically make the right selection about where to store
1786 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1787 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1788 set such that functions always start at even addresses, but the lowest
1789 bit of pointers to functions indicate whether the function at that
1790 address is in ARM or Thumb mode. If this is the case of your
1791 architecture, you should define this macro to
1792 @code{ptrmemfunc_vbit_in_delta}.
1794 In general, you should not have to define this macro. On architectures
1795 in which function addresses are always even, according to
1796 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1797 @code{ptrmemfunc_vbit_in_pfn}.
1800 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1801 Normally, the C++ compiler uses function pointers in vtables. This
1802 macro allows the target to change to use ``function descriptors''
1803 instead. Function descriptors are found on targets for whom a
1804 function pointer is actually a small data structure. Normally the
1805 data structure consists of the actual code address plus a data
1806 pointer to which the function's data is relative.
1808 If vtables are used, the value of this macro should be the number
1809 of words that the function descriptor occupies.
1812 @defmac TARGET_VTABLE_ENTRY_ALIGN
1813 By default, the vtable entries are void pointers, the so the alignment
1814 is the same as pointer alignment. The value of this macro specifies
1815 the alignment of the vtable entry in bits. It should be defined only
1816 when special alignment is necessary. */
1819 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1820 There are a few non-descriptor entries in the vtable at offsets below
1821 zero. If these entries must be padded (say, to preserve the alignment
1822 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1823 of words in each data entry.
1827 @section Register Usage
1828 @cindex register usage
1830 This section explains how to describe what registers the target machine
1831 has, and how (in general) they can be used.
1833 The description of which registers a specific instruction can use is
1834 done with register classes; see @ref{Register Classes}. For information
1835 on using registers to access a stack frame, see @ref{Frame Registers}.
1836 For passing values in registers, see @ref{Register Arguments}.
1837 For returning values in registers, see @ref{Scalar Return}.
1840 * Register Basics:: Number and kinds of registers.
1841 * Allocation Order:: Order in which registers are allocated.
1842 * Values in Registers:: What kinds of values each reg can hold.
1843 * Leaf Functions:: Renumbering registers for leaf functions.
1844 * Stack Registers:: Handling a register stack such as 80387.
1847 @node Register Basics
1848 @subsection Basic Characteristics of Registers
1850 @c prevent bad page break with this line
1851 Registers have various characteristics.
1853 @defmac FIRST_PSEUDO_REGISTER
1854 Number of hardware registers known to the compiler. They receive
1855 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1856 pseudo register's number really is assigned the number
1857 @code{FIRST_PSEUDO_REGISTER}.
1860 @defmac FIXED_REGISTERS
1861 @cindex fixed register
1862 An initializer that says which registers are used for fixed purposes
1863 all throughout the compiled code and are therefore not available for
1864 general allocation. These would include the stack pointer, the frame
1865 pointer (except on machines where that can be used as a general
1866 register when no frame pointer is needed), the program counter on
1867 machines where that is considered one of the addressable registers,
1868 and any other numbered register with a standard use.
1870 This information is expressed as a sequence of numbers, separated by
1871 commas and surrounded by braces. The @var{n}th number is 1 if
1872 register @var{n} is fixed, 0 otherwise.
1874 The table initialized from this macro, and the table initialized by
1875 the following one, may be overridden at run time either automatically,
1876 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1877 the user with the command options @option{-ffixed-@var{reg}},
1878 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1881 @defmac CALL_USED_REGISTERS
1882 @cindex call-used register
1883 @cindex call-clobbered register
1884 @cindex call-saved register
1885 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1886 clobbered (in general) by function calls as well as for fixed
1887 registers. This macro therefore identifies the registers that are not
1888 available for general allocation of values that must live across
1891 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1892 automatically saves it on function entry and restores it on function
1893 exit, if the register is used within the function.
1896 @defmac CALL_REALLY_USED_REGISTERS
1897 @cindex call-used register
1898 @cindex call-clobbered register
1899 @cindex call-saved register
1900 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1901 that the entire set of @code{FIXED_REGISTERS} be included.
1902 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1903 This macro is optional. If not specified, it defaults to the value
1904 of @code{CALL_USED_REGISTERS}.
1907 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1908 @cindex call-used register
1909 @cindex call-clobbered register
1910 @cindex call-saved register
1911 A C expression that is nonzero if it is not permissible to store a
1912 value of mode @var{mode} in hard register number @var{regno} across a
1913 call without some part of it being clobbered. For most machines this
1914 macro need not be defined. It is only required for machines that do not
1915 preserve the entire contents of a register across a call.
1919 @findex call_used_regs
1922 @findex reg_class_contents
1923 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1924 This hook may conditionally modify five variables
1925 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1926 @code{reg_names}, and @code{reg_class_contents}, to take into account
1927 any dependence of these register sets on target flags. The first three
1928 of these are of type @code{char []} (interpreted as Boolean vectors).
1929 @code{global_regs} is a @code{const char *[]}, and
1930 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1931 called, @code{fixed_regs}, @code{call_used_regs},
1932 @code{reg_class_contents}, and @code{reg_names} have been initialized
1933 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1934 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1935 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1936 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1937 command options have been applied.
1939 @cindex disabling certain registers
1940 @cindex controlling register usage
1941 If the usage of an entire class of registers depends on the target
1942 flags, you may indicate this to GCC by using this macro to modify
1943 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1944 registers in the classes which should not be used by GCC@. Also define
1945 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1946 to return @code{NO_REGS} if it
1947 is called with a letter for a class that shouldn't be used.
1949 (However, if this class is not included in @code{GENERAL_REGS} and all
1950 of the insn patterns whose constraints permit this class are
1951 controlled by target switches, then GCC will automatically avoid using
1952 these registers when the target switches are opposed to them.)
1955 @defmac INCOMING_REGNO (@var{out})
1956 Define this macro if the target machine has register windows. This C
1957 expression returns the register number as seen by the called function
1958 corresponding to the register number @var{out} as seen by the calling
1959 function. Return @var{out} if register number @var{out} is not an
1963 @defmac OUTGOING_REGNO (@var{in})
1964 Define this macro if the target machine has register windows. This C
1965 expression returns the register number as seen by the calling function
1966 corresponding to the register number @var{in} as seen by the called
1967 function. Return @var{in} if register number @var{in} is not an inbound
1971 @defmac LOCAL_REGNO (@var{regno})
1972 Define this macro if the target machine has register windows. This C
1973 expression returns true if the register is call-saved but is in the
1974 register window. Unlike most call-saved registers, such registers
1975 need not be explicitly restored on function exit or during non-local
1980 If the program counter has a register number, define this as that
1981 register number. Otherwise, do not define it.
1984 @node Allocation Order
1985 @subsection Order of Allocation of Registers
1986 @cindex order of register allocation
1987 @cindex register allocation order
1989 @c prevent bad page break with this line
1990 Registers are allocated in order.
1992 @defmac REG_ALLOC_ORDER
1993 If defined, an initializer for a vector of integers, containing the
1994 numbers of hard registers in the order in which GCC should prefer
1995 to use them (from most preferred to least).
1997 If this macro is not defined, registers are used lowest numbered first
1998 (all else being equal).
2000 One use of this macro is on machines where the highest numbered
2001 registers must always be saved and the save-multiple-registers
2002 instruction supports only sequences of consecutive registers. On such
2003 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2004 the highest numbered allocable register first.
2007 @defmac ADJUST_REG_ALLOC_ORDER
2008 A C statement (sans semicolon) to choose the order in which to allocate
2009 hard registers for pseudo-registers local to a basic block.
2011 Store the desired register order in the array @code{reg_alloc_order}.
2012 Element 0 should be the register to allocate first; element 1, the next
2013 register; and so on.
2015 The macro body should not assume anything about the contents of
2016 @code{reg_alloc_order} before execution of the macro.
2018 On most machines, it is not necessary to define this macro.
2021 @defmac HONOR_REG_ALLOC_ORDER
2022 Normally, IRA tries to estimate the costs for saving a register in the
2023 prologue and restoring it in the epilogue. This discourages it from
2024 using call-saved registers. If a machine wants to ensure that IRA
2025 allocates registers in the order given by REG_ALLOC_ORDER even if some
2026 call-saved registers appear earlier than call-used ones, this macro
2030 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2031 In some case register allocation order is not enough for the
2032 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2033 If this macro is defined, it should return a floating point value
2034 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2035 be increased by approximately the pseudo's usage frequency times the
2036 value returned by this macro. Not defining this macro is equivalent
2037 to having it always return @code{0.0}.
2039 On most machines, it is not necessary to define this macro.
2042 @node Values in Registers
2043 @subsection How Values Fit in Registers
2045 This section discusses the macros that describe which kinds of values
2046 (specifically, which machine modes) each register can hold, and how many
2047 consecutive registers are needed for a given mode.
2049 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2050 A C expression for the number of consecutive hard registers, starting
2051 at register number @var{regno}, required to hold a value of mode
2052 @var{mode}. This macro must never return zero, even if a register
2053 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2054 and/or CANNOT_CHANGE_MODE_CLASS instead.
2056 On a machine where all registers are exactly one word, a suitable
2057 definition of this macro is
2060 #define HARD_REGNO_NREGS(REGNO, MODE) \
2061 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2066 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2067 A C expression that is nonzero if a value of mode @var{mode}, stored
2068 in memory, ends with padding that causes it to take up more space than
2069 in registers starting at register number @var{regno} (as determined by
2070 multiplying GCC's notion of the size of the register when containing
2071 this mode by the number of registers returned by
2072 @code{HARD_REGNO_NREGS}). By default this is zero.
2074 For example, if a floating-point value is stored in three 32-bit
2075 registers but takes up 128 bits in memory, then this would be
2078 This macros only needs to be defined if there are cases where
2079 @code{subreg_get_info}
2080 would otherwise wrongly determine that a @code{subreg} can be
2081 represented by an offset to the register number, when in fact such a
2082 @code{subreg} would contain some of the padding not stored in
2083 registers and so not be representable.
2086 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2087 For values of @var{regno} and @var{mode} for which
2088 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2089 returning the greater number of registers required to hold the value
2090 including any padding. In the example above, the value would be four.
2093 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2094 Define this macro if the natural size of registers that hold values
2095 of mode @var{mode} is not the word size. It is a C expression that
2096 should give the natural size in bytes for the specified mode. It is
2097 used by the register allocator to try to optimize its results. This
2098 happens for example on SPARC 64-bit where the natural size of
2099 floating-point registers is still 32-bit.
2102 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2103 A C expression that is nonzero if it is permissible to store a value
2104 of mode @var{mode} in hard register number @var{regno} (or in several
2105 registers starting with that one). For a machine where all registers
2106 are equivalent, a suitable definition is
2109 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2112 You need not include code to check for the numbers of fixed registers,
2113 because the allocation mechanism considers them to be always occupied.
2115 @cindex register pairs
2116 On some machines, double-precision values must be kept in even/odd
2117 register pairs. You can implement that by defining this macro to reject
2118 odd register numbers for such modes.
2120 The minimum requirement for a mode to be OK in a register is that the
2121 @samp{mov@var{mode}} instruction pattern support moves between the
2122 register and other hard register in the same class and that moving a
2123 value into the register and back out not alter it.
2125 Since the same instruction used to move @code{word_mode} will work for
2126 all narrower integer modes, it is not necessary on any machine for
2127 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2128 you define patterns @samp{movhi}, etc., to take advantage of this. This
2129 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2130 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2133 Many machines have special registers for floating point arithmetic.
2134 Often people assume that floating point machine modes are allowed only
2135 in floating point registers. This is not true. Any registers that
2136 can hold integers can safely @emph{hold} a floating point machine
2137 mode, whether or not floating arithmetic can be done on it in those
2138 registers. Integer move instructions can be used to move the values.
2140 On some machines, though, the converse is true: fixed-point machine
2141 modes may not go in floating registers. This is true if the floating
2142 registers normalize any value stored in them, because storing a
2143 non-floating value there would garble it. In this case,
2144 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2145 floating registers. But if the floating registers do not automatically
2146 normalize, if you can store any bit pattern in one and retrieve it
2147 unchanged without a trap, then any machine mode may go in a floating
2148 register, so you can define this macro to say so.
2150 The primary significance of special floating registers is rather that
2151 they are the registers acceptable in floating point arithmetic
2152 instructions. However, this is of no concern to
2153 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2154 constraints for those instructions.
2156 On some machines, the floating registers are especially slow to access,
2157 so that it is better to store a value in a stack frame than in such a
2158 register if floating point arithmetic is not being done. As long as the
2159 floating registers are not in class @code{GENERAL_REGS}, they will not
2160 be used unless some pattern's constraint asks for one.
2163 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2164 A C expression that is nonzero if it is OK to rename a hard register
2165 @var{from} to another hard register @var{to}.
2167 One common use of this macro is to prevent renaming of a register to
2168 another register that is not saved by a prologue in an interrupt
2171 The default is always nonzero.
2174 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2175 A C expression that is nonzero if a value of mode
2176 @var{mode1} is accessible in mode @var{mode2} without copying.
2178 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2179 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2180 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2181 should be nonzero. If they differ for any @var{r}, you should define
2182 this macro to return zero unless some other mechanism ensures the
2183 accessibility of the value in a narrower mode.
2185 You should define this macro to return nonzero in as many cases as
2186 possible since doing so will allow GCC to perform better register
2190 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2191 This target hook should return @code{true} if it is OK to use a hard register
2192 @var{regno} as scratch reg in peephole2.
2194 One common use of this macro is to prevent using of a register that
2195 is not saved by a prologue in an interrupt handler.
2197 The default version of this hook always returns @code{true}.
2200 @defmac AVOID_CCMODE_COPIES
2201 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2202 registers. You should only define this macro if support for copying to/from
2203 @code{CCmode} is incomplete.
2206 @node Leaf Functions
2207 @subsection Handling Leaf Functions
2209 @cindex leaf functions
2210 @cindex functions, leaf
2211 On some machines, a leaf function (i.e., one which makes no calls) can run
2212 more efficiently if it does not make its own register window. Often this
2213 means it is required to receive its arguments in the registers where they
2214 are passed by the caller, instead of the registers where they would
2217 The special treatment for leaf functions generally applies only when
2218 other conditions are met; for example, often they may use only those
2219 registers for its own variables and temporaries. We use the term ``leaf
2220 function'' to mean a function that is suitable for this special
2221 handling, so that functions with no calls are not necessarily ``leaf
2224 GCC assigns register numbers before it knows whether the function is
2225 suitable for leaf function treatment. So it needs to renumber the
2226 registers in order to output a leaf function. The following macros
2229 @defmac LEAF_REGISTERS
2230 Name of a char vector, indexed by hard register number, which
2231 contains 1 for a register that is allowable in a candidate for leaf
2234 If leaf function treatment involves renumbering the registers, then the
2235 registers marked here should be the ones before renumbering---those that
2236 GCC would ordinarily allocate. The registers which will actually be
2237 used in the assembler code, after renumbering, should not be marked with 1
2240 Define this macro only if the target machine offers a way to optimize
2241 the treatment of leaf functions.
2244 @defmac LEAF_REG_REMAP (@var{regno})
2245 A C expression whose value is the register number to which @var{regno}
2246 should be renumbered, when a function is treated as a leaf function.
2248 If @var{regno} is a register number which should not appear in a leaf
2249 function before renumbering, then the expression should yield @minus{}1, which
2250 will cause the compiler to abort.
2252 Define this macro only if the target machine offers a way to optimize the
2253 treatment of leaf functions, and registers need to be renumbered to do
2257 @findex current_function_is_leaf
2258 @findex current_function_uses_only_leaf_regs
2259 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2260 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2261 specially. They can test the C variable @code{current_function_is_leaf}
2262 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2263 set prior to local register allocation and is valid for the remaining
2264 compiler passes. They can also test the C variable
2265 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2266 functions which only use leaf registers.
2267 @code{current_function_uses_only_leaf_regs} is valid after all passes
2268 that modify the instructions have been run and is only useful if
2269 @code{LEAF_REGISTERS} is defined.
2270 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2271 @c of the next paragraph?! --mew 2feb93
2273 @node Stack Registers
2274 @subsection Registers That Form a Stack
2276 There are special features to handle computers where some of the
2277 ``registers'' form a stack. Stack registers are normally written by
2278 pushing onto the stack, and are numbered relative to the top of the
2281 Currently, GCC can only handle one group of stack-like registers, and
2282 they must be consecutively numbered. Furthermore, the existing
2283 support for stack-like registers is specific to the 80387 floating
2284 point coprocessor. If you have a new architecture that uses
2285 stack-like registers, you will need to do substantial work on
2286 @file{reg-stack.c} and write your machine description to cooperate
2287 with it, as well as defining these macros.
2290 Define this if the machine has any stack-like registers.
2293 @defmac STACK_REG_COVER_CLASS
2294 This is a cover class containing the stack registers. Define this if
2295 the machine has any stack-like registers.
2298 @defmac FIRST_STACK_REG
2299 The number of the first stack-like register. This one is the top
2303 @defmac LAST_STACK_REG
2304 The number of the last stack-like register. This one is the bottom of
2308 @node Register Classes
2309 @section Register Classes
2310 @cindex register class definitions
2311 @cindex class definitions, register
2313 On many machines, the numbered registers are not all equivalent.
2314 For example, certain registers may not be allowed for indexed addressing;
2315 certain registers may not be allowed in some instructions. These machine
2316 restrictions are described to the compiler using @dfn{register classes}.
2318 You define a number of register classes, giving each one a name and saying
2319 which of the registers belong to it. Then you can specify register classes
2320 that are allowed as operands to particular instruction patterns.
2324 In general, each register will belong to several classes. In fact, one
2325 class must be named @code{ALL_REGS} and contain all the registers. Another
2326 class must be named @code{NO_REGS} and contain no registers. Often the
2327 union of two classes will be another class; however, this is not required.
2329 @findex GENERAL_REGS
2330 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2331 terribly special about the name, but the operand constraint letters
2332 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2333 the same as @code{ALL_REGS}, just define it as a macro which expands
2336 Order the classes so that if class @var{x} is contained in class @var{y}
2337 then @var{x} has a lower class number than @var{y}.
2339 The way classes other than @code{GENERAL_REGS} are specified in operand
2340 constraints is through machine-dependent operand constraint letters.
2341 You can define such letters to correspond to various classes, then use
2342 them in operand constraints.
2344 You must define the narrowest register classes for allocatable
2345 registers, so that each class either has no subclasses, or that for
2346 some mode, the move cost between registers within the class is
2347 cheaper than moving a register in the class to or from memory
2350 You should define a class for the union of two classes whenever some
2351 instruction allows both classes. For example, if an instruction allows
2352 either a floating point (coprocessor) register or a general register for a
2353 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2354 which includes both of them. Otherwise you will get suboptimal code,
2355 or even internal compiler errors when reload cannot find a register in the
2356 class computed via @code{reg_class_subunion}.
2358 You must also specify certain redundant information about the register
2359 classes: for each class, which classes contain it and which ones are
2360 contained in it; for each pair of classes, the largest class contained
2363 When a value occupying several consecutive registers is expected in a
2364 certain class, all the registers used must belong to that class.
2365 Therefore, register classes cannot be used to enforce a requirement for
2366 a register pair to start with an even-numbered register. The way to
2367 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2369 Register classes used for input-operands of bitwise-and or shift
2370 instructions have a special requirement: each such class must have, for
2371 each fixed-point machine mode, a subclass whose registers can transfer that
2372 mode to or from memory. For example, on some machines, the operations for
2373 single-byte values (@code{QImode}) are limited to certain registers. When
2374 this is so, each register class that is used in a bitwise-and or shift
2375 instruction must have a subclass consisting of registers from which
2376 single-byte values can be loaded or stored. This is so that
2377 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2379 @deftp {Data type} {enum reg_class}
2380 An enumerated type that must be defined with all the register class names
2381 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2382 must be the last register class, followed by one more enumerated value,
2383 @code{LIM_REG_CLASSES}, which is not a register class but rather
2384 tells how many classes there are.
2386 Each register class has a number, which is the value of casting
2387 the class name to type @code{int}. The number serves as an index
2388 in many of the tables described below.
2391 @defmac N_REG_CLASSES
2392 The number of distinct register classes, defined as follows:
2395 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2399 @defmac REG_CLASS_NAMES
2400 An initializer containing the names of the register classes as C string
2401 constants. These names are used in writing some of the debugging dumps.
2404 @defmac REG_CLASS_CONTENTS
2405 An initializer containing the contents of the register classes, as integers
2406 which are bit masks. The @var{n}th integer specifies the contents of class
2407 @var{n}. The way the integer @var{mask} is interpreted is that
2408 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2410 When the machine has more than 32 registers, an integer does not suffice.
2411 Then the integers are replaced by sub-initializers, braced groupings containing
2412 several integers. Each sub-initializer must be suitable as an initializer
2413 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2414 In this situation, the first integer in each sub-initializer corresponds to
2415 registers 0 through 31, the second integer to registers 32 through 63, and
2419 @defmac REGNO_REG_CLASS (@var{regno})
2420 A C expression whose value is a register class containing hard register
2421 @var{regno}. In general there is more than one such class; choose a class
2422 which is @dfn{minimal}, meaning that no smaller class also contains the
2426 @defmac BASE_REG_CLASS
2427 A macro whose definition is the name of the class to which a valid
2428 base register must belong. A base register is one used in an address
2429 which is the register value plus a displacement.
2432 @defmac MODE_BASE_REG_CLASS (@var{mode})
2433 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2434 the selection of a base register in a mode dependent manner. If
2435 @var{mode} is VOIDmode then it should return the same value as
2436 @code{BASE_REG_CLASS}.
2439 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2440 A C expression whose value is the register class to which a valid
2441 base register must belong in order to be used in a base plus index
2442 register address. You should define this macro if base plus index
2443 addresses have different requirements than other base register uses.
2446 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2447 A C expression whose value is the register class to which a valid
2448 base register for a memory reference in mode @var{mode} to address
2449 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2450 define the context in which the base register occurs. @var{outer_code} is
2451 the code of the immediately enclosing expression (@code{MEM} for the top level
2452 of an address, @code{ADDRESS} for something that occurs in an
2453 @code{address_operand}). @var{index_code} is the code of the corresponding
2454 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2457 @defmac INDEX_REG_CLASS
2458 A macro whose definition is the name of the class to which a valid
2459 index register must belong. An index register is one used in an
2460 address where its value is either multiplied by a scale factor or
2461 added to another register (as well as added to a displacement).
2464 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2465 A C expression which is nonzero if register number @var{num} is
2466 suitable for use as a base register in operand addresses.
2469 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2470 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2471 that expression may examine the mode of the memory reference in
2472 @var{mode}. You should define this macro if the mode of the memory
2473 reference affects whether a register may be used as a base register. If
2474 you define this macro, the compiler will use it instead of
2475 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2476 addresses that appear outside a @code{MEM}, i.e., as an
2477 @code{address_operand}.
2480 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2481 A C expression which is nonzero if register number @var{num} is suitable for
2482 use as a base register in base plus index operand addresses, accessing
2483 memory in mode @var{mode}. It may be either a suitable hard register or a
2484 pseudo register that has been allocated such a hard register. You should
2485 define this macro if base plus index addresses have different requirements
2486 than other base register uses.
2488 Use of this macro is deprecated; please use the more general
2489 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2492 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2493 A C expression which is nonzero if register number @var{num} is
2494 suitable for use as a base register in operand addresses, accessing
2495 memory in mode @var{mode} in address space @var{address_space}.
2496 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2497 that that expression may examine the context in which the register
2498 appears in the memory reference. @var{outer_code} is the code of the
2499 immediately enclosing expression (@code{MEM} if at the top level of the
2500 address, @code{ADDRESS} for something that occurs in an
2501 @code{address_operand}). @var{index_code} is the code of the
2502 corresponding index expression if @var{outer_code} is @code{PLUS};
2503 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2504 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2507 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2508 A C expression which is nonzero if register number @var{num} is
2509 suitable for use as an index register in operand addresses. It may be
2510 either a suitable hard register or a pseudo register that has been
2511 allocated such a hard register.
2513 The difference between an index register and a base register is that
2514 the index register may be scaled. If an address involves the sum of
2515 two registers, neither one of them scaled, then either one may be
2516 labeled the ``base'' and the other the ``index''; but whichever
2517 labeling is used must fit the machine's constraints of which registers
2518 may serve in each capacity. The compiler will try both labelings,
2519 looking for one that is valid, and will reload one or both registers
2520 only if neither labeling works.
2523 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2524 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.
2527 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2528 A target hook that places additional restrictions on the register class
2529 to use when it is necessary to copy value @var{x} into a register in class
2530 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2531 another, smaller class.
2533 The default version of this hook always returns value of @code{rclass} argument.
2535 Sometimes returning a more restrictive class makes better code. For
2536 example, on the 68000, when @var{x} is an integer constant that is in range
2537 for a @samp{moveq} instruction, the value of this macro is always
2538 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2539 Requiring a data register guarantees that a @samp{moveq} will be used.
2541 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2542 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2543 loaded into some register class. By returning @code{NO_REGS} you can
2544 force @var{x} into a memory location. For example, rs6000 can load
2545 immediate values into general-purpose registers, but does not have an
2546 instruction for loading an immediate value into a floating-point
2547 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2548 @var{x} is a floating-point constant. If the constant can't be loaded
2549 into any kind of register, code generation will be better if
2550 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2551 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2553 If an insn has pseudos in it after register allocation, reload will go
2554 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2555 to find the best one. Returning @code{NO_REGS}, in this case, makes
2556 reload add a @code{!} in front of the constraint: the x86 back-end uses
2557 this feature to discourage usage of 387 registers when math is done in
2558 the SSE registers (and vice versa).
2561 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2562 A C expression that places additional restrictions on the register class
2563 to use when it is necessary to copy value @var{x} into a register in class
2564 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2565 another, smaller class. On many machines, the following definition is
2569 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2572 Sometimes returning a more restrictive class makes better code. For
2573 example, on the 68000, when @var{x} is an integer constant that is in range
2574 for a @samp{moveq} instruction, the value of this macro is always
2575 @code{DATA_REGS} as long as @var{class} includes the data registers.
2576 Requiring a data register guarantees that a @samp{moveq} will be used.
2578 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2579 @var{class} is if @var{x} is a legitimate constant which cannot be
2580 loaded into some register class. By returning @code{NO_REGS} you can
2581 force @var{x} into a memory location. For example, rs6000 can load
2582 immediate values into general-purpose registers, but does not have an
2583 instruction for loading an immediate value into a floating-point
2584 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2585 @var{x} is a floating-point constant. If the constant can't be loaded
2586 into any kind of register, code generation will be better if
2587 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2588 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2590 If an insn has pseudos in it after register allocation, reload will go
2591 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2592 to find the best one. Returning @code{NO_REGS}, in this case, makes
2593 reload add a @code{!} in front of the constraint: the x86 back-end uses
2594 this feature to discourage usage of 387 registers when math is done in
2595 the SSE registers (and vice versa).
2598 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2599 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2602 The default version of this hook always returns value of @code{rclass}
2605 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2606 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2609 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2610 A C expression that places additional restrictions on the register class
2611 to use when it is necessary to be able to hold a value of mode
2612 @var{mode} in a reload register for which class @var{class} would
2615 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2616 there are certain modes that simply can't go in certain reload classes.
2618 The value is a register class; perhaps @var{class}, or perhaps another,
2621 Don't define this macro unless the target machine has limitations which
2622 require the macro to do something nontrivial.
2625 @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})
2626 Many machines have some registers that cannot be copied directly to or
2627 from memory or even from other types of registers. An example is the
2628 @samp{MQ} register, which on most machines, can only be copied to or
2629 from general registers, but not memory. Below, we shall be using the
2630 term 'intermediate register' when a move operation cannot be performed
2631 directly, but has to be done by copying the source into the intermediate
2632 register first, and then copying the intermediate register to the
2633 destination. An intermediate register always has the same mode as
2634 source and destination. Since it holds the actual value being copied,
2635 reload might apply optimizations to re-use an intermediate register
2636 and eliding the copy from the source when it can determine that the
2637 intermediate register still holds the required value.
2639 Another kind of secondary reload is required on some machines which
2640 allow copying all registers to and from memory, but require a scratch
2641 register for stores to some memory locations (e.g., those with symbolic
2642 address on the RT, and those with certain symbolic address on the SPARC
2643 when compiling PIC)@. Scratch registers need not have the same mode
2644 as the value being copied, and usually hold a different value than
2645 that being copied. Special patterns in the md file are needed to
2646 describe how the copy is performed with the help of the scratch register;
2647 these patterns also describe the number, register class(es) and mode(s)
2648 of the scratch register(s).
2650 In some cases, both an intermediate and a scratch register are required.
2652 For input reloads, this target hook is called with nonzero @var{in_p},
2653 and @var{x} is an rtx that needs to be copied to a register of class
2654 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2655 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2656 needs to be copied to rtx @var{x} in @var{reload_mode}.
2658 If copying a register of @var{reload_class} from/to @var{x} requires
2659 an intermediate register, the hook @code{secondary_reload} should
2660 return the register class required for this intermediate register.
2661 If no intermediate register is required, it should return NO_REGS.
2662 If more than one intermediate register is required, describe the one
2663 that is closest in the copy chain to the reload register.
2665 If scratch registers are needed, you also have to describe how to
2666 perform the copy from/to the reload register to/from this
2667 closest intermediate register. Or if no intermediate register is
2668 required, but still a scratch register is needed, describe the
2669 copy from/to the reload register to/from the reload operand @var{x}.
2671 You do this by setting @code{sri->icode} to the instruction code of a pattern
2672 in the md file which performs the move. Operands 0 and 1 are the output
2673 and input of this copy, respectively. Operands from operand 2 onward are
2674 for scratch operands. These scratch operands must have a mode, and a
2675 single-register-class
2676 @c [later: or memory]
2679 When an intermediate register is used, the @code{secondary_reload}
2680 hook will be called again to determine how to copy the intermediate
2681 register to/from the reload operand @var{x}, so your hook must also
2682 have code to handle the register class of the intermediate operand.
2684 @c [For later: maybe we'll allow multi-alternative reload patterns -
2685 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2686 @c and match the constraints of input and output to determine the required
2687 @c alternative. A restriction would be that constraints used to match
2688 @c against reloads registers would have to be written as register class
2689 @c constraints, or we need a new target macro / hook that tells us if an
2690 @c arbitrary constraint can match an unknown register of a given class.
2691 @c Such a macro / hook would also be useful in other places.]
2694 @var{x} might be a pseudo-register or a @code{subreg} of a
2695 pseudo-register, which could either be in a hard register or in memory.
2696 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2697 in memory and the hard register number if it is in a register.
2699 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2700 currently not supported. For the time being, you will have to continue
2701 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2703 @code{copy_cost} also uses this target hook to find out how values are
2704 copied. If you want it to include some extra cost for the need to allocate
2705 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2706 Or if two dependent moves are supposed to have a lower cost than the sum
2707 of the individual moves due to expected fortuitous scheduling and/or special
2708 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2711 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2712 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2713 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2714 These macros are obsolete, new ports should use the target hook
2715 @code{TARGET_SECONDARY_RELOAD} instead.
2717 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2718 target hook. Older ports still define these macros to indicate to the
2719 reload phase that it may
2720 need to allocate at least one register for a reload in addition to the
2721 register to contain the data. Specifically, if copying @var{x} to a
2722 register @var{class} in @var{mode} requires an intermediate register,
2723 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2724 largest register class all of whose registers can be used as
2725 intermediate registers or scratch registers.
2727 If copying a register @var{class} in @var{mode} to @var{x} requires an
2728 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2729 was supposed to be defined be defined to return the largest register
2730 class required. If the
2731 requirements for input and output reloads were the same, the macro
2732 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2735 The values returned by these macros are often @code{GENERAL_REGS}.
2736 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2737 can be directly copied to or from a register of @var{class} in
2738 @var{mode} without requiring a scratch register. Do not define this
2739 macro if it would always return @code{NO_REGS}.
2741 If a scratch register is required (either with or without an
2742 intermediate register), you were supposed to define patterns for
2743 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2744 (@pxref{Standard Names}. These patterns, which were normally
2745 implemented with a @code{define_expand}, should be similar to the
2746 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2749 These patterns need constraints for the reload register and scratch
2751 contain a single register class. If the original reload register (whose
2752 class is @var{class}) can meet the constraint given in the pattern, the
2753 value returned by these macros is used for the class of the scratch
2754 register. Otherwise, two additional reload registers are required.
2755 Their classes are obtained from the constraints in the insn pattern.
2757 @var{x} might be a pseudo-register or a @code{subreg} of a
2758 pseudo-register, which could either be in a hard register or in memory.
2759 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2760 in memory and the hard register number if it is in a register.
2762 These macros should not be used in the case where a particular class of
2763 registers can only be copied to memory and not to another class of
2764 registers. In that case, secondary reload registers are not needed and
2765 would not be helpful. Instead, a stack location must be used to perform
2766 the copy and the @code{mov@var{m}} pattern should use memory as an
2767 intermediate storage. This case often occurs between floating-point and
2771 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2772 Certain machines have the property that some registers cannot be copied
2773 to some other registers without using memory. Define this macro on
2774 those machines to be a C expression that is nonzero if objects of mode
2775 @var{m} in registers of @var{class1} can only be copied to registers of
2776 class @var{class2} by storing a register of @var{class1} into memory
2777 and loading that memory location into a register of @var{class2}.
2779 Do not define this macro if its value would always be zero.
2782 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2783 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2784 allocates a stack slot for a memory location needed for register copies.
2785 If this macro is defined, the compiler instead uses the memory location
2786 defined by this macro.
2788 Do not define this macro if you do not define
2789 @code{SECONDARY_MEMORY_NEEDED}.
2792 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2793 When the compiler needs a secondary memory location to copy between two
2794 registers of mode @var{mode}, it normally allocates sufficient memory to
2795 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2796 load operations in a mode that many bits wide and whose class is the
2797 same as that of @var{mode}.
2799 This is right thing to do on most machines because it ensures that all
2800 bits of the register are copied and prevents accesses to the registers
2801 in a narrower mode, which some machines prohibit for floating-point
2804 However, this default behavior is not correct on some machines, such as
2805 the DEC Alpha, that store short integers in floating-point registers
2806 differently than in integer registers. On those machines, the default
2807 widening will not work correctly and you must define this macro to
2808 suppress that widening in some cases. See the file @file{alpha.h} for
2811 Do not define this macro if you do not define
2812 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2813 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2816 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2817 A target hook which returns @code{true} if pseudos that have been assigned
2818 to registers of class @var{rclass} would likely be spilled because
2819 registers of @var{rclass} are needed for spill registers.
2821 The default version of this target hook returns @code{true} if @var{rclass}
2822 has exactly one register and @code{false} otherwise. On most machines, this
2823 default should be used. Only use this target hook to some other expression
2824 if pseudos allocated by @file{local-alloc.c} end up in memory because their
2825 hard registers were needed for spill registers. If this target hook returns
2826 @code{false} for those classes, those pseudos will only be allocated by
2827 @file{global.c}, which knows how to reallocate the pseudo to another
2828 register. If there would not be another register available for reallocation,
2829 you should not change the implementation of this target hook since
2830 the only effect of such implementation would be to slow down register
2834 @deftypefn {Target Hook} {unsigned char} TARGET_CLASS_MAX_NREGS (reg_class_t @var{rclass}, enum machine_mode @var{mode})
2835 A target hook returns the maximum number of consecutive registers
2836 of class @var{rclass} needed to hold a value of mode @var{mode}.
2838 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2839 the value returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass},
2840 @var{mode})} target hook should be the maximum value of
2841 @code{HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno}
2842 values in the class @var{rclass}.
2844 This target hook helps control the handling of multiple-word values
2847 The default version of this target hook returns the size of @var{mode}
2851 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2852 A C expression for the maximum number of consecutive registers
2853 of class @var{class} needed to hold a value of mode @var{mode}.
2855 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2856 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2857 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2858 @var{mode})} for all @var{regno} values in the class @var{class}.
2860 This macro helps control the handling of multiple-word values
2864 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2865 If defined, a C expression that returns nonzero for a @var{class} for which
2866 a change from mode @var{from} to mode @var{to} is invalid.
2868 For the example, loading 32-bit integer or floating-point objects into
2869 floating-point registers on the Alpha extends them to 64 bits.
2870 Therefore loading a 64-bit object and then storing it as a 32-bit object
2871 does not store the low-order 32 bits, as would be the case for a normal
2872 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2876 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2877 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2878 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2882 @node Old Constraints
2883 @section Obsolete Macros for Defining Constraints
2884 @cindex defining constraints, obsolete method
2885 @cindex constraints, defining, obsolete method
2887 Machine-specific constraints can be defined with these macros instead
2888 of the machine description constructs described in @ref{Define
2889 Constraints}. This mechanism is obsolete. New ports should not use
2890 it; old ports should convert to the new mechanism.
2892 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2893 For the constraint at the start of @var{str}, which starts with the letter
2894 @var{c}, return the length. This allows you to have register class /
2895 constant / extra constraints that are longer than a single letter;
2896 you don't need to define this macro if you can do with single-letter
2897 constraints only. The definition of this macro should use
2898 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2899 to handle specially.
2900 There are some sanity checks in genoutput.c that check the constraint lengths
2901 for the md file, so you can also use this macro to help you while you are
2902 transitioning from a byzantine single-letter-constraint scheme: when you
2903 return a negative length for a constraint you want to re-use, genoutput
2904 will complain about every instance where it is used in the md file.
2907 @defmac REG_CLASS_FROM_LETTER (@var{char})
2908 A C expression which defines the machine-dependent operand constraint
2909 letters for register classes. If @var{char} is such a letter, the
2910 value should be the register class corresponding to it. Otherwise,
2911 the value should be @code{NO_REGS}. The register letter @samp{r},
2912 corresponding to class @code{GENERAL_REGS}, will not be passed
2913 to this macro; you do not need to handle it.
2916 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2917 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2918 passed in @var{str}, so that you can use suffixes to distinguish between
2922 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2923 A C expression that defines the machine-dependent operand constraint
2924 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2925 particular ranges of integer values. If @var{c} is one of those
2926 letters, the expression should check that @var{value}, an integer, is in
2927 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2928 not one of those letters, the value should be 0 regardless of
2932 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2933 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2934 string passed in @var{str}, so that you can use suffixes to distinguish
2935 between different variants.
2938 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2939 A C expression that defines the machine-dependent operand constraint
2940 letters that specify particular ranges of @code{const_double} values
2941 (@samp{G} or @samp{H}).
2943 If @var{c} is one of those letters, the expression should check that
2944 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2945 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2946 letters, the value should be 0 regardless of @var{value}.
2948 @code{const_double} is used for all floating-point constants and for
2949 @code{DImode} fixed-point constants. A given letter can accept either
2950 or both kinds of values. It can use @code{GET_MODE} to distinguish
2951 between these kinds.
2954 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2955 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2956 string passed in @var{str}, so that you can use suffixes to distinguish
2957 between different variants.
2960 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2961 A C expression that defines the optional machine-dependent constraint
2962 letters that can be used to segregate specific types of operands, usually
2963 memory references, for the target machine. Any letter that is not
2964 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2965 @code{REG_CLASS_FROM_CONSTRAINT}
2966 may be used. Normally this macro will not be defined.
2968 If it is required for a particular target machine, it should return 1
2969 if @var{value} corresponds to the operand type represented by the
2970 constraint letter @var{c}. If @var{c} is not defined as an extra
2971 constraint, the value returned should be 0 regardless of @var{value}.
2973 For example, on the ROMP, load instructions cannot have their output
2974 in r0 if the memory reference contains a symbolic address. Constraint
2975 letter @samp{Q} is defined as representing a memory address that does
2976 @emph{not} contain a symbolic address. An alternative is specified with
2977 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2978 alternative specifies @samp{m} on the input and a register class that
2979 does not include r0 on the output.
2982 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2983 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2984 in @var{str}, so that you can use suffixes to distinguish between different
2988 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2989 A C expression that defines the optional machine-dependent constraint
2990 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2991 be treated like memory constraints by the reload pass.
2993 It should return 1 if the operand type represented by the constraint
2994 at the start of @var{str}, the first letter of which is the letter @var{c},
2995 comprises a subset of all memory references including
2996 all those whose address is simply a base register. This allows the reload
2997 pass to reload an operand, if it does not directly correspond to the operand
2998 type of @var{c}, by copying its address into a base register.
3000 For example, on the S/390, some instructions do not accept arbitrary
3001 memory references, but only those that do not make use of an index
3002 register. The constraint letter @samp{Q} is defined via
3003 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
3004 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
3005 a @samp{Q} constraint can handle any memory operand, because the
3006 reload pass knows it can be reloaded by copying the memory address
3007 into a base register if required. This is analogous to the way
3008 an @samp{o} constraint can handle any memory operand.
3011 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
3012 A C expression that defines the optional machine-dependent constraint
3013 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
3014 @code{EXTRA_CONSTRAINT_STR}, that should
3015 be treated like address constraints by the reload pass.
3017 It should return 1 if the operand type represented by the constraint
3018 at the start of @var{str}, which starts with the letter @var{c}, comprises
3019 a subset of all memory addresses including
3020 all those that consist of just a base register. This allows the reload
3021 pass to reload an operand, if it does not directly correspond to the operand
3022 type of @var{str}, by copying it into a base register.
3024 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3025 be used with the @code{address_operand} predicate. It is treated
3026 analogously to the @samp{p} constraint.
3029 @node Stack and Calling
3030 @section Stack Layout and Calling Conventions
3031 @cindex calling conventions
3033 @c prevent bad page break with this line
3034 This describes the stack layout and calling conventions.
3038 * Exception Handling::
3043 * Register Arguments::
3045 * Aggregate Return::
3050 * Stack Smashing Protection::
3054 @subsection Basic Stack Layout
3055 @cindex stack frame layout
3056 @cindex frame layout
3058 @c prevent bad page break with this line
3059 Here is the basic stack layout.
3061 @defmac STACK_GROWS_DOWNWARD
3062 Define this macro if pushing a word onto the stack moves the stack
3063 pointer to a smaller address.
3065 When we say, ``define this macro if @dots{}'', it means that the
3066 compiler checks this macro only with @code{#ifdef} so the precise
3067 definition used does not matter.
3070 @defmac STACK_PUSH_CODE
3071 This macro defines the operation used when something is pushed
3072 on the stack. In RTL, a push operation will be
3073 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3075 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3076 and @code{POST_INC}. Which of these is correct depends on
3077 the stack direction and on whether the stack pointer points
3078 to the last item on the stack or whether it points to the
3079 space for the next item on the stack.
3081 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3082 defined, which is almost always right, and @code{PRE_INC} otherwise,
3083 which is often wrong.
3086 @defmac FRAME_GROWS_DOWNWARD
3087 Define this macro to nonzero value if the addresses of local variable slots
3088 are at negative offsets from the frame pointer.
3091 @defmac ARGS_GROW_DOWNWARD
3092 Define this macro if successive arguments to a function occupy decreasing
3093 addresses on the stack.
3096 @defmac STARTING_FRAME_OFFSET
3097 Offset from the frame pointer to the first local variable slot to be allocated.
3099 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3100 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3101 Otherwise, it is found by adding the length of the first slot to the
3102 value @code{STARTING_FRAME_OFFSET}.
3103 @c i'm not sure if the above is still correct.. had to change it to get
3104 @c rid of an overfull. --mew 2feb93
3107 @defmac STACK_ALIGNMENT_NEEDED
3108 Define to zero to disable final alignment of the stack during reload.
3109 The nonzero default for this macro is suitable for most ports.
3111 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3112 is a register save block following the local block that doesn't require
3113 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3114 stack alignment and do it in the backend.
3117 @defmac STACK_POINTER_OFFSET
3118 Offset from the stack pointer register to the first location at which
3119 outgoing arguments are placed. If not specified, the default value of
3120 zero is used. This is the proper value for most machines.
3122 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3123 the first location at which outgoing arguments are placed.
3126 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3127 Offset from the argument pointer register to the first argument's
3128 address. On some machines it may depend on the data type of the
3131 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3132 the first argument's address.
3135 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3136 Offset from the stack pointer register to an item dynamically allocated
3137 on the stack, e.g., by @code{alloca}.
3139 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3140 length of the outgoing arguments. The default is correct for most
3141 machines. See @file{function.c} for details.
3144 @defmac INITIAL_FRAME_ADDRESS_RTX
3145 A C expression whose value is RTL representing the address of the initial
3146 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3147 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3148 default value will be used. Define this macro in order to make frame pointer
3149 elimination work in the presence of @code{__builtin_frame_address (count)} and
3150 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3153 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3154 A C expression whose value is RTL representing the address in a stack
3155 frame where the pointer to the caller's frame is stored. Assume that
3156 @var{frameaddr} is an RTL expression for the address of the stack frame
3159 If you don't define this macro, the default is to return the value
3160 of @var{frameaddr}---that is, the stack frame address is also the
3161 address of the stack word that points to the previous frame.
3164 @defmac SETUP_FRAME_ADDRESSES
3165 If defined, a C expression that produces the machine-specific code to
3166 setup the stack so that arbitrary frames can be accessed. For example,
3167 on the SPARC, we must flush all of the register windows to the stack
3168 before we can access arbitrary stack frames. You will seldom need to
3172 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3173 This target hook should return an rtx that is used to store
3174 the address of the current frame into the built in @code{setjmp} buffer.
3175 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3176 machines. One reason you may need to define this target hook is if
3177 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3180 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3181 A C expression whose value is RTL representing the value of the frame
3182 address for the current frame. @var{frameaddr} is the frame pointer
3183 of the current frame. This is used for __builtin_frame_address.
3184 You need only define this macro if the frame address is not the same
3185 as the frame pointer. Most machines do not need to define it.
3188 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3189 A C expression whose value is RTL representing the value of the return
3190 address for the frame @var{count} steps up from the current frame, after
3191 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3192 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3193 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3195 The value of the expression must always be the correct address when
3196 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3197 determine the return address of other frames.
3200 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3201 Define this if the return address of a particular stack frame is accessed
3202 from the frame pointer of the previous stack frame.
3205 @defmac INCOMING_RETURN_ADDR_RTX
3206 A C expression whose value is RTL representing the location of the
3207 incoming return address at the beginning of any function, before the
3208 prologue. This RTL is either a @code{REG}, indicating that the return
3209 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3212 You only need to define this macro if you want to support call frame
3213 debugging information like that provided by DWARF 2.
3215 If this RTL is a @code{REG}, you should also define
3216 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3219 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3220 A C expression whose value is an integer giving a DWARF 2 column
3221 number that may be used as an alternative return column. The column
3222 must not correspond to any gcc hard register (that is, it must not
3223 be in the range of @code{DWARF_FRAME_REGNUM}).
3225 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3226 general register, but an alternative column needs to be used for signal
3227 frames. Some targets have also used different frame return columns
3231 @defmac DWARF_ZERO_REG
3232 A C expression whose value is an integer giving a DWARF 2 register
3233 number that is considered to always have the value zero. This should
3234 only be defined if the target has an architected zero register, and
3235 someone decided it was a good idea to use that register number to
3236 terminate the stack backtrace. New ports should avoid this.
3239 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3240 This target hook allows the backend to emit frame-related insns that
3241 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3242 info engine will invoke it on insns of the form
3244 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3248 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3250 to let the backend emit the call frame instructions. @var{label} is
3251 the CFI label attached to the insn, @var{pattern} is the pattern of
3252 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3255 @defmac INCOMING_FRAME_SP_OFFSET
3256 A C expression whose value is an integer giving the offset, in bytes,
3257 from the value of the stack pointer register to the top of the stack
3258 frame at the beginning of any function, before the prologue. The top of
3259 the frame is defined to be the value of the stack pointer in the
3260 previous frame, just before the call instruction.
3262 You only need to define this macro if you want to support call frame
3263 debugging information like that provided by DWARF 2.
3266 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3267 A C expression whose value is an integer giving the offset, in bytes,
3268 from the argument pointer to the canonical frame address (cfa). The
3269 final value should coincide with that calculated by
3270 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3271 during virtual register instantiation.
3273 The default value for this macro is
3274 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3275 which is correct for most machines; in general, the arguments are found
3276 immediately before the stack frame. Note that this is not the case on
3277 some targets that save registers into the caller's frame, such as SPARC
3278 and rs6000, and so such targets need to define this macro.
3280 You only need to define this macro if the default is incorrect, and you
3281 want to support call frame debugging information like that provided by
3285 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3286 If defined, a C expression whose value is an integer giving the offset
3287 in bytes from the frame pointer to the canonical frame address (cfa).
3288 The final value should coincide with that calculated by
3289 @code{INCOMING_FRAME_SP_OFFSET}.
3291 Normally the CFA is calculated as an offset from the argument pointer,
3292 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3293 variable due to the ABI, this may not be possible. If this macro is
3294 defined, it implies that the virtual register instantiation should be
3295 based on the frame pointer instead of the argument pointer. Only one
3296 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3300 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3301 If defined, a C expression whose value is an integer giving the offset
3302 in bytes from the canonical frame address (cfa) to the frame base used
3303 in DWARF 2 debug information. The default is zero. A different value
3304 may reduce the size of debug information on some ports.
3307 @node Exception Handling
3308 @subsection Exception Handling Support
3309 @cindex exception handling
3311 @defmac EH_RETURN_DATA_REGNO (@var{N})
3312 A C expression whose value is the @var{N}th register number used for
3313 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3314 @var{N} registers are usable.
3316 The exception handling library routines communicate with the exception
3317 handlers via a set of agreed upon registers. Ideally these registers
3318 should be call-clobbered; it is possible to use call-saved registers,
3319 but may negatively impact code size. The target must support at least
3320 2 data registers, but should define 4 if there are enough free registers.
3322 You must define this macro if you want to support call frame exception
3323 handling like that provided by DWARF 2.
3326 @defmac EH_RETURN_STACKADJ_RTX
3327 A C expression whose value is RTL representing a location in which
3328 to store a stack adjustment to be applied before function return.
3329 This is used to unwind the stack to an exception handler's call frame.
3330 It will be assigned zero on code paths that return normally.
3332 Typically this is a call-clobbered hard register that is otherwise
3333 untouched by the epilogue, but could also be a stack slot.
3335 Do not define this macro if the stack pointer is saved and restored
3336 by the regular prolog and epilog code in the call frame itself; in
3337 this case, the exception handling library routines will update the
3338 stack location to be restored in place. Otherwise, you must define
3339 this macro if you want to support call frame exception handling like
3340 that provided by DWARF 2.
3343 @defmac EH_RETURN_HANDLER_RTX
3344 A C expression whose value is RTL representing a location in which
3345 to store the address of an exception handler to which we should
3346 return. It will not be assigned on code paths that return normally.
3348 Typically this is the location in the call frame at which the normal
3349 return address is stored. For targets that return by popping an
3350 address off the stack, this might be a memory address just below
3351 the @emph{target} call frame rather than inside the current call
3352 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3353 been assigned, so it may be used to calculate the location of the
3356 Some targets have more complex requirements than storing to an
3357 address calculable during initial code generation. In that case
3358 the @code{eh_return} instruction pattern should be used instead.
3360 If you want to support call frame exception handling, you must
3361 define either this macro or the @code{eh_return} instruction pattern.
3364 @defmac RETURN_ADDR_OFFSET
3365 If defined, an integer-valued C expression for which rtl will be generated
3366 to add it to the exception handler address before it is searched in the
3367 exception handling tables, and to subtract it again from the address before
3368 using it to return to the exception handler.
3371 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3372 This macro chooses the encoding of pointers embedded in the exception
3373 handling sections. If at all possible, this should be defined such
3374 that the exception handling section will not require dynamic relocations,
3375 and so may be read-only.
3377 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3378 @var{global} is true if the symbol may be affected by dynamic relocations.
3379 The macro should return a combination of the @code{DW_EH_PE_*} defines
3380 as found in @file{dwarf2.h}.
3382 If this macro is not defined, pointers will not be encoded but
3383 represented directly.
3386 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3387 This macro allows the target to emit whatever special magic is required
3388 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3389 Generic code takes care of pc-relative and indirect encodings; this must
3390 be defined if the target uses text-relative or data-relative encodings.
3392 This is a C statement that branches to @var{done} if the format was
3393 handled. @var{encoding} is the format chosen, @var{size} is the number
3394 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3398 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3399 This macro allows the target to add CPU and operating system specific
3400 code to the call-frame unwinder for use when there is no unwind data
3401 available. The most common reason to implement this macro is to unwind
3402 through signal frames.
3404 This macro is called from @code{uw_frame_state_for} in
3405 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3406 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3407 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3408 for the address of the code being executed and @code{context->cfa} for
3409 the stack pointer value. If the frame can be decoded, the register
3410 save addresses should be updated in @var{fs} and the macro should
3411 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3412 the macro should evaluate to @code{_URC_END_OF_STACK}.
3414 For proper signal handling in Java this macro is accompanied by
3415 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3418 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3419 This macro allows the target to add operating system specific code to the
3420 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3421 usually used for signal or interrupt frames.
3423 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3424 @var{context} is an @code{_Unwind_Context};
3425 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3426 for the abi and context in the @code{.unwabi} directive. If the
3427 @code{.unwabi} directive can be handled, the register save addresses should
3428 be updated in @var{fs}.
3431 @defmac TARGET_USES_WEAK_UNWIND_INFO
3432 A C expression that evaluates to true if the target requires unwind
3433 info to be given comdat linkage. Define it to be @code{1} if comdat
3434 linkage is necessary. The default is @code{0}.
3437 @node Stack Checking
3438 @subsection Specifying How Stack Checking is Done
3440 GCC will check that stack references are within the boundaries of the
3441 stack, if the option @option{-fstack-check} is specified, in one of
3446 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3447 will assume that you have arranged for full stack checking to be done
3448 at appropriate places in the configuration files. GCC will not do
3449 other special processing.
3452 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3453 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3454 that you have arranged for static stack checking (checking of the
3455 static stack frame of functions) to be done at appropriate places
3456 in the configuration files. GCC will only emit code to do dynamic
3457 stack checking (checking on dynamic stack allocations) using the third
3461 If neither of the above are true, GCC will generate code to periodically
3462 ``probe'' the stack pointer using the values of the macros defined below.
3465 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3466 GCC will change its allocation strategy for large objects if the option
3467 @option{-fstack-check} is specified: they will always be allocated
3468 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3470 @defmac STACK_CHECK_BUILTIN
3471 A nonzero value if stack checking is done by the configuration files in a
3472 machine-dependent manner. You should define this macro if stack checking
3473 is required by the ABI of your machine or if you would like to do stack
3474 checking in some more efficient way than the generic approach. The default
3475 value of this macro is zero.
3478 @defmac STACK_CHECK_STATIC_BUILTIN
3479 A nonzero value if static stack checking is done by the configuration files
3480 in a machine-dependent manner. You should define this macro if you would
3481 like to do static stack checking in some more efficient way than the generic
3482 approach. The default value of this macro is zero.
3485 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3486 An integer specifying the interval at which GCC must generate stack probe
3487 instructions, defined as 2 raised to this integer. You will normally
3488 define this macro so that the interval be no larger than the size of
3489 the ``guard pages'' at the end of a stack area. The default value
3490 of 12 (4096-byte interval) is suitable for most systems.
3493 @defmac STACK_CHECK_MOVING_SP
3494 An integer which is nonzero if GCC should move the stack pointer page by page
3495 when doing probes. This can be necessary on systems where the stack pointer
3496 contains the bottom address of the memory area accessible to the executing
3497 thread at any point in time. In this situation an alternate signal stack
3498 is required in order to be able to recover from a stack overflow. The
3499 default value of this macro is zero.
3502 @defmac STACK_CHECK_PROTECT
3503 The number of bytes of stack needed to recover from a stack overflow, for
3504 languages where such a recovery is supported. The default value of 75 words
3505 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3506 8192 bytes with other exception handling mechanisms should be adequate for
3510 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3511 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3512 in the opposite case.
3514 @defmac STACK_CHECK_MAX_FRAME_SIZE
3515 The maximum size of a stack frame, in bytes. GCC will generate probe
3516 instructions in non-leaf functions to ensure at least this many bytes of
3517 stack are available. If a stack frame is larger than this size, stack
3518 checking will not be reliable and GCC will issue a warning. The
3519 default is chosen so that GCC only generates one instruction on most
3520 systems. You should normally not change the default value of this macro.
3523 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3524 GCC uses this value to generate the above warning message. It
3525 represents the amount of fixed frame used by a function, not including
3526 space for any callee-saved registers, temporaries and user variables.
3527 You need only specify an upper bound for this amount and will normally
3528 use the default of four words.
3531 @defmac STACK_CHECK_MAX_VAR_SIZE
3532 The maximum size, in bytes, of an object that GCC will place in the
3533 fixed area of the stack frame when the user specifies
3534 @option{-fstack-check}.
3535 GCC computed the default from the values of the above macros and you will
3536 normally not need to override that default.
3540 @node Frame Registers
3541 @subsection Registers That Address the Stack Frame
3543 @c prevent bad page break with this line
3544 This discusses registers that address the stack frame.
3546 @defmac STACK_POINTER_REGNUM
3547 The register number of the stack pointer register, which must also be a
3548 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3549 the hardware determines which register this is.
3552 @defmac FRAME_POINTER_REGNUM
3553 The register number of the frame pointer register, which is used to
3554 access automatic variables in the stack frame. On some machines, the
3555 hardware determines which register this is. On other machines, you can
3556 choose any register you wish for this purpose.
3559 @defmac HARD_FRAME_POINTER_REGNUM
3560 On some machines the offset between the frame pointer and starting
3561 offset of the automatic variables is not known until after register
3562 allocation has been done (for example, because the saved registers are
3563 between these two locations). On those machines, define
3564 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3565 be used internally until the offset is known, and define
3566 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3567 used for the frame pointer.
3569 You should define this macro only in the very rare circumstances when it
3570 is not possible to calculate the offset between the frame pointer and
3571 the automatic variables until after register allocation has been
3572 completed. When this macro is defined, you must also indicate in your
3573 definition of @code{ELIMINABLE_REGS} how to eliminate
3574 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3575 or @code{STACK_POINTER_REGNUM}.
3577 Do not define this macro if it would be the same as
3578 @code{FRAME_POINTER_REGNUM}.
3581 @defmac ARG_POINTER_REGNUM
3582 The register number of the arg pointer register, which is used to access
3583 the function's argument list. On some machines, this is the same as the
3584 frame pointer register. On some machines, the hardware determines which
3585 register this is. On other machines, you can choose any register you
3586 wish for this purpose. If this is not the same register as the frame
3587 pointer register, then you must mark it as a fixed register according to
3588 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3589 (@pxref{Elimination}).
3592 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3593 Define this to a preprocessor constant that is nonzero if
3594 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3595 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3596 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3597 definition is not suitable for use in preprocessor conditionals.
3600 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3601 Define this to a preprocessor constant that is nonzero if
3602 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3603 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3604 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3605 definition is not suitable for use in preprocessor conditionals.
3608 @defmac RETURN_ADDRESS_POINTER_REGNUM
3609 The register number of the return address pointer register, which is used to
3610 access the current function's return address from the stack. On some
3611 machines, the return address is not at a fixed offset from the frame
3612 pointer or stack pointer or argument pointer. This register can be defined
3613 to point to the return address on the stack, and then be converted by
3614 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3616 Do not define this macro unless there is no other way to get the return
3617 address from the stack.
3620 @defmac STATIC_CHAIN_REGNUM
3621 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3622 Register numbers used for passing a function's static chain pointer. If
3623 register windows are used, the register number as seen by the called
3624 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3625 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3626 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3629 The static chain register need not be a fixed register.
3631 If the static chain is passed in memory, these macros should not be
3632 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3635 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl}, bool @var{incoming_p})
3636 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3637 targets that may use different static chain locations for different
3638 nested functions. This may be required if the target has function
3639 attributes that affect the calling conventions of the function and
3640 those calling conventions use different static chain locations.
3642 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3644 If the static chain is passed in memory, this hook should be used to
3645 provide rtx giving @code{mem} expressions that denote where they are stored.
3646 Often the @code{mem} expression as seen by the caller will be at an offset
3647 from the stack pointer and the @code{mem} expression as seen by the callee
3648 will be at an offset from the frame pointer.
3649 @findex stack_pointer_rtx
3650 @findex frame_pointer_rtx
3651 @findex arg_pointer_rtx
3652 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3653 @code{arg_pointer_rtx} will have been initialized and should be used
3654 to refer to those items.
3657 @defmac DWARF_FRAME_REGISTERS
3658 This macro specifies the maximum number of hard registers that can be
3659 saved in a call frame. This is used to size data structures used in
3660 DWARF2 exception handling.
3662 Prior to GCC 3.0, this macro was needed in order to establish a stable
3663 exception handling ABI in the face of adding new hard registers for ISA
3664 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3665 in the number of hard registers. Nevertheless, this macro can still be
3666 used to reduce the runtime memory requirements of the exception handling
3667 routines, which can be substantial if the ISA contains a lot of
3668 registers that are not call-saved.
3670 If this macro is not defined, it defaults to
3671 @code{FIRST_PSEUDO_REGISTER}.
3674 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3676 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3677 for backward compatibility in pre GCC 3.0 compiled code.
3679 If this macro is not defined, it defaults to
3680 @code{DWARF_FRAME_REGISTERS}.
3683 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3685 Define this macro if the target's representation for dwarf registers
3686 is different than the internal representation for unwind column.
3687 Given a dwarf register, this macro should return the internal unwind
3688 column number to use instead.
3690 See the PowerPC's SPE target for an example.
3693 @defmac DWARF_FRAME_REGNUM (@var{regno})
3695 Define this macro if the target's representation for dwarf registers
3696 used in .eh_frame or .debug_frame is different from that used in other
3697 debug info sections. Given a GCC hard register number, this macro
3698 should return the .eh_frame register number. The default is
3699 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3703 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3705 Define this macro to map register numbers held in the call frame info
3706 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3707 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3708 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3709 return @code{@var{regno}}.
3713 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3715 Define this macro if the target stores register values as
3716 @code{_Unwind_Word} type in unwind context. It should be defined if
3717 target register size is larger than the size of @code{void *}. The
3718 default is to store register values as @code{void *} type.
3722 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3724 Define this macro to be 1 if the target always uses extended unwind
3725 context with version, args_size and by_value fields. If it is undefined,
3726 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3727 defined and 0 otherwise.
3732 @subsection Eliminating Frame Pointer and Arg Pointer
3734 @c prevent bad page break with this line
3735 This is about eliminating the frame pointer and arg pointer.
3737 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3738 This target hook should return @code{true} if a function must have and use
3739 a frame pointer. This target hook is called in the reload pass. If its return
3740 value is @code{true} the function will have a frame pointer.
3742 This target hook can in principle examine the current function and decide
3743 according to the facts, but on most machines the constant @code{false} or the
3744 constant @code{true} suffices. Use @code{false} when the machine allows code
3745 to be generated with no frame pointer, and doing so saves some time or space.
3746 Use @code{true} when there is no possible advantage to avoiding a frame
3749 In certain cases, the compiler does not know how to produce valid code
3750 without a frame pointer. The compiler recognizes those cases and
3751 automatically gives the function a frame pointer regardless of what
3752 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3755 In a function that does not require a frame pointer, the frame pointer
3756 register can be allocated for ordinary usage, unless you mark it as a
3757 fixed register. See @code{FIXED_REGISTERS} for more information.
3759 Default return value is @code{false}.
3762 @findex get_frame_size
3763 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3764 A C statement to store in the variable @var{depth-var} the difference
3765 between the frame pointer and the stack pointer values immediately after
3766 the function prologue. The value would be computed from information
3767 such as the result of @code{get_frame_size ()} and the tables of
3768 registers @code{regs_ever_live} and @code{call_used_regs}.
3770 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3771 need not be defined. Otherwise, it must be defined even if
3772 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3773 case, you may set @var{depth-var} to anything.
3776 @defmac ELIMINABLE_REGS
3777 If defined, this macro specifies a table of register pairs used to
3778 eliminate unneeded registers that point into the stack frame. If it is not
3779 defined, the only elimination attempted by the compiler is to replace
3780 references to the frame pointer with references to the stack pointer.
3782 The definition of this macro is a list of structure initializations, each
3783 of which specifies an original and replacement register.
3785 On some machines, the position of the argument pointer is not known until
3786 the compilation is completed. In such a case, a separate hard register
3787 must be used for the argument pointer. This register can be eliminated by
3788 replacing it with either the frame pointer or the argument pointer,
3789 depending on whether or not the frame pointer has been eliminated.
3791 In this case, you might specify:
3793 #define ELIMINABLE_REGS \
3794 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3795 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3796 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3799 Note that the elimination of the argument pointer with the stack pointer is
3800 specified first since that is the preferred elimination.
3803 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3804 This target hook should returns @code{true} if the compiler is allowed to
3805 try to replace register number @var{from_reg} with register number
3806 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3807 is defined, and will usually be @code{true}, since most of the cases
3808 preventing register elimination are things that the compiler already
3811 Default return value is @code{true}.
3814 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3815 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3816 specifies the initial difference between the specified pair of
3817 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3821 @node Stack Arguments
3822 @subsection Passing Function Arguments on the Stack
3823 @cindex arguments on stack
3824 @cindex stack arguments
3826 The macros in this section control how arguments are passed
3827 on the stack. See the following section for other macros that
3828 control passing certain arguments in registers.
3830 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3831 This target hook returns @code{true} if an argument declared in a
3832 prototype as an integral type smaller than @code{int} should actually be
3833 passed as an @code{int}. In addition to avoiding errors in certain
3834 cases of mismatch, it also makes for better code on certain machines.
3835 The default is to not promote prototypes.
3839 A C expression. If nonzero, push insns will be used to pass
3841 If the target machine does not have a push instruction, set it to zero.
3842 That directs GCC to use an alternate strategy: to
3843 allocate the entire argument block and then store the arguments into
3844 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3847 @defmac PUSH_ARGS_REVERSED
3848 A C expression. If nonzero, function arguments will be evaluated from
3849 last to first, rather than from first to last. If this macro is not
3850 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3851 and args grow in opposite directions, and 0 otherwise.
3854 @defmac PUSH_ROUNDING (@var{npushed})
3855 A C expression that is the number of bytes actually pushed onto the
3856 stack when an instruction attempts to push @var{npushed} bytes.
3858 On some machines, the definition
3861 #define PUSH_ROUNDING(BYTES) (BYTES)
3865 will suffice. But on other machines, instructions that appear
3866 to push one byte actually push two bytes in an attempt to maintain
3867 alignment. Then the definition should be
3870 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3873 If the value of this macro has a type, it should be an unsigned type.
3876 @findex current_function_outgoing_args_size
3877 @defmac ACCUMULATE_OUTGOING_ARGS
3878 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3879 will be computed and placed into the variable
3880 @code{current_function_outgoing_args_size}. No space will be pushed
3881 onto the stack for each call; instead, the function prologue should
3882 increase the stack frame size by this amount.
3884 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3888 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3889 Define this macro if functions should assume that stack space has been
3890 allocated for arguments even when their values are passed in
3893 The value of this macro is the size, in bytes, of the area reserved for
3894 arguments passed in registers for the function represented by @var{fndecl},
3895 which can be zero if GCC is calling a library function.
3896 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3899 This space can be allocated by the caller, or be a part of the
3900 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3903 @c above is overfull. not sure what to do. --mew 5feb93 did
3904 @c something, not sure if it looks good. --mew 10feb93
3906 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3907 Define this to a nonzero value if it is the responsibility of the
3908 caller to allocate the area reserved for arguments passed in registers
3909 when calling a function of @var{fntype}. @var{fntype} may be NULL
3910 if the function called is a library function.
3912 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3913 whether the space for these arguments counts in the value of
3914 @code{current_function_outgoing_args_size}.
3917 @defmac STACK_PARMS_IN_REG_PARM_AREA
3918 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3919 stack parameters don't skip the area specified by it.
3920 @c i changed this, makes more sens and it should have taken care of the
3921 @c overfull.. not as specific, tho. --mew 5feb93
3923 Normally, when a parameter is not passed in registers, it is placed on the
3924 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3925 suppresses this behavior and causes the parameter to be passed on the
3926 stack in its natural location.
3929 @deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size})
3930 This target hook returns the number of bytes of its own arguments that
3931 a function pops on returning, or 0 if the function pops no arguments
3932 and the caller must therefore pop them all after the function returns.
3934 @var{fundecl} is a C variable whose value is a tree node that describes
3935 the function in question. Normally it is a node of type
3936 @code{FUNCTION_DECL} that describes the declaration of the function.
3937 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3939 @var{funtype} is a C variable whose value is a tree node that
3940 describes the function in question. Normally it is a node of type
3941 @code{FUNCTION_TYPE} that describes the data type of the function.
3942 From this it is possible to obtain the data types of the value and
3943 arguments (if known).
3945 When a call to a library function is being considered, @var{fundecl}
3946 will contain an identifier node for the library function. Thus, if
3947 you need to distinguish among various library functions, you can do so
3948 by their names. Note that ``library function'' in this context means
3949 a function used to perform arithmetic, whose name is known specially
3950 in the compiler and was not mentioned in the C code being compiled.
3952 @var{size} is the number of bytes of arguments passed on the
3953 stack. If a variable number of bytes is passed, it is zero, and
3954 argument popping will always be the responsibility of the calling function.
3956 On the VAX, all functions always pop their arguments, so the definition
3957 of this macro is @var{size}. On the 68000, using the standard
3958 calling convention, no functions pop their arguments, so the value of
3959 the macro is always 0 in this case. But an alternative calling
3960 convention is available in which functions that take a fixed number of
3961 arguments pop them but other functions (such as @code{printf}) pop
3962 nothing (the caller pops all). When this convention is in use,
3963 @var{funtype} is examined to determine whether a function takes a fixed
3964 number of arguments.
3967 @defmac CALL_POPS_ARGS (@var{cum})
3968 A C expression that should indicate the number of bytes a call sequence
3969 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3970 when compiling a function call.
3972 @var{cum} is the variable in which all arguments to the called function
3973 have been accumulated.
3975 On certain architectures, such as the SH5, a call trampoline is used
3976 that pops certain registers off the stack, depending on the arguments
3977 that have been passed to the function. Since this is a property of the
3978 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3982 @node Register Arguments
3983 @subsection Passing Arguments in Registers
3984 @cindex arguments in registers
3985 @cindex registers arguments
3987 This section describes the macros which let you control how various
3988 types of arguments are passed in registers or how they are arranged in
3991 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3992 Return an RTX indicating whether a function argument is passed in a
3993 register and if so, which register.
3995 The arguments are @var{ca}, which summarizes all the previous
3996 arguments; @var{mode}, the machine mode of the argument; @var{type},
3997 the data type of the argument as a tree node or 0 if that is not known
3998 (which happens for C support library functions); and @var{named},
3999 which is @code{true} for an ordinary argument and @code{false} for
4000 nameless arguments that correspond to @samp{@dots{}} in the called
4001 function's prototype. @var{type} can be an incomplete type if a
4002 syntax error has previously occurred.
4004 The return value is usually either a @code{reg} RTX for the hard
4005 register in which to pass the argument, or zero to pass the argument
4008 The value of the expression can also be a @code{parallel} RTX@. This is
4009 used when an argument is passed in multiple locations. The mode of the
4010 @code{parallel} should be the mode of the entire argument. The
4011 @code{parallel} holds any number of @code{expr_list} pairs; each one
4012 describes where part of the argument is passed. In each
4013 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
4014 register in which to pass this part of the argument, and the mode of the
4015 register RTX indicates how large this part of the argument is. The
4016 second operand of the @code{expr_list} is a @code{const_int} which gives
4017 the offset in bytes into the entire argument of where this part starts.
4018 As a special exception the first @code{expr_list} in the @code{parallel}
4019 RTX may have a first operand of zero. This indicates that the entire
4020 argument is also stored on the stack.
4022 The last time this hook is called, it is called with @code{MODE ==
4023 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
4024 pattern as operands 2 and 3 respectively.
4026 @cindex @file{stdarg.h} and register arguments
4027 The usual way to make the ISO library @file{stdarg.h} work on a
4028 machine where some arguments are usually passed in registers, is to
4029 cause nameless arguments to be passed on the stack instead. This is
4030 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
4031 @var{named} is @code{false}.
4033 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
4034 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
4035 You may use the hook @code{targetm.calls.must_pass_in_stack}
4036 in the definition of this macro to determine if this argument is of a
4037 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4038 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
4039 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4040 defined, the argument will be computed in the stack and then loaded into
4044 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, const_tree @var{type})
4045 This target hook should return @code{true} if we should not pass @var{type}
4046 solely in registers. The file @file{expr.h} defines a
4047 definition that is usually appropriate, refer to @file{expr.h} for additional
4051 @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})
4052 Define this hook if the target machine has ``register windows'', so
4053 that the register in which a function sees an arguments is not
4054 necessarily the same as the one in which the caller passed the
4057 For such machines, @code{TARGET_FUNCTION_ARG} computes the register in
4058 which the caller passes the value, and
4059 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4060 fashion to tell the function being called where the arguments will
4063 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4064 @code{TARGET_FUNCTION_ARG} serves both purposes.
4067 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4068 This target hook returns the number of bytes at the beginning of an
4069 argument that must be put in registers. The value must be zero for
4070 arguments that are passed entirely in registers or that are entirely
4071 pushed on the stack.
4073 On some machines, certain arguments must be passed partially in
4074 registers and partially in memory. On these machines, typically the
4075 first few words of arguments are passed in registers, and the rest
4076 on the stack. If a multi-word argument (a @code{double} or a
4077 structure) crosses that boundary, its first few words must be passed
4078 in registers and the rest must be pushed. This macro tells the
4079 compiler when this occurs, and how many bytes should go in registers.
4081 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4082 register to be used by the caller for this argument; likewise
4083 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4086 @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})
4087 This target hook should return @code{true} if an argument at the
4088 position indicated by @var{cum} should be passed by reference. This
4089 predicate is queried after target independent reasons for being
4090 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4092 If the hook returns true, a copy of that argument is made in memory and a
4093 pointer to the argument is passed instead of the argument itself.
4094 The pointer is passed in whatever way is appropriate for passing a pointer
4098 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4099 The function argument described by the parameters to this hook is
4100 known to be passed by reference. The hook should return true if the
4101 function argument should be copied by the callee instead of copied
4104 For any argument for which the hook returns true, if it can be
4105 determined that the argument is not modified, then a copy need
4108 The default version of this hook always returns false.
4111 @defmac CUMULATIVE_ARGS
4112 A C type for declaring a variable that is used as the first argument
4113 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4114 target machines, the type @code{int} suffices and can hold the number
4115 of bytes of argument so far.
4117 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4118 arguments that have been passed on the stack. The compiler has other
4119 variables to keep track of that. For target machines on which all
4120 arguments are passed on the stack, there is no need to store anything in
4121 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4122 should not be empty, so use @code{int}.
4125 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4126 If defined, this macro is called before generating any code for a
4127 function, but after the @var{cfun} descriptor for the function has been
4128 created. The back end may use this macro to update @var{cfun} to
4129 reflect an ABI other than that which would normally be used by default.
4130 If the compiler is generating code for a compiler-generated function,
4131 @var{fndecl} may be @code{NULL}.
4134 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4135 A C statement (sans semicolon) for initializing the variable
4136 @var{cum} for the state at the beginning of the argument list. The
4137 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4138 is the tree node for the data type of the function which will receive
4139 the args, or 0 if the args are to a compiler support library function.
4140 For direct calls that are not libcalls, @var{fndecl} contain the
4141 declaration node of the function. @var{fndecl} is also set when
4142 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4143 being compiled. @var{n_named_args} is set to the number of named
4144 arguments, including a structure return address if it is passed as a
4145 parameter, when making a call. When processing incoming arguments,
4146 @var{n_named_args} is set to @minus{}1.
4148 When processing a call to a compiler support library function,
4149 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4150 contains the name of the function, as a string. @var{libname} is 0 when
4151 an ordinary C function call is being processed. Thus, each time this
4152 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4153 never both of them at once.
4156 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4157 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4158 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4159 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4160 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4161 0)} is used instead.
4164 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4165 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4166 finding the arguments for the function being compiled. If this macro is
4167 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4169 The value passed for @var{libname} is always 0, since library routines
4170 with special calling conventions are never compiled with GCC@. The
4171 argument @var{libname} exists for symmetry with
4172 @code{INIT_CUMULATIVE_ARGS}.
4173 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4174 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4177 @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})
4178 This hook updates the summarizer variable pointed to by @var{ca} to
4179 advance past an argument in the argument list. The values @var{mode},
4180 @var{type} and @var{named} describe that argument. Once this is done,
4181 the variable @var{cum} is suitable for analyzing the @emph{following}
4182 argument with @code{TARGET_FUNCTION_ARG}, etc.
4184 This hook need not do anything if the argument in question was passed
4185 on the stack. The compiler knows how to track the amount of stack space
4186 used for arguments without any special help.
4189 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4190 If defined, a C expression that is the number of bytes to add to the
4191 offset of the argument passed in memory. This is needed for the SPU,
4192 which passes @code{char} and @code{short} arguments in the preferred
4193 slot that is in the middle of the quad word instead of starting at the
4197 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4198 If defined, a C expression which determines whether, and in which direction,
4199 to pad out an argument with extra space. The value should be of type
4200 @code{enum direction}: either @code{upward} to pad above the argument,
4201 @code{downward} to pad below, or @code{none} to inhibit padding.
4203 The @emph{amount} of padding is not controlled by this macro, but by the
4204 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
4205 always just enough to reach the next multiple of that boundary.
4207 This macro has a default definition which is right for most systems.
4208 For little-endian machines, the default is to pad upward. For
4209 big-endian machines, the default is to pad downward for an argument of
4210 constant size shorter than an @code{int}, and upward otherwise.
4213 @defmac PAD_VARARGS_DOWN
4214 If defined, a C expression which determines whether the default
4215 implementation of va_arg will attempt to pad down before reading the
4216 next argument, if that argument is smaller than its aligned space as
4217 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4218 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4221 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4222 Specify padding for the last element of a block move between registers and
4223 memory. @var{first} is nonzero if this is the only element. Defining this
4224 macro allows better control of register function parameters on big-endian
4225 machines, without using @code{PARALLEL} rtl. In particular,
4226 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4227 registers, as there is no longer a "wrong" part of a register; For example,
4228 a three byte aggregate may be passed in the high part of a register if so
4232 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4233 This hook returns the alignment boundary, in bits, of an argument
4234 with the specified mode and type. The default hook returns
4235 @code{PARM_BOUNDARY} for all arguments.
4238 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_ROUND_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4239 Normally, the size of an argument is rounded up to @code{PARM_BOUNDARY},
4240 which is the default value for this hook. You can define this hook to
4241 return a different value if an argument size must be rounded to a larger
4245 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4246 A C expression that is nonzero if @var{regno} is the number of a hard
4247 register in which function arguments are sometimes passed. This does
4248 @emph{not} include implicit arguments such as the static chain and
4249 the structure-value address. On many machines, no registers can be
4250 used for this purpose since all function arguments are pushed on the
4254 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4255 This hook should return true if parameter of type @var{type} are passed
4256 as two scalar parameters. By default, GCC will attempt to pack complex
4257 arguments into the target's word size. Some ABIs require complex arguments
4258 to be split and treated as their individual components. For example, on
4259 AIX64, complex floats should be passed in a pair of floating point
4260 registers, even though a complex float would fit in one 64-bit floating
4263 The default value of this hook is @code{NULL}, which is treated as always
4267 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4268 This hook returns a type node for @code{va_list} for the target.
4269 The default version of the hook returns @code{void*}.
4272 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4273 This target hook is used in function @code{c_common_nodes_and_builtins}
4274 to iterate through the target specific builtin types for va_list. The
4275 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4276 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4278 The arguments @var{pname} and @var{ptree} are used to store the result of
4279 this macro and are set to the name of the va_list builtin type and its
4281 If the return value of this macro is zero, then there is no more element.
4282 Otherwise the @var{IDX} should be increased for the next call of this
4283 macro to iterate through all types.
4286 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4287 This hook returns the va_list type of the calling convention specified by
4289 The default version of this hook returns @code{va_list_type_node}.
4292 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4293 This hook returns the va_list type of the calling convention specified by the
4294 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4298 @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})
4299 This hook performs target-specific gimplification of
4300 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4301 arguments to @code{va_arg}; the latter two are as in
4302 @code{gimplify.c:gimplify_expr}.
4305 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4306 Define this to return nonzero if the port can handle pointers
4307 with machine mode @var{mode}. The default version of this
4308 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4311 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (struct ao_ref_s *@var{ref})
4312 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.
4315 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4316 Define this to return nonzero if the port is prepared to handle
4317 insns involving scalar mode @var{mode}. For a scalar mode to be
4318 considered supported, all the basic arithmetic and comparisons
4321 The default version of this hook returns true for any mode
4322 required to handle the basic C types (as defined by the port).
4323 Included here are the double-word arithmetic supported by the
4324 code in @file{optabs.c}.
4327 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4328 Define this to return nonzero if the port is prepared to handle
4329 insns involving vector mode @var{mode}. At the very least, it
4330 must have move patterns for this mode.
4333 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (enum machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4334 Return true if GCC should try to use a scalar mode to store an array
4335 of @var{nelems} elements, given that each element has mode @var{mode}.
4336 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4337 and allows GCC to use any defined integer mode.
4339 One use of this hook is to support vector load and store operations
4340 that operate on several homogeneous vectors. For example, ARM NEON
4341 has operations like:
4344 int8x8x3_t vld3_s8 (const int8_t *)
4347 where the return type is defined as:
4350 typedef struct int8x8x3_t
4356 If this hook allows @code{val} to have a scalar mode, then
4357 @code{int8x8x3_t} can have the same mode. GCC can then store
4358 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4361 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (enum machine_mode @var{mode})
4362 Define this to return nonzero for machine modes for which the port has
4363 small register classes. If this target hook returns nonzero for a given
4364 @var{mode}, the compiler will try to minimize the lifetime of registers
4365 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4366 In this case, the hook is expected to return nonzero if it returns nonzero
4369 On some machines, it is risky to let hard registers live across arbitrary
4370 insns. Typically, these machines have instructions that require values
4371 to be in specific registers (like an accumulator), and reload will fail
4372 if the required hard register is used for another purpose across such an
4375 Passes before reload do not know which hard registers will be used
4376 in an instruction, but the machine modes of the registers set or used in
4377 the instruction are already known. And for some machines, register
4378 classes are small for, say, integer registers but not for floating point
4379 registers. For example, the AMD x86-64 architecture requires specific
4380 registers for the legacy x86 integer instructions, but there are many
4381 SSE registers for floating point operations. On such targets, a good
4382 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4383 machine modes but zero for the SSE register classes.
4385 The default version of this hook returns false for any mode. It is always
4386 safe to redefine this hook to return with a nonzero value. But if you
4387 unnecessarily define it, you will reduce the amount of optimizations
4388 that can be performed in some cases. If you do not define this hook
4389 to return a nonzero value when it is required, the compiler will run out
4390 of spill registers and print a fatal error message.
4393 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
4394 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.
4398 @subsection How Scalar Function Values Are Returned
4399 @cindex return values in registers
4400 @cindex values, returned by functions
4401 @cindex scalars, returned as values
4403 This section discusses the macros that control returning scalars as
4404 values---values that can fit in registers.
4406 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4408 Define this to return an RTX representing the place where a function
4409 returns or receives a value of data type @var{ret_type}, a tree node
4410 representing a data type. @var{fn_decl_or_type} is a tree node
4411 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4412 function being called. If @var{outgoing} is false, the hook should
4413 compute the register in which the caller will see the return value.
4414 Otherwise, the hook should return an RTX representing the place where
4415 a function returns a value.
4417 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4418 (Actually, on most machines, scalar values are returned in the same
4419 place regardless of mode.) The value of the expression is usually a
4420 @code{reg} RTX for the hard register where the return value is stored.
4421 The value can also be a @code{parallel} RTX, if the return value is in
4422 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4423 @code{parallel} form. Note that the callee will populate every
4424 location specified in the @code{parallel}, but if the first element of
4425 the @code{parallel} contains the whole return value, callers will use
4426 that element as the canonical location and ignore the others. The m68k
4427 port uses this type of @code{parallel} to return pointers in both
4428 @samp{%a0} (the canonical location) and @samp{%d0}.
4430 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4431 the same promotion rules specified in @code{PROMOTE_MODE} if
4432 @var{valtype} is a scalar type.
4434 If the precise function being called is known, @var{func} is a tree
4435 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4436 pointer. This makes it possible to use a different value-returning
4437 convention for specific functions when all their calls are
4440 Some target machines have ``register windows'' so that the register in
4441 which a function returns its value is not the same as the one in which
4442 the caller sees the value. For such machines, you should return
4443 different RTX depending on @var{outgoing}.
4445 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4446 aggregate data types, because these are returned in another way. See
4447 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4450 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4451 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4452 a new target instead.
4455 @defmac LIBCALL_VALUE (@var{mode})
4456 A C expression to create an RTX representing the place where a library
4457 function returns a value of mode @var{mode}.
4459 Note that ``library function'' in this context means a compiler
4460 support routine, used to perform arithmetic, whose name is known
4461 specially by the compiler and was not mentioned in the C code being
4465 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (enum machine_mode @var{mode}, const_rtx @var{fun})
4466 Define this hook if the back-end needs to know the name of the libcall
4467 function in order to determine where the result should be returned.
4469 The mode of the result is given by @var{mode} and the name of the called
4470 library function is given by @var{fun}. The hook should return an RTX
4471 representing the place where the library function result will be returned.
4473 If this hook is not defined, then LIBCALL_VALUE will be used.
4476 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4477 A C expression that is nonzero if @var{regno} is the number of a hard
4478 register in which the values of called function may come back.
4480 A register whose use for returning values is limited to serving as the
4481 second of a pair (for a value of type @code{double}, say) need not be
4482 recognized by this macro. So for most machines, this definition
4486 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4489 If the machine has register windows, so that the caller and the called
4490 function use different registers for the return value, this macro
4491 should recognize only the caller's register numbers.
4493 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4494 for a new target instead.
4497 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4498 A target hook that return @code{true} if @var{regno} is the number of a hard
4499 register in which the values of called function may come back.
4501 A register whose use for returning values is limited to serving as the
4502 second of a pair (for a value of type @code{double}, say) need not be
4503 recognized by this target hook.
4505 If the machine has register windows, so that the caller and the called
4506 function use different registers for the return value, this target hook
4507 should recognize only the caller's register numbers.
4509 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4512 @defmac APPLY_RESULT_SIZE
4513 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4514 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4515 saving and restoring an arbitrary return value.
4518 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4519 This hook should return true if values of type @var{type} are returned
4520 at the most significant end of a register (in other words, if they are
4521 padded at the least significant end). You can assume that @var{type}
4522 is returned in a register; the caller is required to check this.
4524 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4525 be able to hold the complete return value. For example, if a 1-, 2-
4526 or 3-byte structure is returned at the most significant end of a
4527 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4531 @node Aggregate Return
4532 @subsection How Large Values Are Returned
4533 @cindex aggregates as return values
4534 @cindex large return values
4535 @cindex returning aggregate values
4536 @cindex structure value address
4538 When a function value's mode is @code{BLKmode} (and in some other
4539 cases), the value is not returned according to
4540 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4541 caller passes the address of a block of memory in which the value
4542 should be stored. This address is called the @dfn{structure value
4545 This section describes how to control returning structure values in
4548 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4549 This target hook should return a nonzero value to say to return the
4550 function value in memory, just as large structures are always returned.
4551 Here @var{type} will be the data type of the value, and @var{fntype}
4552 will be the type of the function doing the returning, or @code{NULL} for
4555 Note that values of mode @code{BLKmode} must be explicitly handled
4556 by this function. Also, the option @option{-fpcc-struct-return}
4557 takes effect regardless of this macro. On most systems, it is
4558 possible to leave the hook undefined; this causes a default
4559 definition to be used, whose value is the constant 1 for @code{BLKmode}
4560 values, and 0 otherwise.
4562 Do not use this hook to indicate that structures and unions should always
4563 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4567 @defmac DEFAULT_PCC_STRUCT_RETURN
4568 Define this macro to be 1 if all structure and union return values must be
4569 in memory. Since this results in slower code, this should be defined
4570 only if needed for compatibility with other compilers or with an ABI@.
4571 If you define this macro to be 0, then the conventions used for structure
4572 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4575 If not defined, this defaults to the value 1.
4578 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4579 This target hook should return the location of the structure value
4580 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4581 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4582 be @code{NULL}, for libcalls. You do not need to define this target
4583 hook if the address is always passed as an ``invisible'' first
4586 On some architectures the place where the structure value address
4587 is found by the called function is not the same place that the
4588 caller put it. This can be due to register windows, or it could
4589 be because the function prologue moves it to a different place.
4590 @var{incoming} is @code{1} or @code{2} when the location is needed in
4591 the context of the called function, and @code{0} in the context of
4594 If @var{incoming} is nonzero and the address is to be found on the
4595 stack, return a @code{mem} which refers to the frame pointer. If
4596 @var{incoming} is @code{2}, the result is being used to fetch the
4597 structure value address at the beginning of a function. If you need
4598 to emit adjusting code, you should do it at this point.
4601 @defmac PCC_STATIC_STRUCT_RETURN
4602 Define this macro if the usual system convention on the target machine
4603 for returning structures and unions is for the called function to return
4604 the address of a static variable containing the value.
4606 Do not define this if the usual system convention is for the caller to
4607 pass an address to the subroutine.
4609 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4610 nothing when you use @option{-freg-struct-return} mode.
4613 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4614 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.
4617 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_ARG_MODE (int @var{regno})
4618 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.
4622 @subsection Caller-Saves Register Allocation
4624 If you enable it, GCC can save registers around function calls. This
4625 makes it possible to use call-clobbered registers to hold variables that
4626 must live across calls.
4628 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4629 A C expression to determine whether it is worthwhile to consider placing
4630 a pseudo-register in a call-clobbered hard register and saving and
4631 restoring it around each function call. The expression should be 1 when
4632 this is worth doing, and 0 otherwise.
4634 If you don't define this macro, a default is used which is good on most
4635 machines: @code{4 * @var{calls} < @var{refs}}.
4638 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4639 A C expression specifying which mode is required for saving @var{nregs}
4640 of a pseudo-register in call-clobbered hard register @var{regno}. If
4641 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4642 returned. For most machines this macro need not be defined since GCC
4643 will select the smallest suitable mode.
4646 @node Function Entry
4647 @subsection Function Entry and Exit
4648 @cindex function entry and exit
4652 This section describes the macros that output function entry
4653 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4655 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4656 If defined, a function that outputs the assembler code for entry to a
4657 function. The prologue is responsible for setting up the stack frame,
4658 initializing the frame pointer register, saving registers that must be
4659 saved, and allocating @var{size} additional bytes of storage for the
4660 local variables. @var{size} is an integer. @var{file} is a stdio
4661 stream to which the assembler code should be output.
4663 The label for the beginning of the function need not be output by this
4664 macro. That has already been done when the macro is run.
4666 @findex regs_ever_live
4667 To determine which registers to save, the macro can refer to the array
4668 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4669 @var{r} is used anywhere within the function. This implies the function
4670 prologue should save register @var{r}, provided it is not one of the
4671 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4672 @code{regs_ever_live}.)
4674 On machines that have ``register windows'', the function entry code does
4675 not save on the stack the registers that are in the windows, even if
4676 they are supposed to be preserved by function calls; instead it takes
4677 appropriate steps to ``push'' the register stack, if any non-call-used
4678 registers are used in the function.
4680 @findex frame_pointer_needed
4681 On machines where functions may or may not have frame-pointers, the
4682 function entry code must vary accordingly; it must set up the frame
4683 pointer if one is wanted, and not otherwise. To determine whether a
4684 frame pointer is in wanted, the macro can refer to the variable
4685 @code{frame_pointer_needed}. The variable's value will be 1 at run
4686 time in a function that needs a frame pointer. @xref{Elimination}.
4688 The function entry code is responsible for allocating any stack space
4689 required for the function. This stack space consists of the regions
4690 listed below. In most cases, these regions are allocated in the
4691 order listed, with the last listed region closest to the top of the
4692 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4693 the highest address if it is not defined). You can use a different order
4694 for a machine if doing so is more convenient or required for
4695 compatibility reasons. Except in cases where required by standard
4696 or by a debugger, there is no reason why the stack layout used by GCC
4697 need agree with that used by other compilers for a machine.
4700 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4701 If defined, a function that outputs assembler code at the end of a
4702 prologue. This should be used when the function prologue is being
4703 emitted as RTL, and you have some extra assembler that needs to be
4704 emitted. @xref{prologue instruction pattern}.
4707 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4708 If defined, a function that outputs assembler code at the start of an
4709 epilogue. This should be used when the function epilogue is being
4710 emitted as RTL, and you have some extra assembler that needs to be
4711 emitted. @xref{epilogue instruction pattern}.
4714 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4715 If defined, a function that outputs the assembler code for exit from a
4716 function. The epilogue is responsible for restoring the saved
4717 registers and stack pointer to their values when the function was
4718 called, and returning control to the caller. This macro takes the
4719 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4720 registers to restore are determined from @code{regs_ever_live} and
4721 @code{CALL_USED_REGISTERS} in the same way.
4723 On some machines, there is a single instruction that does all the work
4724 of returning from the function. On these machines, give that
4725 instruction the name @samp{return} and do not define the macro
4726 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4728 Do not define a pattern named @samp{return} if you want the
4729 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4730 switches to control whether return instructions or epilogues are used,
4731 define a @samp{return} pattern with a validity condition that tests the
4732 target switches appropriately. If the @samp{return} pattern's validity
4733 condition is false, epilogues will be used.
4735 On machines where functions may or may not have frame-pointers, the
4736 function exit code must vary accordingly. Sometimes the code for these
4737 two cases is completely different. To determine whether a frame pointer
4738 is wanted, the macro can refer to the variable
4739 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4740 a function that needs a frame pointer.
4742 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4743 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4744 The C variable @code{current_function_is_leaf} is nonzero for such a
4745 function. @xref{Leaf Functions}.
4747 On some machines, some functions pop their arguments on exit while
4748 others leave that for the caller to do. For example, the 68020 when
4749 given @option{-mrtd} pops arguments in functions that take a fixed
4750 number of arguments.
4752 @findex current_function_pops_args
4753 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4754 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4755 needs to know what was decided. The number of bytes of the current
4756 function's arguments that this function should pop is available in
4757 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4762 @findex current_function_pretend_args_size
4763 A region of @code{current_function_pretend_args_size} bytes of
4764 uninitialized space just underneath the first argument arriving on the
4765 stack. (This may not be at the very start of the allocated stack region
4766 if the calling sequence has pushed anything else since pushing the stack
4767 arguments. But usually, on such machines, nothing else has been pushed
4768 yet, because the function prologue itself does all the pushing.) This
4769 region is used on machines where an argument may be passed partly in
4770 registers and partly in memory, and, in some cases to support the
4771 features in @code{<stdarg.h>}.
4774 An area of memory used to save certain registers used by the function.
4775 The size of this area, which may also include space for such things as
4776 the return address and pointers to previous stack frames, is
4777 machine-specific and usually depends on which registers have been used
4778 in the function. Machines with register windows often do not require
4782 A region of at least @var{size} bytes, possibly rounded up to an allocation
4783 boundary, to contain the local variables of the function. On some machines,
4784 this region and the save area may occur in the opposite order, with the
4785 save area closer to the top of the stack.
4788 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4789 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4790 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4791 argument lists of the function. @xref{Stack Arguments}.
4794 @defmac EXIT_IGNORE_STACK
4795 Define this macro as a C expression that is nonzero if the return
4796 instruction or the function epilogue ignores the value of the stack
4797 pointer; in other words, if it is safe to delete an instruction to
4798 adjust the stack pointer before a return from the function. The
4801 Note that this macro's value is relevant only for functions for which
4802 frame pointers are maintained. It is never safe to delete a final
4803 stack adjustment in a function that has no frame pointer, and the
4804 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4807 @defmac EPILOGUE_USES (@var{regno})
4808 Define this macro as a C expression that is nonzero for registers that are
4809 used by the epilogue or the @samp{return} pattern. The stack and frame
4810 pointer registers are already assumed to be used as needed.
4813 @defmac EH_USES (@var{regno})
4814 Define this macro as a C expression that is nonzero for registers that are
4815 used by the exception handling mechanism, and so should be considered live
4816 on entry to an exception edge.
4819 @defmac DELAY_SLOTS_FOR_EPILOGUE
4820 Define this macro if the function epilogue contains delay slots to which
4821 instructions from the rest of the function can be ``moved''. The
4822 definition should be a C expression whose value is an integer
4823 representing the number of delay slots there.
4826 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4827 A C expression that returns 1 if @var{insn} can be placed in delay
4828 slot number @var{n} of the epilogue.
4830 The argument @var{n} is an integer which identifies the delay slot now
4831 being considered (since different slots may have different rules of
4832 eligibility). It is never negative and is always less than the number
4833 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4834 If you reject a particular insn for a given delay slot, in principle, it
4835 may be reconsidered for a subsequent delay slot. Also, other insns may
4836 (at least in principle) be considered for the so far unfilled delay
4839 @findex current_function_epilogue_delay_list
4840 @findex final_scan_insn
4841 The insns accepted to fill the epilogue delay slots are put in an RTL
4842 list made with @code{insn_list} objects, stored in the variable
4843 @code{current_function_epilogue_delay_list}. The insn for the first
4844 delay slot comes first in the list. Your definition of the macro
4845 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4846 outputting the insns in this list, usually by calling
4847 @code{final_scan_insn}.
4849 You need not define this macro if you did not define
4850 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4853 @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})
4854 A function that outputs the assembler code for a thunk
4855 function, used to implement C++ virtual function calls with multiple
4856 inheritance. The thunk acts as a wrapper around a virtual function,
4857 adjusting the implicit object parameter before handing control off to
4860 First, emit code to add the integer @var{delta} to the location that
4861 contains the incoming first argument. Assume that this argument
4862 contains a pointer, and is the one used to pass the @code{this} pointer
4863 in C++. This is the incoming argument @emph{before} the function prologue,
4864 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4865 all other incoming arguments.
4867 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4868 made after adding @code{delta}. In particular, if @var{p} is the
4869 adjusted pointer, the following adjustment should be made:
4872 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4875 After the additions, emit code to jump to @var{function}, which is a
4876 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4877 not touch the return address. Hence returning from @var{FUNCTION} will
4878 return to whoever called the current @samp{thunk}.
4880 The effect must be as if @var{function} had been called directly with
4881 the adjusted first argument. This macro is responsible for emitting all
4882 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4883 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4885 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4886 have already been extracted from it.) It might possibly be useful on
4887 some targets, but probably not.
4889 If you do not define this macro, the target-independent code in the C++
4890 front end will generate a less efficient heavyweight thunk that calls
4891 @var{function} instead of jumping to it. The generic approach does
4892 not support varargs.
4895 @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})
4896 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4897 to output the assembler code for the thunk function specified by the
4898 arguments it is passed, and false otherwise. In the latter case, the
4899 generic approach will be used by the C++ front end, with the limitations
4904 @subsection Generating Code for Profiling
4905 @cindex profiling, code generation
4907 These macros will help you generate code for profiling.
4909 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4910 A C statement or compound statement to output to @var{file} some
4911 assembler code to call the profiling subroutine @code{mcount}.
4914 The details of how @code{mcount} expects to be called are determined by
4915 your operating system environment, not by GCC@. To figure them out,
4916 compile a small program for profiling using the system's installed C
4917 compiler and look at the assembler code that results.
4919 Older implementations of @code{mcount} expect the address of a counter
4920 variable to be loaded into some register. The name of this variable is
4921 @samp{LP} followed by the number @var{labelno}, so you would generate
4922 the name using @samp{LP%d} in a @code{fprintf}.
4925 @defmac PROFILE_HOOK
4926 A C statement or compound statement to output to @var{file} some assembly
4927 code to call the profiling subroutine @code{mcount} even the target does
4928 not support profiling.
4931 @defmac NO_PROFILE_COUNTERS
4932 Define this macro to be an expression with a nonzero value if the
4933 @code{mcount} subroutine on your system does not need a counter variable
4934 allocated for each function. This is true for almost all modern
4935 implementations. If you define this macro, you must not use the
4936 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4939 @defmac PROFILE_BEFORE_PROLOGUE
4940 Define this macro if the code for function profiling should come before
4941 the function prologue. Normally, the profiling code comes after.
4945 @subsection Permitting tail calls
4948 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4949 True if it is ok to do sibling call optimization for the specified
4950 call expression @var{exp}. @var{decl} will be the called function,
4951 or @code{NULL} if this is an indirect call.
4953 It is not uncommon for limitations of calling conventions to prevent
4954 tail calls to functions outside the current unit of translation, or
4955 during PIC compilation. The hook is used to enforce these restrictions,
4956 as the @code{sibcall} md pattern can not fail, or fall over to a
4957 ``normal'' call. The criteria for successful sibling call optimization
4958 may vary greatly between different architectures.
4961 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4962 Add any hard registers to @var{regs} that are live on entry to the
4963 function. This hook only needs to be defined to provide registers that
4964 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4965 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4966 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4967 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4970 @deftypefn {Target Hook} void TARGET_SET_UP_BY_PROLOGUE (struct hard_reg_set_container *@var{})
4971 This hook should add additional registers that are computed by the prologue to the hard regset for shrink-wrapping optimization purposes.
4974 @node Stack Smashing Protection
4975 @subsection Stack smashing protection
4976 @cindex stack smashing protection
4978 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4979 This hook returns a @code{DECL} node for the external variable to use
4980 for the stack protection guard. This variable is initialized by the
4981 runtime to some random value and is used to initialize the guard value
4982 that is placed at the top of the local stack frame. The type of this
4983 variable must be @code{ptr_type_node}.
4985 The default version of this hook creates a variable called
4986 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4989 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4990 This hook returns a tree expression that alerts the runtime that the
4991 stack protect guard variable has been modified. This expression should
4992 involve a call to a @code{noreturn} function.
4994 The default version of this hook invokes a function called
4995 @samp{__stack_chk_fail}, taking no arguments. This function is
4996 normally defined in @file{libgcc2.c}.
4999 @deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
5000 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
5004 @section Implementing the Varargs Macros
5005 @cindex varargs implementation
5007 GCC comes with an implementation of @code{<varargs.h>} and
5008 @code{<stdarg.h>} that work without change on machines that pass arguments
5009 on the stack. Other machines require their own implementations of
5010 varargs, and the two machine independent header files must have
5011 conditionals to include it.
5013 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
5014 the calling convention for @code{va_start}. The traditional
5015 implementation takes just one argument, which is the variable in which
5016 to store the argument pointer. The ISO implementation of
5017 @code{va_start} takes an additional second argument. The user is
5018 supposed to write the last named argument of the function here.
5020 However, @code{va_start} should not use this argument. The way to find
5021 the end of the named arguments is with the built-in functions described
5024 @defmac __builtin_saveregs ()
5025 Use this built-in function to save the argument registers in memory so
5026 that the varargs mechanism can access them. Both ISO and traditional
5027 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
5028 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
5030 On some machines, @code{__builtin_saveregs} is open-coded under the
5031 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
5032 other machines, it calls a routine written in assembler language,
5033 found in @file{libgcc2.c}.
5035 Code generated for the call to @code{__builtin_saveregs} appears at the
5036 beginning of the function, as opposed to where the call to
5037 @code{__builtin_saveregs} is written, regardless of what the code is.
5038 This is because the registers must be saved before the function starts
5039 to use them for its own purposes.
5040 @c i rewrote the first sentence above to fix an overfull hbox. --mew
5044 @defmac __builtin_next_arg (@var{lastarg})
5045 This builtin returns the address of the first anonymous stack
5046 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
5047 returns the address of the location above the first anonymous stack
5048 argument. Use it in @code{va_start} to initialize the pointer for
5049 fetching arguments from the stack. Also use it in @code{va_start} to
5050 verify that the second parameter @var{lastarg} is the last named argument
5051 of the current function.
5054 @defmac __builtin_classify_type (@var{object})
5055 Since each machine has its own conventions for which data types are
5056 passed in which kind of register, your implementation of @code{va_arg}
5057 has to embody these conventions. The easiest way to categorize the
5058 specified data type is to use @code{__builtin_classify_type} together
5059 with @code{sizeof} and @code{__alignof__}.
5061 @code{__builtin_classify_type} ignores the value of @var{object},
5062 considering only its data type. It returns an integer describing what
5063 kind of type that is---integer, floating, pointer, structure, and so on.
5065 The file @file{typeclass.h} defines an enumeration that you can use to
5066 interpret the values of @code{__builtin_classify_type}.
5069 These machine description macros help implement varargs:
5071 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5072 If defined, this hook produces the machine-specific code for a call to
5073 @code{__builtin_saveregs}. This code will be moved to the very
5074 beginning of the function, before any parameter access are made. The
5075 return value of this function should be an RTX that contains the value
5076 to use as the return of @code{__builtin_saveregs}.
5079 @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})
5080 This target hook offers an alternative to using
5081 @code{__builtin_saveregs} and defining the hook
5082 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5083 register arguments into the stack so that all the arguments appear to
5084 have been passed consecutively on the stack. Once this is done, you can
5085 use the standard implementation of varargs that works for machines that
5086 pass all their arguments on the stack.
5088 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5089 structure, containing the values that are obtained after processing the
5090 named arguments. The arguments @var{mode} and @var{type} describe the
5091 last named argument---its machine mode and its data type as a tree node.
5093 The target hook should do two things: first, push onto the stack all the
5094 argument registers @emph{not} used for the named arguments, and second,
5095 store the size of the data thus pushed into the @code{int}-valued
5096 variable pointed to by @var{pretend_args_size}. The value that you
5097 store here will serve as additional offset for setting up the stack
5100 Because you must generate code to push the anonymous arguments at
5101 compile time without knowing their data types,
5102 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5103 have just a single category of argument register and use it uniformly
5106 If the argument @var{second_time} is nonzero, it means that the
5107 arguments of the function are being analyzed for the second time. This
5108 happens for an inline function, which is not actually compiled until the
5109 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5110 not generate any instructions in this case.
5113 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca})
5114 Define this hook to return @code{true} if the location where a function
5115 argument is passed depends on whether or not it is a named argument.
5117 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5118 is set for varargs and stdarg functions. If this hook returns
5119 @code{true}, the @var{named} argument is always true for named
5120 arguments, and false for unnamed arguments. If it returns @code{false},
5121 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5122 then all arguments are treated as named. Otherwise, all named arguments
5123 except the last are treated as named.
5125 You need not define this hook if it always returns @code{false}.
5128 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca})
5129 If you need to conditionally change ABIs so that one works with
5130 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5131 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5132 defined, then define this hook to return @code{true} if
5133 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5134 Otherwise, you should not define this hook.
5138 @section Trampolines for Nested Functions
5139 @cindex trampolines for nested functions
5140 @cindex nested functions, trampolines for
5142 A @dfn{trampoline} is a small piece of code that is created at run time
5143 when the address of a nested function is taken. It normally resides on
5144 the stack, in the stack frame of the containing function. These macros
5145 tell GCC how to generate code to allocate and initialize a
5148 The instructions in the trampoline must do two things: load a constant
5149 address into the static chain register, and jump to the real address of
5150 the nested function. On CISC machines such as the m68k, this requires
5151 two instructions, a move immediate and a jump. Then the two addresses
5152 exist in the trampoline as word-long immediate operands. On RISC
5153 machines, it is often necessary to load each address into a register in
5154 two parts. Then pieces of each address form separate immediate
5157 The code generated to initialize the trampoline must store the variable
5158 parts---the static chain value and the function address---into the
5159 immediate operands of the instructions. On a CISC machine, this is
5160 simply a matter of copying each address to a memory reference at the
5161 proper offset from the start of the trampoline. On a RISC machine, it
5162 may be necessary to take out pieces of the address and store them
5165 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5166 This hook is called by @code{assemble_trampoline_template} to output,
5167 on the stream @var{f}, assembler code for a block of data that contains
5168 the constant parts of a trampoline. This code should not include a
5169 label---the label is taken care of automatically.
5171 If you do not define this hook, it means no template is needed
5172 for the target. Do not define this hook on systems where the block move
5173 code to copy the trampoline into place would be larger than the code
5174 to generate it on the spot.
5177 @defmac TRAMPOLINE_SECTION
5178 Return the section into which the trampoline template is to be placed
5179 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5182 @defmac TRAMPOLINE_SIZE
5183 A C expression for the size in bytes of the trampoline, as an integer.
5186 @defmac TRAMPOLINE_ALIGNMENT
5187 Alignment required for trampolines, in bits.
5189 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5190 is used for aligning trampolines.
5193 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5194 This hook is called to initialize a trampoline.
5195 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5196 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5197 RTX for the static chain value that should be passed to the function
5200 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5201 first thing this hook should do is emit a block move into @var{m_tramp}
5202 from the memory block returned by @code{assemble_trampoline_template}.
5203 Note that the block move need only cover the constant parts of the
5204 trampoline. If the target isolates the variable parts of the trampoline
5205 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5207 If the target requires any other actions, such as flushing caches or
5208 enabling stack execution, these actions should be performed after
5209 initializing the trampoline proper.
5212 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5213 This hook should perform any machine-specific adjustment in
5214 the address of the trampoline. Its argument contains the address of the
5215 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5216 the address to be used for a function call should be different from the
5217 address at which the template was stored, the different address should
5218 be returned; otherwise @var{addr} should be returned unchanged.
5219 If this hook is not defined, @var{addr} will be used for function calls.
5222 Implementing trampolines is difficult on many machines because they have
5223 separate instruction and data caches. Writing into a stack location
5224 fails to clear the memory in the instruction cache, so when the program
5225 jumps to that location, it executes the old contents.
5227 Here are two possible solutions. One is to clear the relevant parts of
5228 the instruction cache whenever a trampoline is set up. The other is to
5229 make all trampolines identical, by having them jump to a standard
5230 subroutine. The former technique makes trampoline execution faster; the
5231 latter makes initialization faster.
5233 To clear the instruction cache when a trampoline is initialized, define
5234 the following macro.
5236 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5237 If defined, expands to a C expression clearing the @emph{instruction
5238 cache} in the specified interval. The definition of this macro would
5239 typically be a series of @code{asm} statements. Both @var{beg} and
5240 @var{end} are both pointer expressions.
5243 To use a standard subroutine, define the following macro. In addition,
5244 you must make sure that the instructions in a trampoline fill an entire
5245 cache line with identical instructions, or else ensure that the
5246 beginning of the trampoline code is always aligned at the same point in
5247 its cache line. Look in @file{m68k.h} as a guide.
5249 @defmac TRANSFER_FROM_TRAMPOLINE
5250 Define this macro if trampolines need a special subroutine to do their
5251 work. The macro should expand to a series of @code{asm} statements
5252 which will be compiled with GCC@. They go in a library function named
5253 @code{__transfer_from_trampoline}.
5255 If you need to avoid executing the ordinary prologue code of a compiled
5256 C function when you jump to the subroutine, you can do so by placing a
5257 special label of your own in the assembler code. Use one @code{asm}
5258 statement to generate an assembler label, and another to make the label
5259 global. Then trampolines can use that label to jump directly to your
5260 special assembler code.
5264 @section Implicit Calls to Library Routines
5265 @cindex library subroutine names
5266 @cindex @file{libgcc.a}
5268 @c prevent bad page break with this line
5269 Here is an explanation of implicit calls to library routines.
5271 @defmac DECLARE_LIBRARY_RENAMES
5272 This macro, if defined, should expand to a piece of C code that will get
5273 expanded when compiling functions for libgcc.a. It can be used to
5274 provide alternate names for GCC's internal library functions if there
5275 are ABI-mandated names that the compiler should provide.
5278 @findex set_optab_libfunc
5279 @findex init_one_libfunc
5280 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5281 This hook should declare additional library routines or rename
5282 existing ones, using the functions @code{set_optab_libfunc} and
5283 @code{init_one_libfunc} defined in @file{optabs.c}.
5284 @code{init_optabs} calls this macro after initializing all the normal
5287 The default is to do nothing. Most ports don't need to define this hook.
5290 @deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX
5291 If false (the default), internal library routines start with two
5292 underscores. If set to true, these routines start with @code{__gnu_}
5293 instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This
5294 currently only affects functions defined in @file{libgcc2.c}. If this
5295 is set to true, the @file{tm.h} file must also
5296 @code{#define LIBGCC2_GNU_PREFIX}.
5299 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5300 This macro should return @code{true} if the library routine that
5301 implements the floating point comparison operator @var{comparison} in
5302 mode @var{mode} will return a boolean, and @var{false} if it will
5305 GCC's own floating point libraries return tristates from the
5306 comparison operators, so the default returns false always. Most ports
5307 don't need to define this macro.
5310 @defmac TARGET_LIB_INT_CMP_BIASED
5311 This macro should evaluate to @code{true} if the integer comparison
5312 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5313 operand is smaller than the second, 1 to indicate that they are equal,
5314 and 2 to indicate that the first operand is greater than the second.
5315 If this macro evaluates to @code{false} the comparison functions return
5316 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5317 in @file{libgcc.a}, you do not need to define this macro.
5320 @cindex @code{EDOM}, implicit usage
5323 The value of @code{EDOM} on the target machine, as a C integer constant
5324 expression. If you don't define this macro, GCC does not attempt to
5325 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5326 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5329 If you do not define @code{TARGET_EDOM}, then compiled code reports
5330 domain errors by calling the library function and letting it report the
5331 error. If mathematical functions on your system use @code{matherr} when
5332 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5333 that @code{matherr} is used normally.
5336 @cindex @code{errno}, implicit usage
5337 @defmac GEN_ERRNO_RTX
5338 Define this macro as a C expression to create an rtl expression that
5339 refers to the global ``variable'' @code{errno}. (On certain systems,
5340 @code{errno} may not actually be a variable.) If you don't define this
5341 macro, a reasonable default is used.
5344 @cindex C99 math functions, implicit usage
5345 @defmac TARGET_C99_FUNCTIONS
5346 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5347 @code{sinf} and similarly for other functions defined by C99 standard. The
5348 default is zero because a number of existing systems lack support for these
5349 functions in their runtime so this macro needs to be redefined to one on
5350 systems that do support the C99 runtime.
5353 @cindex sincos math function, implicit usage
5354 @defmac TARGET_HAS_SINCOS
5355 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5356 and @code{cos} with the same argument to a call to @code{sincos}. The
5357 default is zero. The target has to provide the following functions:
5359 void sincos(double x, double *sin, double *cos);
5360 void sincosf(float x, float *sin, float *cos);
5361 void sincosl(long double x, long double *sin, long double *cos);
5365 @defmac NEXT_OBJC_RUNTIME
5366 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
5367 by default. This calling convention involves passing the object, the selector
5368 and the method arguments all at once to the method-lookup library function.
5369 This is the usual setting when targeting Darwin/Mac OS X systems, which have
5370 the NeXT runtime installed.
5372 If the macro is set to 0, the "GNU" Objective-C message sending convention
5373 will be used by default. This convention passes just the object and the
5374 selector to the method-lookup function, which returns a pointer to the method.
5376 In either case, it remains possible to select code-generation for the alternate
5377 scheme, by means of compiler command line switches.
5380 @node Addressing Modes
5381 @section Addressing Modes
5382 @cindex addressing modes
5384 @c prevent bad page break with this line
5385 This is about addressing modes.
5387 @defmac HAVE_PRE_INCREMENT
5388 @defmacx HAVE_PRE_DECREMENT
5389 @defmacx HAVE_POST_INCREMENT
5390 @defmacx HAVE_POST_DECREMENT
5391 A C expression that is nonzero if the machine supports pre-increment,
5392 pre-decrement, post-increment, or post-decrement addressing respectively.
5395 @defmac HAVE_PRE_MODIFY_DISP
5396 @defmacx HAVE_POST_MODIFY_DISP
5397 A C expression that is nonzero if the machine supports pre- or
5398 post-address side-effect generation involving constants other than
5399 the size of the memory operand.
5402 @defmac HAVE_PRE_MODIFY_REG
5403 @defmacx HAVE_POST_MODIFY_REG
5404 A C expression that is nonzero if the machine supports pre- or
5405 post-address side-effect generation involving a register displacement.
5408 @defmac CONSTANT_ADDRESS_P (@var{x})
5409 A C expression that is 1 if the RTX @var{x} is a constant which
5410 is a valid address. On most machines the default definition of
5411 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5412 is acceptable, but a few machines are more restrictive as to which
5413 constant addresses are supported.
5416 @defmac CONSTANT_P (@var{x})
5417 @code{CONSTANT_P}, which is defined by target-independent code,
5418 accepts integer-values expressions whose values are not explicitly
5419 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5420 expressions and @code{const} arithmetic expressions, in addition to
5421 @code{const_int} and @code{const_double} expressions.
5424 @defmac MAX_REGS_PER_ADDRESS
5425 A number, the maximum number of registers that can appear in a valid
5426 memory address. Note that it is up to you to specify a value equal to
5427 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5431 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5432 A function that returns whether @var{x} (an RTX) is a legitimate memory
5433 address on the target machine for a memory operand of mode @var{mode}.
5435 Legitimate addresses are defined in two variants: a strict variant and a
5436 non-strict one. The @var{strict} parameter chooses which variant is
5437 desired by the caller.
5439 The strict variant is used in the reload pass. It must be defined so
5440 that any pseudo-register that has not been allocated a hard register is
5441 considered a memory reference. This is because in contexts where some
5442 kind of register is required, a pseudo-register with no hard register
5443 must be rejected. For non-hard registers, the strict variant should look
5444 up the @code{reg_renumber} array; it should then proceed using the hard
5445 register number in the array, or treat the pseudo as a memory reference
5446 if the array holds @code{-1}.
5448 The non-strict variant is used in other passes. It must be defined to
5449 accept all pseudo-registers in every context where some kind of
5450 register is required.
5452 Normally, constant addresses which are the sum of a @code{symbol_ref}
5453 and an integer are stored inside a @code{const} RTX to mark them as
5454 constant. Therefore, there is no need to recognize such sums
5455 specifically as legitimate addresses. Normally you would simply
5456 recognize any @code{const} as legitimate.
5458 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5459 sums that are not marked with @code{const}. It assumes that a naked
5460 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5461 naked constant sums as illegitimate addresses, so that none of them will
5462 be given to @code{PRINT_OPERAND_ADDRESS}.
5464 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5465 On some machines, whether a symbolic address is legitimate depends on
5466 the section that the address refers to. On these machines, define the
5467 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5468 into the @code{symbol_ref}, and then check for it here. When you see a
5469 @code{const}, you will have to look inside it to find the
5470 @code{symbol_ref} in order to determine the section. @xref{Assembler
5473 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5474 Some ports are still using a deprecated legacy substitute for
5475 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5479 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5483 and should @code{goto @var{label}} if the address @var{x} is a valid
5484 address on the target machine for a memory operand of mode @var{mode}.
5486 @findex REG_OK_STRICT
5487 Compiler source files that want to use the strict variant of this
5488 macro define the macro @code{REG_OK_STRICT}. You should use an
5489 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5490 that case and the non-strict variant otherwise.
5492 Using the hook is usually simpler because it limits the number of
5493 files that are recompiled when changes are made.
5496 @defmac TARGET_MEM_CONSTRAINT
5497 A single character to be used instead of the default @code{'m'}
5498 character for general memory addresses. This defines the constraint
5499 letter which matches the memory addresses accepted by
5500 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5501 support new address formats in your back end without changing the
5502 semantics of the @code{'m'} constraint. This is necessary in order to
5503 preserve functionality of inline assembly constructs using the
5504 @code{'m'} constraint.
5507 @defmac FIND_BASE_TERM (@var{x})
5508 A C expression to determine the base term of address @var{x},
5509 or to provide a simplified version of @var{x} from which @file{alias.c}
5510 can easily find the base term. This macro is used in only two places:
5511 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5513 It is always safe for this macro to not be defined. It exists so
5514 that alias analysis can understand machine-dependent addresses.
5516 The typical use of this macro is to handle addresses containing
5517 a label_ref or symbol_ref within an UNSPEC@.
5520 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode})
5521 This hook is given an invalid memory address @var{x} for an
5522 operand of mode @var{mode} and should try to return a valid memory
5525 @findex break_out_memory_refs
5526 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5527 and @var{oldx} will be the operand that was given to that function to produce
5530 The code of the hook should not alter the substructure of
5531 @var{x}. If it transforms @var{x} into a more legitimate form, it
5532 should return the new @var{x}.
5534 It is not necessary for this hook to come up with a legitimate address.
5535 The compiler has standard ways of doing so in all cases. In fact, it
5536 is safe to omit this hook or make it return @var{x} if it cannot find
5537 a valid way to legitimize the address. But often a machine-dependent
5538 strategy can generate better code.
5541 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5542 A C compound statement that attempts to replace @var{x}, which is an address
5543 that needs reloading, with a valid memory address for an operand of mode
5544 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5545 It is not necessary to define this macro, but it might be useful for
5546 performance reasons.
5548 For example, on the i386, it is sometimes possible to use a single
5549 reload register instead of two by reloading a sum of two pseudo
5550 registers into a register. On the other hand, for number of RISC
5551 processors offsets are limited so that often an intermediate address
5552 needs to be generated in order to address a stack slot. By defining
5553 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5554 generated for adjacent some stack slots can be made identical, and thus
5557 @emph{Note}: This macro should be used with caution. It is necessary
5558 to know something of how reload works in order to effectively use this,
5559 and it is quite easy to produce macros that build in too much knowledge
5560 of reload internals.
5562 @emph{Note}: This macro must be able to reload an address created by a
5563 previous invocation of this macro. If it fails to handle such addresses
5564 then the compiler may generate incorrect code or abort.
5567 The macro definition should use @code{push_reload} to indicate parts that
5568 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5569 suitable to be passed unaltered to @code{push_reload}.
5571 The code generated by this macro must not alter the substructure of
5572 @var{x}. If it transforms @var{x} into a more legitimate form, it
5573 should assign @var{x} (which will always be a C variable) a new value.
5574 This also applies to parts that you change indirectly by calling
5577 @findex strict_memory_address_p
5578 The macro definition may use @code{strict_memory_address_p} to test if
5579 the address has become legitimate.
5582 If you want to change only a part of @var{x}, one standard way of doing
5583 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5584 single level of rtl. Thus, if the part to be changed is not at the
5585 top level, you'll need to replace first the top level.
5586 It is not necessary for this macro to come up with a legitimate
5587 address; but often a machine-dependent strategy can generate better code.
5590 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr})
5591 This hook returns @code{true} if memory address @var{addr} can have
5592 different meanings depending on the machine mode of the memory
5593 reference it is used for or if the address is valid for some modes
5596 Autoincrement and autodecrement addresses typically have mode-dependent
5597 effects because the amount of the increment or decrement is the size
5598 of the operand being addressed. Some machines have other mode-dependent
5599 addresses. Many RISC machines have no mode-dependent addresses.
5601 You may assume that @var{addr} is a valid address for the machine.
5603 The default version of this hook returns @code{false}.
5606 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5607 A C statement or compound statement with a conditional @code{goto
5608 @var{label};} executed if memory address @var{x} (an RTX) can have
5609 different meanings depending on the machine mode of the memory
5610 reference it is used for or if the address is valid for some modes
5613 Autoincrement and autodecrement addresses typically have mode-dependent
5614 effects because the amount of the increment or decrement is the size
5615 of the operand being addressed. Some machines have other mode-dependent
5616 addresses. Many RISC machines have no mode-dependent addresses.
5618 You may assume that @var{addr} is a valid address for the machine.
5620 These are obsolete macros, replaced by the
5621 @code{TARGET_MODE_DEPENDENT_ADDRESS_P} target hook.
5624 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5625 This hook returns true if @var{x} is a legitimate constant for a
5626 @var{mode}-mode immediate operand on the target machine. You can assume that
5627 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5629 The default definition returns true.
5632 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5633 This hook is used to undo the possibly obfuscating effects of the
5634 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5635 macros. Some backend implementations of these macros wrap symbol
5636 references inside an @code{UNSPEC} rtx to represent PIC or similar
5637 addressing modes. This target hook allows GCC's optimizers to understand
5638 the semantics of these opaque @code{UNSPEC}s by converting them back
5639 into their original form.
5642 @deftypefn {Target Hook} bool TARGET_CONST_NOT_OK_FOR_DEBUG_P (rtx @var{x})
5643 This hook should return true if @var{x} should not be emitted into
5647 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (enum machine_mode @var{mode}, rtx @var{x})
5648 This hook should return true if @var{x} is of a form that cannot (or
5649 should not) be spilled to the constant pool. @var{mode} is the mode
5652 The default version of this hook returns false.
5654 The primary reason to define this hook is to prevent reload from
5655 deciding that a non-legitimate constant would be better reloaded
5656 from the constant pool instead of spilling and reloading a register
5657 holding the constant. This restriction is often true of addresses
5658 of TLS symbols for various targets.
5661 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, const_rtx @var{x})
5662 This hook should return true if pool entries for constant @var{x} can
5663 be placed in an @code{object_block} structure. @var{mode} is the mode
5666 The default version returns false for all constants.
5669 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (unsigned @var{fn}, bool @var{md_fn}, bool @var{sqrt})
5670 This hook should return the DECL of a function that implements reciprocal of
5671 the builtin function with builtin function code @var{fn}, or
5672 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5673 when @var{fn} is a code of a machine-dependent builtin function. When
5674 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5675 of a square root function are performed, and only reciprocals of @code{sqrt}
5679 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5680 This hook should return the DECL of a function @var{f} that given an
5681 address @var{addr} as an argument returns a mask @var{m} that can be
5682 used to extract from two vectors the relevant data that resides in
5683 @var{addr} in case @var{addr} is not properly aligned.
5685 The autovectorizer, when vectorizing a load operation from an address
5686 @var{addr} that may be unaligned, will generate two vector loads from
5687 the two aligned addresses around @var{addr}. It then generates a
5688 @code{REALIGN_LOAD} operation to extract the relevant data from the
5689 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5690 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5691 the third argument, @var{OFF}, defines how the data will be extracted
5692 from these two vectors: if @var{OFF} is 0, then the returned vector is
5693 @var{v2}; otherwise, the returned vector is composed from the last
5694 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5695 @var{OFF} elements of @var{v2}.
5697 If this hook is defined, the autovectorizer will generate a call
5698 to @var{f} (using the DECL tree that this hook returns) and will
5699 use the return value of @var{f} as the argument @var{OFF} to
5700 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5701 should comply with the semantics expected by @code{REALIGN_LOAD}
5703 If this hook is not defined, then @var{addr} will be used as
5704 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5705 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5708 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5709 This hook should return the DECL of a function @var{f} that implements
5710 widening multiplication of the even elements of two input vectors of type @var{x}.
5712 If this hook is defined, the autovectorizer will use it along with the
5713 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5714 widening multiplication in cases that the order of the results does not have to be
5715 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5716 @code{widen_mult_hi/lo} idioms will be used.
5719 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5720 This hook should return the DECL of a function @var{f} that implements
5721 widening multiplication of the odd elements of two input vectors of type @var{x}.
5723 If this hook is defined, the autovectorizer will use it along with the
5724 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5725 widening multiplication in cases that the order of the results does not have to be
5726 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5727 @code{widen_mult_hi/lo} idioms will be used.
5730 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5731 Returns cost of different scalar or vector statements for vectorization cost model.
5732 For vector memory operations the cost may depend on type (@var{vectype}) and
5733 misalignment value (@var{misalign}).
5736 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5737 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5740 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VEC_PERM_CONST_OK (enum @var{machine_mode}, const unsigned char *@var{sel})
5741 Return true if a vector created for @code{vec_perm_const} is valid.
5744 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5745 This hook should return the DECL of a function that implements conversion of the
5746 input vector of type @var{src_type} to type @var{dest_type}.
5747 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5748 specifies how the conversion is to be applied
5749 (truncation, rounding, etc.).
5751 If this hook is defined, the autovectorizer will use the
5752 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5753 conversion. Otherwise, it will return @code{NULL_TREE}.
5756 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5757 This hook should return the decl of a function that implements the
5758 vectorized variant of the builtin function with builtin function code
5759 @var{code} or @code{NULL_TREE} if such a function is not available.
5760 The value of @var{fndecl} is the builtin function declaration. The
5761 return type of the vectorized function shall be of vector type
5762 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5765 @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})
5766 This hook should return true if the target supports misaligned vector
5767 store/load of a specific factor denoted in the @var{misalignment}
5768 parameter. The vector store/load should be of machine mode @var{mode} and
5769 the elements in the vectors should be of type @var{type}. @var{is_packed}
5770 parameter is true if the memory access is defined in a packed struct.
5773 @deftypefn {Target Hook} {enum machine_mode} TARGET_VECTORIZE_PREFERRED_SIMD_MODE (enum machine_mode @var{mode})
5774 This hook should return the preferred mode for vectorizing scalar
5775 mode @var{mode}. The default is
5776 equal to @code{word_mode}, because the vectorizer can do some
5777 transformations even in absence of specialized @acronym{SIMD} hardware.
5780 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void)
5781 This hook should return a mask of sizes that should be iterated over
5782 after trying to autovectorize using the vector size derived from the
5783 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5784 The default is zero which means to not iterate over other vector sizes.
5787 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_TM_LOAD (tree)
5788 This hook should return the built-in decl needed to load a vector of the given type within a transaction.
5791 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_TM_STORE (tree)
5792 This hook should return the built-in decl needed to store a vector of the given type within a transaction.
5795 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_GATHER (const_tree @var{mem_vectype}, const_tree @var{index_type}, int @var{scale})
5796 Target builtin that implements vector gather operation. @var{mem_vectype}
5797 is the vector type of the load and @var{index_type} is scalar type of
5798 the index, scaled by @var{scale}.
5799 The default is @code{NULL_TREE} which means to not vectorize gather
5803 @node Anchored Addresses
5804 @section Anchored Addresses
5805 @cindex anchored addresses
5806 @cindex @option{-fsection-anchors}
5808 GCC usually addresses every static object as a separate entity.
5809 For example, if we have:
5813 int foo (void) @{ return a + b + c; @}
5816 the code for @code{foo} will usually calculate three separate symbolic
5817 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5818 it would be better to calculate just one symbolic address and access
5819 the three variables relative to it. The equivalent pseudocode would
5825 register int *xr = &x;
5826 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5830 (which isn't valid C). We refer to shared addresses like @code{x} as
5831 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5833 The hooks below describe the target properties that GCC needs to know
5834 in order to make effective use of section anchors. It won't use
5835 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5836 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5838 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5839 The minimum offset that should be applied to a section anchor.
5840 On most targets, it should be the smallest offset that can be
5841 applied to a base register while still giving a legitimate address
5842 for every mode. The default value is 0.
5845 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5846 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5847 offset that should be applied to section anchors. The default
5851 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5852 Write the assembly code to define section anchor @var{x}, which is a
5853 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5854 The hook is called with the assembly output position set to the beginning
5855 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5857 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5858 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5859 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5860 is @code{NULL}, which disables the use of section anchors altogether.
5863 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
5864 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5865 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5866 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5868 The default version is correct for most targets, but you might need to
5869 intercept this hook to handle things like target-specific attributes
5870 or target-specific sections.
5873 @node Condition Code
5874 @section Condition Code Status
5875 @cindex condition code status
5877 The macros in this section can be split in two families, according to the
5878 two ways of representing condition codes in GCC.
5880 The first representation is the so called @code{(cc0)} representation
5881 (@pxref{Jump Patterns}), where all instructions can have an implicit
5882 clobber of the condition codes. The second is the condition code
5883 register representation, which provides better schedulability for
5884 architectures that do have a condition code register, but on which
5885 most instructions do not affect it. The latter category includes
5888 The implicit clobbering poses a strong restriction on the placement of
5889 the definition and use of the condition code, which need to be in adjacent
5890 insns for machines using @code{(cc0)}. This can prevent important
5891 optimizations on some machines. For example, on the IBM RS/6000, there
5892 is a delay for taken branches unless the condition code register is set
5893 three instructions earlier than the conditional branch. The instruction
5894 scheduler cannot perform this optimization if it is not permitted to
5895 separate the definition and use of the condition code register.
5897 For this reason, it is possible and suggested to use a register to
5898 represent the condition code for new ports. If there is a specific
5899 condition code register in the machine, use a hard register. If the
5900 condition code or comparison result can be placed in any general register,
5901 or if there are multiple condition registers, use a pseudo register.
5902 Registers used to store the condition code value will usually have a mode
5903 that is in class @code{MODE_CC}.
5905 Alternatively, you can use @code{BImode} if the comparison operator is
5906 specified already in the compare instruction. In this case, you are not
5907 interested in most macros in this section.
5910 * CC0 Condition Codes:: Old style representation of condition codes.
5911 * MODE_CC Condition Codes:: Modern representation of condition codes.
5912 * Cond Exec Macros:: Macros to control conditional execution.
5915 @node CC0 Condition Codes
5916 @subsection Representation of condition codes using @code{(cc0)}
5920 The file @file{conditions.h} defines a variable @code{cc_status} to
5921 describe how the condition code was computed (in case the interpretation of
5922 the condition code depends on the instruction that it was set by). This
5923 variable contains the RTL expressions on which the condition code is
5924 currently based, and several standard flags.
5926 Sometimes additional machine-specific flags must be defined in the machine
5927 description header file. It can also add additional machine-specific
5928 information by defining @code{CC_STATUS_MDEP}.
5930 @defmac CC_STATUS_MDEP
5931 C code for a data type which is used for declaring the @code{mdep}
5932 component of @code{cc_status}. It defaults to @code{int}.
5934 This macro is not used on machines that do not use @code{cc0}.
5937 @defmac CC_STATUS_MDEP_INIT
5938 A C expression to initialize the @code{mdep} field to ``empty''.
5939 The default definition does nothing, since most machines don't use
5940 the field anyway. If you want to use the field, you should probably
5941 define this macro to initialize it.
5943 This macro is not used on machines that do not use @code{cc0}.
5946 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5947 A C compound statement to set the components of @code{cc_status}
5948 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5949 this macro's responsibility to recognize insns that set the condition
5950 code as a byproduct of other activity as well as those that explicitly
5953 This macro is not used on machines that do not use @code{cc0}.
5955 If there are insns that do not set the condition code but do alter
5956 other machine registers, this macro must check to see whether they
5957 invalidate the expressions that the condition code is recorded as
5958 reflecting. For example, on the 68000, insns that store in address
5959 registers do not set the condition code, which means that usually
5960 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5961 insns. But suppose that the previous insn set the condition code
5962 based on location @samp{a4@@(102)} and the current insn stores a new
5963 value in @samp{a4}. Although the condition code is not changed by
5964 this, it will no longer be true that it reflects the contents of
5965 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5966 @code{cc_status} in this case to say that nothing is known about the
5967 condition code value.
5969 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5970 with the results of peephole optimization: insns whose patterns are
5971 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5972 constants which are just the operands. The RTL structure of these
5973 insns is not sufficient to indicate what the insns actually do. What
5974 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5975 @code{CC_STATUS_INIT}.
5977 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5978 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5979 @samp{cc}. This avoids having detailed information about patterns in
5980 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5983 @node MODE_CC Condition Codes
5984 @subsection Representation of condition codes using registers
5988 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5989 On many machines, the condition code may be produced by other instructions
5990 than compares, for example the branch can use directly the condition
5991 code set by a subtract instruction. However, on some machines
5992 when the condition code is set this way some bits (such as the overflow
5993 bit) are not set in the same way as a test instruction, so that a different
5994 branch instruction must be used for some conditional branches. When
5995 this happens, use the machine mode of the condition code register to
5996 record different formats of the condition code register. Modes can
5997 also be used to record which compare instruction (e.g. a signed or an
5998 unsigned comparison) produced the condition codes.
6000 If other modes than @code{CCmode} are required, add them to
6001 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
6002 a mode given an operand of a compare. This is needed because the modes
6003 have to be chosen not only during RTL generation but also, for example,
6004 by instruction combination. The result of @code{SELECT_CC_MODE} should
6005 be consistent with the mode used in the patterns; for example to support
6006 the case of the add on the SPARC discussed above, we have the pattern
6010 [(set (reg:CC_NOOV 0)
6012 (plus:SI (match_operand:SI 0 "register_operand" "%r")
6013 (match_operand:SI 1 "arith_operand" "rI"))
6020 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
6021 for comparisons whose argument is a @code{plus}:
6024 #define SELECT_CC_MODE(OP,X,Y) \
6025 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
6026 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
6027 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
6028 || GET_CODE (X) == NEG) \
6029 ? CC_NOOVmode : CCmode))
6032 Another reason to use modes is to retain information on which operands
6033 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
6036 You should define this macro if and only if you define extra CC modes
6037 in @file{@var{machine}-modes.def}.
6040 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
6041 On some machines not all possible comparisons are defined, but you can
6042 convert an invalid comparison into a valid one. For example, the Alpha
6043 does not have a @code{GT} comparison, but you can use an @code{LT}
6044 comparison instead and swap the order of the operands.
6046 On such machines, define this macro to be a C statement to do any
6047 required conversions. @var{code} is the initial comparison code
6048 and @var{op0} and @var{op1} are the left and right operands of the
6049 comparison, respectively. You should modify @var{code}, @var{op0}, and
6050 @var{op1} as required.
6052 GCC will not assume that the comparison resulting from this macro is
6053 valid but will see if the resulting insn matches a pattern in the
6056 You need not define this macro if it would never change the comparison
6060 @defmac REVERSIBLE_CC_MODE (@var{mode})
6061 A C expression whose value is one if it is always safe to reverse a
6062 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6063 can ever return @var{mode} for a floating-point inequality comparison,
6064 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6066 You need not define this macro if it would always returns zero or if the
6067 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6068 For example, here is the definition used on the SPARC, where floating-point
6069 inequality comparisons are always given @code{CCFPEmode}:
6072 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
6076 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6077 A C expression whose value is reversed condition code of the @var{code} for
6078 comparison done in CC_MODE @var{mode}. The macro is used only in case
6079 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6080 machine has some non-standard way how to reverse certain conditionals. For
6081 instance in case all floating point conditions are non-trapping, compiler may
6082 freely convert unordered compares to ordered one. Then definition may look
6086 #define REVERSE_CONDITION(CODE, MODE) \
6087 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6088 : reverse_condition_maybe_unordered (CODE))
6092 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6093 On targets which do not use @code{(cc0)}, and which use a hard
6094 register rather than a pseudo-register to hold condition codes, the
6095 regular CSE passes are often not able to identify cases in which the
6096 hard register is set to a common value. Use this hook to enable a
6097 small pass which optimizes such cases. This hook should return true
6098 to enable this pass, and it should set the integers to which its
6099 arguments point to the hard register numbers used for condition codes.
6100 When there is only one such register, as is true on most systems, the
6101 integer pointed to by @var{p2} should be set to
6102 @code{INVALID_REGNUM}.
6104 The default version of this hook returns false.
6107 @deftypefn {Target Hook} {enum machine_mode} TARGET_CC_MODES_COMPATIBLE (enum machine_mode @var{m1}, enum machine_mode @var{m2})
6108 On targets which use multiple condition code modes in class
6109 @code{MODE_CC}, it is sometimes the case that a comparison can be
6110 validly done in more than one mode. On such a system, define this
6111 target hook to take two mode arguments and to return a mode in which
6112 both comparisons may be validly done. If there is no such mode,
6113 return @code{VOIDmode}.
6115 The default version of this hook checks whether the modes are the
6116 same. If they are, it returns that mode. If they are different, it
6117 returns @code{VOIDmode}.
6120 @node Cond Exec Macros
6121 @subsection Macros to control conditional execution
6122 @findex conditional execution
6125 There is one macro that may need to be defined for targets
6126 supporting conditional execution, independent of how they
6127 represent conditional branches.
6129 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6130 A C expression that returns true if the conditional execution predicate
6131 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6132 versa. Define this to return 0 if the target has conditional execution
6133 predicates that cannot be reversed safely. There is no need to validate
6134 that the arguments of op1 and op2 are the same, this is done separately.
6135 If no expansion is specified, this macro is defined as follows:
6138 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6139 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6144 @section Describing Relative Costs of Operations
6145 @cindex costs of instructions
6146 @cindex relative costs
6147 @cindex speed of instructions
6149 These macros let you describe the relative speed of various operations
6150 on the target machine.
6152 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6153 A C expression for the cost of moving data of mode @var{mode} from a
6154 register in class @var{from} to one in class @var{to}. The classes are
6155 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6156 value of 2 is the default; other values are interpreted relative to
6159 It is not required that the cost always equal 2 when @var{from} is the
6160 same as @var{to}; on some machines it is expensive to move between
6161 registers if they are not general registers.
6163 If reload sees an insn consisting of a single @code{set} between two
6164 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6165 classes returns a value of 2, reload does not check to ensure that the
6166 constraints of the insn are met. Setting a cost of other than 2 will
6167 allow reload to verify that the constraints are met. You should do this
6168 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6170 These macros are obsolete, new ports should use the target hook
6171 @code{TARGET_REGISTER_MOVE_COST} instead.
6174 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6175 This target hook should return the cost of moving data of mode @var{mode}
6176 from a register in class @var{from} to one in class @var{to}. The classes
6177 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6178 A value of 2 is the default; other values are interpreted relative to
6181 It is not required that the cost always equal 2 when @var{from} is the
6182 same as @var{to}; on some machines it is expensive to move between
6183 registers if they are not general registers.
6185 If reload sees an insn consisting of a single @code{set} between two
6186 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6187 classes returns a value of 2, reload does not check to ensure that the
6188 constraints of the insn are met. Setting a cost of other than 2 will
6189 allow reload to verify that the constraints are met. You should do this
6190 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6192 The default version of this function returns 2.
6195 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6196 A C expression for the cost of moving data of mode @var{mode} between a
6197 register of class @var{class} and memory; @var{in} is zero if the value
6198 is to be written to memory, nonzero if it is to be read in. This cost
6199 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6200 registers and memory is more expensive than between two registers, you
6201 should define this macro to express the relative cost.
6203 If you do not define this macro, GCC uses a default cost of 4 plus
6204 the cost of copying via a secondary reload register, if one is
6205 needed. If your machine requires a secondary reload register to copy
6206 between memory and a register of @var{class} but the reload mechanism is
6207 more complex than copying via an intermediate, define this macro to
6208 reflect the actual cost of the move.
6210 GCC defines the function @code{memory_move_secondary_cost} if
6211 secondary reloads are needed. It computes the costs due to copying via
6212 a secondary register. If your machine copies from memory using a
6213 secondary register in the conventional way but the default base value of
6214 4 is not correct for your machine, define this macro to add some other
6215 value to the result of that function. The arguments to that function
6216 are the same as to this macro.
6218 These macros are obsolete, new ports should use the target hook
6219 @code{TARGET_MEMORY_MOVE_COST} instead.
6222 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6223 This target hook should return the cost of moving data of mode @var{mode}
6224 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6225 if the value is to be written to memory, @code{true} if it is to be read in.
6226 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6227 If moving between registers and memory is more expensive than between two
6228 registers, you should add this target hook to express the relative cost.
6230 If you do not add this target hook, GCC uses a default cost of 4 plus
6231 the cost of copying via a secondary reload register, if one is
6232 needed. If your machine requires a secondary reload register to copy
6233 between memory and a register of @var{rclass} but the reload mechanism is
6234 more complex than copying via an intermediate, use this target hook to
6235 reflect the actual cost of the move.
6237 GCC defines the function @code{memory_move_secondary_cost} if
6238 secondary reloads are needed. It computes the costs due to copying via
6239 a secondary register. If your machine copies from memory using a
6240 secondary register in the conventional way but the default base value of
6241 4 is not correct for your machine, use this target hook to add some other
6242 value to the result of that function. The arguments to that function
6243 are the same as to this target hook.
6246 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6247 A C expression for the cost of a branch instruction. A value of 1 is
6248 the default; other values are interpreted relative to that. Parameter
6249 @var{speed_p} is true when the branch in question should be optimized
6250 for speed. When it is false, @code{BRANCH_COST} should return a value
6251 optimal for code size rather than performance. @var{predictable_p} is
6252 true for well-predicted branches. On many architectures the
6253 @code{BRANCH_COST} can be reduced then.
6256 Here are additional macros which do not specify precise relative costs,
6257 but only that certain actions are more expensive than GCC would
6260 @defmac SLOW_BYTE_ACCESS
6261 Define this macro as a C expression which is nonzero if accessing less
6262 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6263 faster than accessing a word of memory, i.e., if such access
6264 require more than one instruction or if there is no difference in cost
6265 between byte and (aligned) word loads.
6267 When this macro is not defined, the compiler will access a field by
6268 finding the smallest containing object; when it is defined, a fullword
6269 load will be used if alignment permits. Unless bytes accesses are
6270 faster than word accesses, using word accesses is preferable since it
6271 may eliminate subsequent memory access if subsequent accesses occur to
6272 other fields in the same word of the structure, but to different bytes.
6275 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6276 Define this macro to be the value 1 if memory accesses described by the
6277 @var{mode} and @var{alignment} parameters have a cost many times greater
6278 than aligned accesses, for example if they are emulated in a trap
6281 When this macro is nonzero, the compiler will act as if
6282 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6283 moves. This can cause significantly more instructions to be produced.
6284 Therefore, do not set this macro nonzero if unaligned accesses only add a
6285 cycle or two to the time for a memory access.
6287 If the value of this macro is always zero, it need not be defined. If
6288 this macro is defined, it should produce a nonzero value when
6289 @code{STRICT_ALIGNMENT} is nonzero.
6292 @defmac MOVE_RATIO (@var{speed})
6293 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6294 which a sequence of insns should be generated instead of a
6295 string move insn or a library call. Increasing the value will always
6296 make code faster, but eventually incurs high cost in increased code size.
6298 Note that on machines where the corresponding move insn is a
6299 @code{define_expand} that emits a sequence of insns, this macro counts
6300 the number of such sequences.
6302 The parameter @var{speed} is true if the code is currently being
6303 optimized for speed rather than size.
6305 If you don't define this, a reasonable default is used.
6308 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6309 A C expression used to determine whether @code{move_by_pieces} will be used to
6310 copy a chunk of memory, or whether some other block move mechanism
6311 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6312 than @code{MOVE_RATIO}.
6315 @defmac MOVE_MAX_PIECES
6316 A C expression used by @code{move_by_pieces} to determine the largest unit
6317 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6320 @defmac CLEAR_RATIO (@var{speed})
6321 The threshold of number of scalar move insns, @emph{below} which a sequence
6322 of insns should be generated to clear memory instead of a string clear insn
6323 or a library call. Increasing the value will always make code faster, but
6324 eventually incurs high cost in increased code size.
6326 The parameter @var{speed} is true if the code is currently being
6327 optimized for speed rather than size.
6329 If you don't define this, a reasonable default is used.
6332 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6333 A C expression used to determine whether @code{clear_by_pieces} will be used
6334 to clear a chunk of memory, or whether some other block clear mechanism
6335 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6336 than @code{CLEAR_RATIO}.
6339 @defmac SET_RATIO (@var{speed})
6340 The threshold of number of scalar move insns, @emph{below} which a sequence
6341 of insns should be generated to set memory to a constant value, instead of
6342 a block set insn or a library call.
6343 Increasing the value will always make code faster, but
6344 eventually incurs high cost in increased code size.
6346 The parameter @var{speed} is true if the code is currently being
6347 optimized for speed rather than size.
6349 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6352 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6353 A C expression used to determine whether @code{store_by_pieces} will be
6354 used to set a chunk of memory to a constant value, or whether some
6355 other mechanism will be used. Used by @code{__builtin_memset} when
6356 storing values other than constant zero.
6357 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6358 than @code{SET_RATIO}.
6361 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6362 A C expression used to determine whether @code{store_by_pieces} will be
6363 used to set a chunk of memory to a constant string value, or whether some
6364 other mechanism will be used. Used by @code{__builtin_strcpy} when
6365 called with a constant source string.
6366 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6367 than @code{MOVE_RATIO}.
6370 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6371 A C expression used to determine whether a load postincrement is a good
6372 thing to use for a given mode. Defaults to the value of
6373 @code{HAVE_POST_INCREMENT}.
6376 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6377 A C expression used to determine whether a load postdecrement is a good
6378 thing to use for a given mode. Defaults to the value of
6379 @code{HAVE_POST_DECREMENT}.
6382 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6383 A C expression used to determine whether a load preincrement is a good
6384 thing to use for a given mode. Defaults to the value of
6385 @code{HAVE_PRE_INCREMENT}.
6388 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6389 A C expression used to determine whether a load predecrement is a good
6390 thing to use for a given mode. Defaults to the value of
6391 @code{HAVE_PRE_DECREMENT}.
6394 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6395 A C expression used to determine whether a store postincrement is a good
6396 thing to use for a given mode. Defaults to the value of
6397 @code{HAVE_POST_INCREMENT}.
6400 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6401 A C expression used to determine whether a store postdecrement is a good
6402 thing to use for a given mode. Defaults to the value of
6403 @code{HAVE_POST_DECREMENT}.
6406 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6407 This macro is used to determine whether a store preincrement is a good
6408 thing to use for a given mode. Defaults to the value of
6409 @code{HAVE_PRE_INCREMENT}.
6412 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6413 This macro is used to determine whether a store predecrement is a good
6414 thing to use for a given mode. Defaults to the value of
6415 @code{HAVE_PRE_DECREMENT}.
6418 @defmac NO_FUNCTION_CSE
6419 Define this macro if it is as good or better to call a constant
6420 function address than to call an address kept in a register.
6423 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6424 Define this macro if a non-short-circuit operation produced by
6425 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6426 @code{BRANCH_COST} is greater than or equal to the value 2.
6429 @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})
6430 This target hook describes the relative costs of RTL expressions.
6432 The cost may depend on the precise form of the expression, which is
6433 available for examination in @var{x}, and the fact that @var{x} appears
6434 as operand @var{opno} of an expression with rtx code @var{outer_code}.
6435 That is, the hook can assume that there is some rtx @var{y} such
6436 that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that
6437 either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or
6438 (b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}.
6440 @var{code} is @var{x}'s expression code---redundant, since it can be
6441 obtained with @code{GET_CODE (@var{x})}.
6443 In implementing this hook, you can use the construct
6444 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6447 On entry to the hook, @code{*@var{total}} contains a default estimate
6448 for the cost of the expression. The hook should modify this value as
6449 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6450 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6451 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6453 When optimizing for code size, i.e.@: when @code{speed} is
6454 false, this target hook should be used to estimate the relative
6455 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6457 The hook returns true when all subexpressions of @var{x} have been
6458 processed, and false when @code{rtx_cost} should recurse.
6461 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, bool @var{speed})
6462 This hook computes the cost of an addressing mode that contains
6463 @var{address}. If not defined, the cost is computed from
6464 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6466 For most CISC machines, the default cost is a good approximation of the
6467 true cost of the addressing mode. However, on RISC machines, all
6468 instructions normally have the same length and execution time. Hence
6469 all addresses will have equal costs.
6471 In cases where more than one form of an address is known, the form with
6472 the lowest cost will be used. If multiple forms have the same, lowest,
6473 cost, the one that is the most complex will be used.
6475 For example, suppose an address that is equal to the sum of a register
6476 and a constant is used twice in the same basic block. When this macro
6477 is not defined, the address will be computed in a register and memory
6478 references will be indirect through that register. On machines where
6479 the cost of the addressing mode containing the sum is no higher than
6480 that of a simple indirect reference, this will produce an additional
6481 instruction and possibly require an additional register. Proper
6482 specification of this macro eliminates this overhead for such machines.
6484 This hook is never called with an invalid address.
6486 On machines where an address involving more than one register is as
6487 cheap as an address computation involving only one register, defining
6488 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6489 be live over a region of code where only one would have been if
6490 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6491 should be considered in the definition of this macro. Equivalent costs
6492 should probably only be given to addresses with different numbers of
6493 registers on machines with lots of registers.
6497 @section Adjusting the Instruction Scheduler
6499 The instruction scheduler may need a fair amount of machine-specific
6500 adjustment in order to produce good code. GCC provides several target
6501 hooks for this purpose. It is usually enough to define just a few of
6502 them: try the first ones in this list first.
6504 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6505 This hook returns the maximum number of instructions that can ever
6506 issue at the same time on the target machine. The default is one.
6507 Although the insn scheduler can define itself the possibility of issue
6508 an insn on the same cycle, the value can serve as an additional
6509 constraint to issue insns on the same simulated processor cycle (see
6510 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6511 This value must be constant over the entire compilation. If you need
6512 it to vary depending on what the instructions are, you must use
6513 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6516 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6517 This hook is executed by the scheduler after it has scheduled an insn
6518 from the ready list. It should return the number of insns which can
6519 still be issued in the current cycle. The default is
6520 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6521 @code{USE}, which normally are not counted against the issue rate.
6522 You should define this hook if some insns take more machine resources
6523 than others, so that fewer insns can follow them in the same cycle.
6524 @var{file} is either a null pointer, or a stdio stream to write any
6525 debug output to. @var{verbose} is the verbose level provided by
6526 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6530 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6531 This function corrects the value of @var{cost} based on the
6532 relationship between @var{insn} and @var{dep_insn} through the
6533 dependence @var{link}. It should return the new value. The default
6534 is to make no adjustment to @var{cost}. This can be used for example
6535 to specify to the scheduler using the traditional pipeline description
6536 that an output- or anti-dependence does not incur the same cost as a
6537 data-dependence. If the scheduler using the automaton based pipeline
6538 description, the cost of anti-dependence is zero and the cost of
6539 output-dependence is maximum of one and the difference of latency
6540 times of the first and the second insns. If these values are not
6541 acceptable, you could use the hook to modify them too. See also
6542 @pxref{Processor pipeline description}.
6545 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6546 This hook adjusts the integer scheduling priority @var{priority} of
6547 @var{insn}. It should return the new priority. Increase the priority to
6548 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6549 later. Do not define this hook if you do not need to adjust the
6550 scheduling priorities of insns.
6553 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6554 This hook is executed by the scheduler after it has scheduled the ready
6555 list, to allow the machine description to reorder it (for example to
6556 combine two small instructions together on @samp{VLIW} machines).
6557 @var{file} is either a null pointer, or a stdio stream to write any
6558 debug output to. @var{verbose} is the verbose level provided by
6559 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6560 list of instructions that are ready to be scheduled. @var{n_readyp} is
6561 a pointer to the number of elements in the ready list. The scheduler
6562 reads the ready list in reverse order, starting with
6563 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6564 is the timer tick of the scheduler. You may modify the ready list and
6565 the number of ready insns. The return value is the number of insns that
6566 can issue this cycle; normally this is just @code{issue_rate}. See also
6567 @samp{TARGET_SCHED_REORDER2}.
6570 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6571 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6572 function is called whenever the scheduler starts a new cycle. This one
6573 is called once per iteration over a cycle, immediately after
6574 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6575 return the number of insns to be scheduled in the same cycle. Defining
6576 this hook can be useful if there are frequent situations where
6577 scheduling one insn causes other insns to become ready in the same
6578 cycle. These other insns can then be taken into account properly.
6581 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6582 This hook is called after evaluation forward dependencies of insns in
6583 chain given by two parameter values (@var{head} and @var{tail}
6584 correspondingly) but before insns scheduling of the insn chain. For
6585 example, it can be used for better insn classification if it requires
6586 analysis of dependencies. This hook can use backward and forward
6587 dependencies of the insn scheduler because they are already
6591 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6592 This hook is executed by the scheduler at the beginning of each block of
6593 instructions that are to be scheduled. @var{file} is either a null
6594 pointer, or a stdio stream to write any debug output to. @var{verbose}
6595 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6596 @var{max_ready} is the maximum number of insns in the current scheduling
6597 region that can be live at the same time. This can be used to allocate
6598 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6601 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6602 This hook is executed by the scheduler at the end of each block of
6603 instructions that are to be scheduled. It can be used to perform
6604 cleanup of any actions done by the other scheduling hooks. @var{file}
6605 is either a null pointer, or a stdio stream to write any debug output
6606 to. @var{verbose} is the verbose level provided by
6607 @option{-fsched-verbose-@var{n}}.
6610 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6611 This hook is executed by the scheduler after function level initializations.
6612 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6613 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6614 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6617 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6618 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6619 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6620 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6623 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6624 The hook returns an RTL insn. The automaton state used in the
6625 pipeline hazard recognizer is changed as if the insn were scheduled
6626 when the new simulated processor cycle starts. Usage of the hook may
6627 simplify the automaton pipeline description for some @acronym{VLIW}
6628 processors. If the hook is defined, it is used only for the automaton
6629 based pipeline description. The default is not to change the state
6630 when the new simulated processor cycle starts.
6633 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6634 The hook can be used to initialize data used by the previous hook.
6637 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6638 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6639 to changed the state as if the insn were scheduled when the new
6640 simulated processor cycle finishes.
6643 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6644 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6645 used to initialize data used by the previous hook.
6648 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
6649 The hook to notify target that the current simulated cycle is about to finish.
6650 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6651 to change the state in more complicated situations - e.g., when advancing
6652 state on a single insn is not enough.
6655 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
6656 The hook to notify target that new simulated cycle has just started.
6657 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6658 to change the state in more complicated situations - e.g., when advancing
6659 state on a single insn is not enough.
6662 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6663 This hook controls better choosing an insn from the ready insn queue
6664 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6665 chooses the first insn from the queue. If the hook returns a positive
6666 value, an additional scheduler code tries all permutations of
6667 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6668 subsequent ready insns to choose an insn whose issue will result in
6669 maximal number of issued insns on the same cycle. For the
6670 @acronym{VLIW} processor, the code could actually solve the problem of
6671 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6672 rules of @acronym{VLIW} packing are described in the automaton.
6674 This code also could be used for superscalar @acronym{RISC}
6675 processors. Let us consider a superscalar @acronym{RISC} processor
6676 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6677 @var{B}, some insns can be executed only in pipelines @var{B} or
6678 @var{C}, and one insn can be executed in pipeline @var{B}. The
6679 processor may issue the 1st insn into @var{A} and the 2nd one into
6680 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6681 until the next cycle. If the scheduler issues the 3rd insn the first,
6682 the processor could issue all 3 insns per cycle.
6684 Actually this code demonstrates advantages of the automaton based
6685 pipeline hazard recognizer. We try quickly and easy many insn
6686 schedules to choose the best one.
6688 The default is no multipass scheduling.
6691 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx @var{insn})
6693 This hook controls what insns from the ready insn queue will be
6694 considered for the multipass insn scheduling. If the hook returns
6695 zero for @var{insn}, the insn will be not chosen to
6698 The default is that any ready insns can be chosen to be issued.
6701 @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})
6702 This hook prepares the target backend for a new round of multipass
6706 @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})
6707 This hook is called when multipass scheduling evaluates instruction INSN.
6710 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, char *@var{ready_try}, int @var{n_ready})
6711 This is called when multipass scheduling backtracks from evaluation of
6715 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
6716 This hook notifies the target about the result of the concluded current
6717 round of multipass scheduling.
6720 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
6721 This hook initializes target-specific data used in multipass scheduling.
6724 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
6725 This hook finalizes target-specific data used in multipass scheduling.
6728 @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})
6729 This hook is called by the insn scheduler before issuing @var{insn}
6730 on cycle @var{clock}. If the hook returns nonzero,
6731 @var{insn} is not issued on this processor cycle. Instead,
6732 the processor cycle is advanced. If *@var{sort_p}
6733 is zero, the insn ready queue is not sorted on the new cycle
6734 start as usually. @var{dump} and @var{verbose} specify the file and
6735 verbosity level to use for debugging output.
6736 @var{last_clock} and @var{clock} are, respectively, the
6737 processor cycle on which the previous insn has been issued,
6738 and the current processor cycle.
6741 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
6742 This hook is used to define which dependences are considered costly by
6743 the target, so costly that it is not advisable to schedule the insns that
6744 are involved in the dependence too close to one another. The parameters
6745 to this hook are as follows: The first parameter @var{_dep} is the dependence
6746 being evaluated. The second parameter @var{cost} is the cost of the
6747 dependence as estimated by the scheduler, and the third
6748 parameter @var{distance} is the distance in cycles between the two insns.
6749 The hook returns @code{true} if considering the distance between the two
6750 insns the dependence between them is considered costly by the target,
6751 and @code{false} otherwise.
6753 Defining this hook can be useful in multiple-issue out-of-order machines,
6754 where (a) it's practically hopeless to predict the actual data/resource
6755 delays, however: (b) there's a better chance to predict the actual grouping
6756 that will be formed, and (c) correctly emulating the grouping can be very
6757 important. In such targets one may want to allow issuing dependent insns
6758 closer to one another---i.e., closer than the dependence distance; however,
6759 not in cases of ``costly dependences'', which this hooks allows to define.
6762 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6763 This hook is called by the insn scheduler after emitting a new instruction to
6764 the instruction stream. The hook notifies a target backend to extend its
6765 per instruction data structures.
6768 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6769 Return a pointer to a store large enough to hold target scheduling context.
6772 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6773 Initialize store pointed to by @var{tc} to hold target scheduling context.
6774 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6775 beginning of the block. Otherwise, copy the current context into @var{tc}.
6778 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6779 Copy target scheduling context pointed to by @var{tc} to the current context.
6782 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6783 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6786 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6787 Deallocate a store for target scheduling context pointed to by @var{tc}.
6790 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6791 This hook is called by the insn scheduler when @var{insn} has only
6792 speculative dependencies and therefore can be scheduled speculatively.
6793 The hook is used to check if the pattern of @var{insn} has a speculative
6794 version and, in case of successful check, to generate that speculative
6795 pattern. The hook should return 1, if the instruction has a speculative form,
6796 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6797 speculation. If the return value equals 1 then @var{new_pat} is assigned
6798 the generated speculative pattern.
6801 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (int @var{dep_status})
6802 This hook is called by the insn scheduler during generation of recovery code
6803 for @var{insn}. It should return @code{true}, if the corresponding check
6804 instruction should branch to recovery code, or @code{false} otherwise.
6807 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6808 This hook is called by the insn scheduler to generate a pattern for recovery
6809 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6810 speculative instruction for which the check should be generated.
6811 @var{label} is either a label of a basic block, where recovery code should
6812 be emitted, or a null pointer, when requested check doesn't branch to
6813 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6814 a pattern for a branchy check corresponding to a simple check denoted by
6815 @var{insn} should be generated. In this case @var{label} can't be null.
6818 @deftypefn {Target Hook} bool TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (const_rtx @var{insn})
6819 This hook is used as a workaround for
6820 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6821 called on the first instruction of the ready list. The hook is used to
6822 discard speculative instructions that stand first in the ready list from
6823 being scheduled on the current cycle. If the hook returns @code{false},
6824 @var{insn} will not be chosen to be issued.
6825 For non-speculative instructions,
6826 the hook should always return @code{true}. For example, in the ia64 backend
6827 the hook is used to cancel data speculative insns when the ALAT table
6831 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
6832 This hook is used by the insn scheduler to find out what features should be
6834 The structure *@var{spec_info} should be filled in by the target.
6835 The structure describes speculation types that can be used in the scheduler.
6838 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6839 This hook is called by the swing modulo scheduler to calculate a
6840 resource-based lower bound which is based on the resources available in
6841 the machine and the resources required by each instruction. The target
6842 backend can use @var{g} to calculate such bound. A very simple lower
6843 bound will be used in case this hook is not implemented: the total number
6844 of instructions divided by the issue rate.
6847 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx @var{insn}, int @var{x})
6848 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6849 is supported in hardware and the condition specified in the parameter is true.
6852 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx @var{insn}, int @var{x})
6853 This hook is called by Haifa Scheduler. It performs the operation specified
6854 in its second parameter.
6857 @deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE
6858 True if the processor has an exposed pipeline, which means that not just
6859 the order of instructions is important for correctness when scheduling, but
6860 also the latencies of operations.
6863 @deftypefn {Target Hook} int TARGET_SCHED_REASSOCIATION_WIDTH (unsigned int @var{opc}, enum machine_mode @var{mode})
6864 This hook is called by tree reassociator to determine a level of
6865 parallelism required in output calculations chain.
6869 @section Dividing the Output into Sections (Texts, Data, @dots{})
6870 @c the above section title is WAY too long. maybe cut the part between
6871 @c the (...)? --mew 10feb93
6873 An object file is divided into sections containing different types of
6874 data. In the most common case, there are three sections: the @dfn{text
6875 section}, which holds instructions and read-only data; the @dfn{data
6876 section}, which holds initialized writable data; and the @dfn{bss
6877 section}, which holds uninitialized data. Some systems have other kinds
6880 @file{varasm.c} provides several well-known sections, such as
6881 @code{text_section}, @code{data_section} and @code{bss_section}.
6882 The normal way of controlling a @code{@var{foo}_section} variable
6883 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6884 as described below. The macros are only read once, when @file{varasm.c}
6885 initializes itself, so their values must be run-time constants.
6886 They may however depend on command-line flags.
6888 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6889 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6890 to be string literals.
6892 Some assemblers require a different string to be written every time a
6893 section is selected. If your assembler falls into this category, you
6894 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6895 @code{get_unnamed_section} to set up the sections.
6897 You must always create a @code{text_section}, either by defining
6898 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6899 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6900 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6901 create a distinct @code{readonly_data_section}, the default is to
6902 reuse @code{text_section}.
6904 All the other @file{varasm.c} sections are optional, and are null
6905 if the target does not provide them.
6907 @defmac TEXT_SECTION_ASM_OP
6908 A C expression whose value is a string, including spacing, containing the
6909 assembler operation that should precede instructions and read-only data.
6910 Normally @code{"\t.text"} is right.
6913 @defmac HOT_TEXT_SECTION_NAME
6914 If defined, a C string constant for the name of the section containing most
6915 frequently executed functions of the program. If not defined, GCC will provide
6916 a default definition if the target supports named sections.
6919 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6920 If defined, a C string constant for the name of the section containing unlikely
6921 executed functions in the program.
6924 @defmac DATA_SECTION_ASM_OP
6925 A C expression whose value is a string, including spacing, containing the
6926 assembler operation to identify the following data as writable initialized
6927 data. Normally @code{"\t.data"} is right.
6930 @defmac SDATA_SECTION_ASM_OP
6931 If defined, a C expression whose value is a string, including spacing,
6932 containing the assembler operation to identify the following data as
6933 initialized, writable small data.
6936 @defmac READONLY_DATA_SECTION_ASM_OP
6937 A C expression whose value is a string, including spacing, containing the
6938 assembler operation to identify the following data as read-only initialized
6942 @defmac BSS_SECTION_ASM_OP
6943 If defined, a C expression whose value is a string, including spacing,
6944 containing the assembler operation to identify the following data as
6945 uninitialized global data. If not defined, and
6946 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
6947 uninitialized global data will be output in the data section if
6948 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6952 @defmac SBSS_SECTION_ASM_OP
6953 If defined, a C expression whose value is a string, including spacing,
6954 containing the assembler operation to identify the following data as
6955 uninitialized, writable small data.
6958 @defmac TLS_COMMON_ASM_OP
6959 If defined, a C expression whose value is a string containing the
6960 assembler operation to identify the following data as thread-local
6961 common data. The default is @code{".tls_common"}.
6964 @defmac TLS_SECTION_ASM_FLAG
6965 If defined, a C expression whose value is a character constant
6966 containing the flag used to mark a section as a TLS section. The
6967 default is @code{'T'}.
6970 @defmac INIT_SECTION_ASM_OP
6971 If defined, a C expression whose value is a string, including spacing,
6972 containing the assembler operation to identify the following data as
6973 initialization code. If not defined, GCC will assume such a section does
6974 not exist. This section has no corresponding @code{init_section}
6975 variable; it is used entirely in runtime code.
6978 @defmac FINI_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 finalization code. If not defined, GCC will assume such a section does
6982 not exist. This section has no corresponding @code{fini_section}
6983 variable; it is used entirely in runtime code.
6986 @defmac INIT_ARRAY_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 part of the @code{.init_array} (or equivalent) section. If not
6990 defined, GCC will assume such a section does not exist. Do not define
6991 both this macro and @code{INIT_SECTION_ASM_OP}.
6994 @defmac FINI_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{.fini_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{FINI_SECTION_ASM_OP}.
7002 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
7003 If defined, an ASM statement that switches to a different section
7004 via @var{section_op}, calls @var{function}, and switches back to
7005 the text section. This is used in @file{crtstuff.c} if
7006 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
7007 to initialization and finalization functions from the init and fini
7008 sections. By default, this macro uses a simple function call. Some
7009 ports need hand-crafted assembly code to avoid dependencies on
7010 registers initialized in the function prologue or to ensure that
7011 constant pools don't end up too far way in the text section.
7014 @defmac TARGET_LIBGCC_SDATA_SECTION
7015 If defined, a string which names the section into which small
7016 variables defined in crtstuff and libgcc should go. This is useful
7017 when the target has options for optimizing access to small data, and
7018 you want the crtstuff and libgcc routines to be conservative in what
7019 they expect of your application yet liberal in what your application
7020 expects. For example, for targets with a @code{.sdata} section (like
7021 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
7022 require small data support from your application, but use this macro
7023 to put small data into @code{.sdata} so that your application can
7024 access these variables whether it uses small data or not.
7027 @defmac FORCE_CODE_SECTION_ALIGN
7028 If defined, an ASM statement that aligns a code section to some
7029 arbitrary boundary. This is used to force all fragments of the
7030 @code{.init} and @code{.fini} sections to have to same alignment
7031 and thus prevent the linker from having to add any padding.
7034 @defmac JUMP_TABLES_IN_TEXT_SECTION
7035 Define this macro to be an expression with a nonzero value if jump
7036 tables (for @code{tablejump} insns) should be output in the text
7037 section, along with the assembler instructions. Otherwise, the
7038 readonly data section is used.
7040 This macro is irrelevant if there is no separate readonly data section.
7043 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
7044 Define this hook if you need to do something special to set up the
7045 @file{varasm.c} sections, or if your target has some special sections
7046 of its own that you need to create.
7048 GCC calls this hook after processing the command line, but before writing
7049 any assembly code, and before calling any of the section-returning hooks
7053 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
7054 Return a mask describing how relocations should be treated when
7055 selecting sections. Bit 1 should be set if global relocations
7056 should be placed in a read-write section; bit 0 should be set if
7057 local relocations should be placed in a read-write section.
7059 The default version of this function returns 3 when @option{-fpic}
7060 is in effect, and 0 otherwise. The hook is typically redefined
7061 when the target cannot support (some kinds of) dynamic relocations
7062 in read-only sections even in executables.
7065 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
7066 Return the section into which @var{exp} should be placed. You can
7067 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
7068 some sort. @var{reloc} indicates whether the initial value of @var{exp}
7069 requires link-time relocations. Bit 0 is set when variable contains
7070 local relocations only, while bit 1 is set for global relocations.
7071 @var{align} is the constant alignment in bits.
7073 The default version of this function takes care of putting read-only
7074 variables in @code{readonly_data_section}.
7076 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7079 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7080 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7081 for @code{FUNCTION_DECL}s as well as for variables and constants.
7083 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7084 function has been determined to be likely to be called, and nonzero if
7085 it is unlikely to be called.
7088 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7089 Build up a unique section name, expressed as a @code{STRING_CST} node,
7090 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7091 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7092 the initial value of @var{exp} requires link-time relocations.
7094 The default version of this function appends the symbol name to the
7095 ELF section name that would normally be used for the symbol. For
7096 example, the function @code{foo} would be placed in @code{.text.foo}.
7097 Whatever the actual target object format, this is often good enough.
7100 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7101 Return the readonly data section associated with
7102 @samp{DECL_SECTION_NAME (@var{decl})}.
7103 The default version of this function selects @code{.gnu.linkonce.r.name} if
7104 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7105 if function is in @code{.text.name}, and the normal readonly-data section
7109 @deftypevr {Target Hook} {const char *} TARGET_ASM_MERGEABLE_RODATA_PREFIX
7110 Usually, the compiler uses the prefix @code{".rodata"} to construct
7111 section names for mergeable constant data. Define this macro to override
7112 the string if a different section name should be used.
7115 @deftypefn {Target Hook} {section *} TARGET_ASM_TM_CLONE_TABLE_SECTION (void)
7116 Return the section that should be used for transactional memory clone tables.
7119 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7120 Return the section into which a constant @var{x}, of mode @var{mode},
7121 should be placed. You can assume that @var{x} is some kind of
7122 constant in RTL@. The argument @var{mode} is redundant except in the
7123 case of a @code{const_int} rtx. @var{align} is the constant alignment
7126 The default version of this function takes care of putting symbolic
7127 constants in @code{flag_pic} mode in @code{data_section} and everything
7128 else in @code{readonly_data_section}.
7131 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7132 Define this hook if you need to postprocess the assembler name generated
7133 by target-independent code. The @var{id} provided to this hook will be
7134 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7135 or the mangled name of the @var{decl} in C++). The return value of the
7136 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7137 your target system. The default implementation of this hook just
7138 returns the @var{id} provided.
7141 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7142 Define this hook if references to a symbol or a constant must be
7143 treated differently depending on something about the variable or
7144 function named by the symbol (such as what section it is in).
7146 The hook is executed immediately after rtl has been created for
7147 @var{decl}, which may be a variable or function declaration or
7148 an entry in the constant pool. In either case, @var{rtl} is the
7149 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7150 in this hook; that field may not have been initialized yet.
7152 In the case of a constant, it is safe to assume that the rtl is
7153 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7154 will also have this form, but that is not guaranteed. Global
7155 register variables, for instance, will have a @code{reg} for their
7156 rtl. (Normally the right thing to do with such unusual rtl is
7159 The @var{new_decl_p} argument will be true if this is the first time
7160 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7161 be false for subsequent invocations, which will happen for duplicate
7162 declarations. Whether or not anything must be done for the duplicate
7163 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7164 @var{new_decl_p} is always true when the hook is called for a constant.
7166 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7167 The usual thing for this hook to do is to record flags in the
7168 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7169 Historically, the name string was modified if it was necessary to
7170 encode more than one bit of information, but this practice is now
7171 discouraged; use @code{SYMBOL_REF_FLAGS}.
7173 The default definition of this hook, @code{default_encode_section_info}
7174 in @file{varasm.c}, sets a number of commonly-useful bits in
7175 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7176 before overriding it.
7179 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7180 Decode @var{name} and return the real name part, sans
7181 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7185 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7186 Returns true if @var{exp} should be placed into a ``small data'' section.
7187 The default version of this hook always returns false.
7190 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7191 Contains the value true if the target places read-only
7192 ``small data'' into a separate section. The default value is false.
7195 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7196 It returns true if target wants profile code emitted before prologue.
7198 The default version of this hook use the target macro
7199 @code{PROFILE_BEFORE_PROLOGUE}.
7202 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7203 Returns true if @var{exp} names an object for which name resolution
7204 rules must resolve to the current ``module'' (dynamic shared library
7205 or executable image).
7207 The default version of this hook implements the name resolution rules
7208 for ELF, which has a looser model of global name binding than other
7209 currently supported object file formats.
7212 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7213 Contains the value true if the target supports thread-local storage.
7214 The default value is false.
7219 @section Position Independent Code
7220 @cindex position independent code
7223 This section describes macros that help implement generation of position
7224 independent code. Simply defining these macros is not enough to
7225 generate valid PIC; you must also add support to the hook
7226 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7227 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7228 must modify the definition of @samp{movsi} to do something appropriate
7229 when the source operand contains a symbolic address. You may also
7230 need to alter the handling of switch statements so that they use
7232 @c i rearranged the order of the macros above to try to force one of
7233 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7235 @defmac PIC_OFFSET_TABLE_REGNUM
7236 The register number of the register used to address a table of static
7237 data addresses in memory. In some cases this register is defined by a
7238 processor's ``application binary interface'' (ABI)@. When this macro
7239 is defined, RTL is generated for this register once, as with the stack
7240 pointer and frame pointer registers. If this macro is not defined, it
7241 is up to the machine-dependent files to allocate such a register (if
7242 necessary). Note that this register must be fixed when in use (e.g.@:
7243 when @code{flag_pic} is true).
7246 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7247 A C expression that is nonzero if the register defined by
7248 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7249 the default is zero. Do not define
7250 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7253 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7254 A C expression that is nonzero if @var{x} is a legitimate immediate
7255 operand on the target machine when generating position independent code.
7256 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7257 check this. You can also assume @var{flag_pic} is true, so you need not
7258 check it either. You need not define this macro if all constants
7259 (including @code{SYMBOL_REF}) can be immediate operands when generating
7260 position independent code.
7263 @node Assembler Format
7264 @section Defining the Output Assembler Language
7266 This section describes macros whose principal purpose is to describe how
7267 to write instructions in assembler language---rather than what the
7271 * File Framework:: Structural information for the assembler file.
7272 * Data Output:: Output of constants (numbers, strings, addresses).
7273 * Uninitialized Data:: Output of uninitialized variables.
7274 * Label Output:: Output and generation of labels.
7275 * Initialization:: General principles of initialization
7276 and termination routines.
7277 * Macros for Initialization::
7278 Specific macros that control the handling of
7279 initialization and termination routines.
7280 * Instruction Output:: Output of actual instructions.
7281 * Dispatch Tables:: Output of jump tables.
7282 * Exception Region Output:: Output of exception region code.
7283 * Alignment Output:: Pseudo ops for alignment and skipping data.
7286 @node File Framework
7287 @subsection The Overall Framework of an Assembler File
7288 @cindex assembler format
7289 @cindex output of assembler code
7291 @c prevent bad page break with this line
7292 This describes the overall framework of an assembly file.
7294 @findex default_file_start
7295 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7296 Output to @code{asm_out_file} any text which the assembler expects to
7297 find at the beginning of a file. The default behavior is controlled
7298 by two flags, documented below. Unless your target's assembler is
7299 quite unusual, if you override the default, you should call
7300 @code{default_file_start} at some point in your target hook. This
7301 lets other target files rely on these variables.
7304 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7305 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7306 printed as the very first line in the assembly file, unless
7307 @option{-fverbose-asm} is in effect. (If that macro has been defined
7308 to the empty string, this variable has no effect.) With the normal
7309 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7310 assembler that it need not bother stripping comments or extra
7311 whitespace from its input. This allows it to work a bit faster.
7313 The default is false. You should not set it to true unless you have
7314 verified that your port does not generate any extra whitespace or
7315 comments that will cause GAS to issue errors in NO_APP mode.
7318 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7319 If this flag is true, @code{output_file_directive} will be called
7320 for the primary source file, immediately after printing
7321 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7322 this to be done. The default is false.
7325 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7326 Output to @code{asm_out_file} any text which the assembler expects
7327 to find at the end of a file. The default is to output nothing.
7330 @deftypefun void file_end_indicate_exec_stack ()
7331 Some systems use a common convention, the @samp{.note.GNU-stack}
7332 special section, to indicate whether or not an object file relies on
7333 the stack being executable. If your system uses this convention, you
7334 should define @code{TARGET_ASM_FILE_END} to this function. If you
7335 need to do other things in that hook, have your hook function call
7339 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7340 Output to @code{asm_out_file} any text which the assembler expects
7341 to find at the start of an LTO section. The default is to output
7345 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7346 Output to @code{asm_out_file} any text which the assembler expects
7347 to find at the end of an LTO section. The default is to output
7351 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7352 Output to @code{asm_out_file} any text which is needed before emitting
7353 unwind info and debug info at the end of a file. Some targets emit
7354 here PIC setup thunks that cannot be emitted at the end of file,
7355 because they couldn't have unwind info then. The default is to output
7359 @defmac ASM_COMMENT_START
7360 A C string constant describing how to begin a comment in the target
7361 assembler language. The compiler assumes that the comment will end at
7362 the end of the line.
7366 A C string constant for text to be output before each @code{asm}
7367 statement or group of consecutive ones. Normally this is
7368 @code{"#APP"}, which is a comment that has no effect on most
7369 assemblers but tells the GNU assembler that it must check the lines
7370 that follow for all valid assembler constructs.
7374 A C string constant for text to be output after each @code{asm}
7375 statement or group of consecutive ones. Normally this is
7376 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7377 time-saving assumptions that are valid for ordinary compiler output.
7380 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7381 A C statement to output COFF information or DWARF debugging information
7382 which indicates that filename @var{name} is the current source file to
7383 the stdio stream @var{stream}.
7385 This macro need not be defined if the standard form of output
7386 for the file format in use is appropriate.
7389 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7390 Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7392 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7395 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7396 A C statement to output the string @var{string} to the stdio stream
7397 @var{stream}. If you do not call the function @code{output_quoted_string}
7398 in your config files, GCC will only call it to output filenames to
7399 the assembler source. So you can use it to canonicalize the format
7400 of the filename using this macro.
7403 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7404 A C statement to output something to the assembler file to handle a
7405 @samp{#ident} directive containing the text @var{string}. If this
7406 macro is not defined, nothing is output for a @samp{#ident} directive.
7409 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7410 Output assembly directives to switch to section @var{name}. The section
7411 should have attributes as specified by @var{flags}, which is a bit mask
7412 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7413 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7414 this section is associated.
7417 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
7418 Return preferred text (sub)section for function @var{decl}.
7419 Main purpose of this function is to separate cold, normal and hot
7420 functions. @var{startup} is true when function is known to be used only
7421 at startup (from static constructors or it is @code{main()}).
7422 @var{exit} is true when function is known to be used only at exit
7423 (from static destructors).
7424 Return NULL if function should go to default text section.
7427 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
7428 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}.
7431 @deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7432 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7433 It must not be modified by command-line option processing.
7436 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7437 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7438 This flag is true if we can create zeroed data by switching to a BSS
7439 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7440 This is true on most ELF targets.
7443 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7444 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7445 based on a variable or function decl, a section name, and whether or not the
7446 declaration's initializer may contain runtime relocations. @var{decl} may be
7447 null, in which case read-write data should be assumed.
7449 The default version of this function handles choosing code vs data,
7450 read-only vs read-write data, and @code{flag_pic}. You should only
7451 need to override this if your target has special flags that might be
7452 set via @code{__attribute__}.
7455 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
7456 Provides the target with the ability to record the gcc command line
7457 switches that have been passed to the compiler, and options that are
7458 enabled. The @var{type} argument specifies what is being recorded.
7459 It can take the following values:
7462 @item SWITCH_TYPE_PASSED
7463 @var{text} is a command line switch that has been set by the user.
7465 @item SWITCH_TYPE_ENABLED
7466 @var{text} is an option which has been enabled. This might be as a
7467 direct result of a command line switch, or because it is enabled by
7468 default or because it has been enabled as a side effect of a different
7469 command line switch. For example, the @option{-O2} switch enables
7470 various different individual optimization passes.
7472 @item SWITCH_TYPE_DESCRIPTIVE
7473 @var{text} is either NULL or some descriptive text which should be
7474 ignored. If @var{text} is NULL then it is being used to warn the
7475 target hook that either recording is starting or ending. The first
7476 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7477 warning is for start up and the second time the warning is for
7478 wind down. This feature is to allow the target hook to make any
7479 necessary preparations before it starts to record switches and to
7480 perform any necessary tidying up after it has finished recording
7483 @item SWITCH_TYPE_LINE_START
7484 This option can be ignored by this target hook.
7486 @item SWITCH_TYPE_LINE_END
7487 This option can be ignored by this target hook.
7490 The hook's return value must be zero. Other return values may be
7491 supported in the future.
7493 By default this hook is set to NULL, but an example implementation is
7494 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7495 it records the switches as ASCII text inside a new, string mergeable
7496 section in the assembler output file. The name of the new section is
7497 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7501 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7502 This is the name of the section that will be created by the example
7503 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7509 @subsection Output of Data
7512 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7513 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7514 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7515 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7516 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7517 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7518 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7519 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7520 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7521 These hooks specify assembly directives for creating certain kinds
7522 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7523 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7524 aligned two-byte object, and so on. Any of the hooks may be
7525 @code{NULL}, indicating that no suitable directive is available.
7527 The compiler will print these strings at the start of a new line,
7528 followed immediately by the object's initial value. In most cases,
7529 the string should contain a tab, a pseudo-op, and then another tab.
7532 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7533 The @code{assemble_integer} function uses this hook to output an
7534 integer object. @var{x} is the object's value, @var{size} is its size
7535 in bytes and @var{aligned_p} indicates whether it is aligned. The
7536 function should return @code{true} if it was able to output the
7537 object. If it returns false, @code{assemble_integer} will try to
7538 split the object into smaller parts.
7540 The default implementation of this hook will use the
7541 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7542 when the relevant string is @code{NULL}.
7545 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
7546 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7547 can't deal with, and output assembly code to @var{file} corresponding to
7548 the pattern @var{x}. This may be used to allow machine-dependent
7549 @code{UNSPEC}s to appear within constants.
7551 If target hook fails to recognize a pattern, it must return @code{false},
7552 so that a standard error message is printed. If it prints an error message
7553 itself, by calling, for example, @code{output_operand_lossage}, it may just
7557 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7558 A C statement to output to the stdio stream @var{stream} an assembler
7559 instruction to assemble a string constant containing the @var{len}
7560 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7561 @code{char *} and @var{len} a C expression of type @code{int}.
7563 If the assembler has a @code{.ascii} pseudo-op as found in the
7564 Berkeley Unix assembler, do not define the macro
7565 @code{ASM_OUTPUT_ASCII}.
7568 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7569 A C statement to output word @var{n} of a function descriptor for
7570 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7571 is defined, and is otherwise unused.
7574 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7575 You may define this macro as a C expression. You should define the
7576 expression to have a nonzero value if GCC should output the constant
7577 pool for a function before the code for the function, or a zero value if
7578 GCC should output the constant pool after the function. If you do
7579 not define this macro, the usual case, GCC will output the constant
7580 pool before the function.
7583 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7584 A C statement to output assembler commands to define the start of the
7585 constant pool for a function. @var{funname} is a string giving
7586 the name of the function. Should the return type of the function
7587 be required, it can be obtained via @var{fundecl}. @var{size}
7588 is the size, in bytes, of the constant pool that will be written
7589 immediately after this call.
7591 If no constant-pool prefix is required, the usual case, this macro need
7595 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7596 A C statement (with or without semicolon) to output a constant in the
7597 constant pool, if it needs special treatment. (This macro need not do
7598 anything for RTL expressions that can be output normally.)
7600 The argument @var{file} is the standard I/O stream to output the
7601 assembler code on. @var{x} is the RTL expression for the constant to
7602 output, and @var{mode} is the machine mode (in case @var{x} is a
7603 @samp{const_int}). @var{align} is the required alignment for the value
7604 @var{x}; you should output an assembler directive to force this much
7607 The argument @var{labelno} is a number to use in an internal label for
7608 the address of this pool entry. The definition of this macro is
7609 responsible for outputting the label definition at the proper place.
7610 Here is how to do this:
7613 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7616 When you output a pool entry specially, you should end with a
7617 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7618 entry from being output a second time in the usual manner.
7620 You need not define this macro if it would do nothing.
7623 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7624 A C statement to output assembler commands to at the end of the constant
7625 pool for a function. @var{funname} is a string giving the name of the
7626 function. Should the return type of the function be required, you can
7627 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7628 constant pool that GCC wrote immediately before this call.
7630 If no constant-pool epilogue is required, the usual case, you need not
7634 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7635 Define this macro as a C expression which is nonzero if @var{C} is
7636 used as a logical line separator by the assembler. @var{STR} points
7637 to the position in the string where @var{C} was found; this can be used if
7638 a line separator uses multiple characters.
7640 If you do not define this macro, the default is that only
7641 the character @samp{;} is treated as a logical line separator.
7644 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7645 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7646 These target hooks are C string constants, describing the syntax in the
7647 assembler for grouping arithmetic expressions. If not overridden, they
7648 default to normal parentheses, which is correct for most assemblers.
7651 These macros are provided by @file{real.h} for writing the definitions
7652 of @code{ASM_OUTPUT_DOUBLE} and the like:
7654 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7655 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7656 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7657 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7658 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7659 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7660 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7661 target's floating point representation, and store its bit pattern in
7662 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7663 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7664 simple @code{long int}. For the others, it should be an array of
7665 @code{long int}. The number of elements in this array is determined
7666 by the size of the desired target floating point data type: 32 bits of
7667 it go in each @code{long int} array element. Each array element holds
7668 32 bits of the result, even if @code{long int} is wider than 32 bits
7669 on the host machine.
7671 The array element values are designed so that you can print them out
7672 using @code{fprintf} in the order they should appear in the target
7676 @node Uninitialized Data
7677 @subsection Output of Uninitialized Variables
7679 Each of the macros in this section is used to do the whole job of
7680 outputting a single uninitialized variable.
7682 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7683 A C statement (sans semicolon) to output to the stdio stream
7684 @var{stream} the assembler definition of a common-label named
7685 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7686 is the size rounded up to whatever alignment the caller wants. It is
7687 possible that @var{size} may be zero, for instance if a struct with no
7688 other member than a zero-length array is defined. In this case, the
7689 backend must output a symbol definition that allocates at least one
7690 byte, both so that the address of the resulting object does not compare
7691 equal to any other, and because some object formats cannot even express
7692 the concept of a zero-sized common symbol, as that is how they represent
7693 an ordinary undefined external.
7695 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7696 output the name itself; before and after that, output the additional
7697 assembler syntax for defining the name, and a newline.
7699 This macro controls how the assembler definitions of uninitialized
7700 common global variables are output.
7703 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7704 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7705 separate, explicit argument. If you define this macro, it is used in
7706 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7707 handling the required alignment of the variable. The alignment is specified
7708 as the number of bits.
7711 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7712 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7713 variable to be output, if there is one, or @code{NULL_TREE} if there
7714 is no corresponding variable. If you define this macro, GCC will use it
7715 in place of both @code{ASM_OUTPUT_COMMON} and
7716 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7717 the variable's decl in order to chose what to output.
7720 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7721 A C statement (sans semicolon) to output to the stdio stream
7722 @var{stream} the assembler definition of uninitialized global @var{decl} named
7723 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
7724 is the alignment specified as the number of bits.
7726 Try to use function @code{asm_output_aligned_bss} defined in file
7727 @file{varasm.c} when defining this macro. If unable, use the expression
7728 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7729 before and after that, output the additional assembler syntax for defining
7730 the name, and a newline.
7732 There are two ways of handling global BSS@. One is to define this macro.
7733 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7734 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7735 You do not need to do both.
7737 Some languages do not have @code{common} data, and require a
7738 non-common form of global BSS in order to handle uninitialized globals
7739 efficiently. C++ is one example of this. However, if the target does
7740 not support global BSS, the front end may choose to make globals
7741 common in order to save space in the object file.
7744 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7745 A C statement (sans semicolon) to output to the stdio stream
7746 @var{stream} the assembler definition of a local-common-label named
7747 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7748 is the size rounded up to whatever alignment the caller wants.
7750 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7751 output the name itself; before and after that, output the additional
7752 assembler syntax for defining the name, and a newline.
7754 This macro controls how the assembler definitions of uninitialized
7755 static variables are output.
7758 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7759 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7760 separate, explicit argument. If you define this macro, it is used in
7761 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7762 handling the required alignment of the variable. The alignment is specified
7763 as the number of bits.
7766 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7767 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7768 variable to be output, if there is one, or @code{NULL_TREE} if there
7769 is no corresponding variable. If you define this macro, GCC will use it
7770 in place of both @code{ASM_OUTPUT_DECL} and
7771 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7772 the variable's decl in order to chose what to output.
7776 @subsection Output and Generation of Labels
7778 @c prevent bad page break with this line
7779 This is about outputting labels.
7781 @findex assemble_name
7782 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7783 A C statement (sans semicolon) to output to the stdio stream
7784 @var{stream} the assembler definition of a label named @var{name}.
7785 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7786 output the name itself; before and after that, output the additional
7787 assembler syntax for defining the name, and a newline. A default
7788 definition of this macro is provided which is correct for most systems.
7791 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7792 A C statement (sans semicolon) to output to the stdio stream
7793 @var{stream} the assembler definition of a label named @var{name} of
7795 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7796 output the name itself; before and after that, output the additional
7797 assembler syntax for defining the name, and a newline. A default
7798 definition of this macro is provided which is correct for most systems.
7800 If this macro is not defined, then the function name is defined in the
7801 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7804 @findex assemble_name_raw
7805 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7806 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7807 to refer to a compiler-generated label. The default definition uses
7808 @code{assemble_name_raw}, which is like @code{assemble_name} except
7809 that it is more efficient.
7813 A C string containing the appropriate assembler directive to specify the
7814 size of a symbol, without any arguments. On systems that use ELF, the
7815 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7816 systems, the default is not to define this macro.
7818 Define this macro only if it is correct to use the default definitions
7819 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7820 for your system. If you need your own custom definitions of those
7821 macros, or if you do not need explicit symbol sizes at all, do not
7825 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7826 A C statement (sans semicolon) to output to the stdio stream
7827 @var{stream} a directive telling the assembler that the size of the
7828 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7829 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7833 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7834 A C statement (sans semicolon) to output to the stdio stream
7835 @var{stream} a directive telling the assembler to calculate the size of
7836 the symbol @var{name} by subtracting its address from the current
7839 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7840 provided. The default assumes that the assembler recognizes a special
7841 @samp{.} symbol as referring to the current address, and can calculate
7842 the difference between this and another symbol. If your assembler does
7843 not recognize @samp{.} or cannot do calculations with it, you will need
7844 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7848 A C string containing the appropriate assembler directive to specify the
7849 type of a symbol, without any arguments. On systems that use ELF, the
7850 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7851 systems, the default is not to define this macro.
7853 Define this macro only if it is correct to use the default definition of
7854 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7855 custom definition of this macro, or if you do not need explicit symbol
7856 types at all, do not define this macro.
7859 @defmac TYPE_OPERAND_FMT
7860 A C string which specifies (using @code{printf} syntax) the format of
7861 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7862 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7863 the default is not to define this macro.
7865 Define this macro only if it is correct to use the default definition of
7866 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7867 custom definition of this macro, or if you do not need explicit symbol
7868 types at all, do not define this macro.
7871 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7872 A C statement (sans semicolon) to output to the stdio stream
7873 @var{stream} a directive telling the assembler that the type of the
7874 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7875 that string is always either @samp{"function"} or @samp{"object"}, but
7876 you should not count on this.
7878 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7879 definition of this macro is provided.
7882 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7883 A C statement (sans semicolon) to output to the stdio stream
7884 @var{stream} any text necessary for declaring the name @var{name} of a
7885 function which is being defined. This macro is responsible for
7886 outputting the label definition (perhaps using
7887 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7888 @code{FUNCTION_DECL} tree node representing the function.
7890 If this macro is not defined, then the function name is defined in the
7891 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7893 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7897 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7898 A C statement (sans semicolon) to output to the stdio stream
7899 @var{stream} any text necessary for declaring the size of a function
7900 which is being defined. The argument @var{name} is the name of the
7901 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7902 representing the function.
7904 If this macro is not defined, then the function size is not defined.
7906 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7910 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7911 A C statement (sans semicolon) to output to the stdio stream
7912 @var{stream} any text necessary for declaring the name @var{name} of an
7913 initialized variable which is being defined. This macro must output the
7914 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7915 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7917 If this macro is not defined, then the variable name is defined in the
7918 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7920 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7921 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7924 @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})
7925 A target hook to output to the stdio stream @var{file} any text necessary
7926 for declaring the name @var{name} of a constant which is being defined. This
7927 target hook is responsible for outputting the label definition (perhaps using
7928 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7929 and @var{size} is the size of the constant in bytes. The @var{name}
7930 will be an internal label.
7932 The default version of this target hook, define the @var{name} in the
7933 usual manner as a label (by means of @code{assemble_label}).
7935 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7938 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7939 A C statement (sans semicolon) to output to the stdio stream
7940 @var{stream} any text necessary for claiming a register @var{regno}
7941 for a global variable @var{decl} with name @var{name}.
7943 If you don't define this macro, that is equivalent to defining it to do
7947 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7948 A C statement (sans semicolon) to finish up declaring a variable name
7949 once the compiler has processed its initializer fully and thus has had a
7950 chance to determine the size of an array when controlled by an
7951 initializer. This is used on systems where it's necessary to declare
7952 something about the size of the object.
7954 If you don't define this macro, that is equivalent to defining it to do
7957 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7958 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7961 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7962 This target hook is a function to output to the stdio stream
7963 @var{stream} some commands that will make the label @var{name} global;
7964 that is, available for reference from other files.
7966 The default implementation relies on a proper definition of
7967 @code{GLOBAL_ASM_OP}.
7970 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7971 This target hook is a function to output to the stdio stream
7972 @var{stream} some commands that will make the name associated with @var{decl}
7973 global; that is, available for reference from other files.
7975 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7978 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7979 A C statement (sans semicolon) to output to the stdio stream
7980 @var{stream} some commands that will make the label @var{name} weak;
7981 that is, available for reference from other files but only used if
7982 no other definition is available. Use the expression
7983 @code{assemble_name (@var{stream}, @var{name})} to output the name
7984 itself; before and after that, output the additional assembler syntax
7985 for making that name weak, and a newline.
7987 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7988 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7992 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7993 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7994 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7995 or variable decl. If @var{value} is not @code{NULL}, this C statement
7996 should output to the stdio stream @var{stream} assembler code which
7997 defines (equates) the weak symbol @var{name} to have the value
7998 @var{value}. If @var{value} is @code{NULL}, it should output commands
7999 to make @var{name} weak.
8002 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
8003 Outputs a directive that enables @var{name} to be used to refer to
8004 symbol @var{value} with weak-symbol semantics. @code{decl} is the
8005 declaration of @code{name}.
8008 @defmac SUPPORTS_WEAK
8009 A preprocessor constant expression which evaluates to true if the target
8010 supports weak symbols.
8012 If you don't define this macro, @file{defaults.h} provides a default
8013 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
8014 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
8017 @defmac TARGET_SUPPORTS_WEAK
8018 A C expression which evaluates to true if the target supports weak symbols.
8020 If you don't define this macro, @file{defaults.h} provides a default
8021 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
8022 this macro if you want to control weak symbol support with a compiler
8023 flag such as @option{-melf}.
8026 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
8027 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
8028 public symbol such that extra copies in multiple translation units will
8029 be discarded by the linker. Define this macro if your object file
8030 format provides support for this concept, such as the @samp{COMDAT}
8031 section flags in the Microsoft Windows PE/COFF format, and this support
8032 requires changes to @var{decl}, such as putting it in a separate section.
8035 @defmac SUPPORTS_ONE_ONLY
8036 A C expression which evaluates to true if the target supports one-only
8039 If you don't define this macro, @file{varasm.c} provides a default
8040 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
8041 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
8042 you want to control one-only symbol support with a compiler flag, or if
8043 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
8044 be emitted as one-only.
8047 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
8048 This target hook is a function to output to @var{asm_out_file} some
8049 commands that will make the symbol(s) associated with @var{decl} have
8050 hidden, protected or internal visibility as specified by @var{visibility}.
8053 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
8054 A C expression that evaluates to true if the target's linker expects
8055 that weak symbols do not appear in a static archive's table of contents.
8056 The default is @code{0}.
8058 Leaving weak symbols out of an archive's table of contents means that,
8059 if a symbol will only have a definition in one translation unit and
8060 will have undefined references from other translation units, that
8061 symbol should not be weak. Defining this macro to be nonzero will
8062 thus have the effect that certain symbols that would normally be weak
8063 (explicit template instantiations, and vtables for polymorphic classes
8064 with noninline key methods) will instead be nonweak.
8066 The C++ ABI requires this macro to be zero. Define this macro for
8067 targets where full C++ ABI compliance is impossible and where linker
8068 restrictions require weak symbols to be left out of a static archive's
8072 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8073 A C statement (sans semicolon) to output to the stdio stream
8074 @var{stream} any text necessary for declaring the name of an external
8075 symbol named @var{name} which is referenced in this compilation but
8076 not defined. The value of @var{decl} is the tree node for the
8079 This macro need not be defined if it does not need to output anything.
8080 The GNU assembler and most Unix assemblers don't require anything.
8083 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8084 This target hook is a function to output to @var{asm_out_file} an assembler
8085 pseudo-op to declare a library function name external. The name of the
8086 library function is given by @var{symref}, which is a @code{symbol_ref}.
8089 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8090 This target hook is a function to output to @var{asm_out_file} an assembler
8091 directive to annotate @var{symbol} as used. The Darwin target uses the
8092 .no_dead_code_strip directive.
8095 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8096 A C statement (sans semicolon) to output to the stdio stream
8097 @var{stream} a reference in assembler syntax to a label named
8098 @var{name}. This should add @samp{_} to the front of the name, if that
8099 is customary on your operating system, as it is in most Berkeley Unix
8100 systems. This macro is used in @code{assemble_name}.
8103 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8104 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.
8107 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8108 A C statement (sans semicolon) to output a reference to
8109 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8110 will be used to output the name of the symbol. This macro may be used
8111 to modify the way a symbol is referenced depending on information
8112 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8115 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8116 A C statement (sans semicolon) to output a reference to @var{buf}, the
8117 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8118 @code{assemble_name} will be used to output the name of the symbol.
8119 This macro is not used by @code{output_asm_label}, or the @code{%l}
8120 specifier that calls it; the intention is that this macro should be set
8121 when it is necessary to output a label differently when its address is
8125 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8126 A function to output to the stdio stream @var{stream} a label whose
8127 name is made from the string @var{prefix} and the number @var{labelno}.
8129 It is absolutely essential that these labels be distinct from the labels
8130 used for user-level functions and variables. Otherwise, certain programs
8131 will have name conflicts with internal labels.
8133 It is desirable to exclude internal labels from the symbol table of the
8134 object file. Most assemblers have a naming convention for labels that
8135 should be excluded; on many systems, the letter @samp{L} at the
8136 beginning of a label has this effect. You should find out what
8137 convention your system uses, and follow it.
8139 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8142 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8143 A C statement to output to the stdio stream @var{stream} a debug info
8144 label whose name is made from the string @var{prefix} and the number
8145 @var{num}. This is useful for VLIW targets, where debug info labels
8146 may need to be treated differently than branch target labels. On some
8147 systems, branch target labels must be at the beginning of instruction
8148 bundles, but debug info labels can occur in the middle of instruction
8151 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8155 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8156 A C statement to store into the string @var{string} a label whose name
8157 is made from the string @var{prefix} and the number @var{num}.
8159 This string, when output subsequently by @code{assemble_name}, should
8160 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8161 with the same @var{prefix} and @var{num}.
8163 If the string begins with @samp{*}, then @code{assemble_name} will
8164 output the rest of the string unchanged. It is often convenient for
8165 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8166 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8167 to output the string, and may change it. (Of course,
8168 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8169 you should know what it does on your machine.)
8172 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8173 A C expression to assign to @var{outvar} (which is a variable of type
8174 @code{char *}) a newly allocated string made from the string
8175 @var{name} and the number @var{number}, with some suitable punctuation
8176 added. Use @code{alloca} to get space for the string.
8178 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8179 produce an assembler label for an internal static variable whose name is
8180 @var{name}. Therefore, the string must be such as to result in valid
8181 assembler code. The argument @var{number} is different each time this
8182 macro is executed; it prevents conflicts between similarly-named
8183 internal static variables in different scopes.
8185 Ideally this string should not be a valid C identifier, to prevent any
8186 conflict with the user's own symbols. Most assemblers allow periods
8187 or percent signs in assembler symbols; putting at least one of these
8188 between the name and the number will suffice.
8190 If this macro is not defined, a default definition will be provided
8191 which is correct for most systems.
8194 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8195 A C statement to output to the stdio stream @var{stream} assembler code
8196 which defines (equates) the symbol @var{name} to have the value @var{value}.
8199 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8200 correct for most systems.
8203 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8204 A C statement to output to the stdio stream @var{stream} assembler code
8205 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8206 to have the value of the tree node @var{decl_of_value}. This macro will
8207 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8208 the tree nodes are available.
8211 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8212 correct for most systems.
8215 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8216 A C statement that evaluates to true if the assembler code which defines
8217 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8218 of the tree node @var{decl_of_value} should be emitted near the end of the
8219 current compilation unit. The default is to not defer output of defines.
8220 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8221 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8224 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8225 A C statement to output to the stdio stream @var{stream} assembler code
8226 which defines (equates) the weak symbol @var{name} to have the value
8227 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8228 an undefined weak symbol.
8230 Define this macro if the target only supports weak aliases; define
8231 @code{ASM_OUTPUT_DEF} instead if possible.
8234 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8235 Define this macro to override the default assembler names used for
8236 Objective-C methods.
8238 The default name is a unique method number followed by the name of the
8239 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8240 the category is also included in the assembler name (e.g.@:
8243 These names are safe on most systems, but make debugging difficult since
8244 the method's selector is not present in the name. Therefore, particular
8245 systems define other ways of computing names.
8247 @var{buf} is an expression of type @code{char *} which gives you a
8248 buffer in which to store the name; its length is as long as
8249 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8250 50 characters extra.
8252 The argument @var{is_inst} specifies whether the method is an instance
8253 method or a class method; @var{class_name} is the name of the class;
8254 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8255 in a category); and @var{sel_name} is the name of the selector.
8257 On systems where the assembler can handle quoted names, you can use this
8258 macro to provide more human-readable names.
8261 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
8262 A C statement (sans semicolon) to output to the stdio stream
8263 @var{stream} commands to declare that the label @var{name} is an
8264 Objective-C class reference. This is only needed for targets whose
8265 linkers have special support for NeXT-style runtimes.
8268 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
8269 A C statement (sans semicolon) to output to the stdio stream
8270 @var{stream} commands to declare that the label @var{name} is an
8271 unresolved Objective-C class reference. This is only needed for targets
8272 whose linkers have special support for NeXT-style runtimes.
8275 @node Initialization
8276 @subsection How Initialization Functions Are Handled
8277 @cindex initialization routines
8278 @cindex termination routines
8279 @cindex constructors, output of
8280 @cindex destructors, output of
8282 The compiled code for certain languages includes @dfn{constructors}
8283 (also called @dfn{initialization routines})---functions to initialize
8284 data in the program when the program is started. These functions need
8285 to be called before the program is ``started''---that is to say, before
8286 @code{main} is called.
8288 Compiling some languages generates @dfn{destructors} (also called
8289 @dfn{termination routines}) that should be called when the program
8292 To make the initialization and termination functions work, the compiler
8293 must output something in the assembler code to cause those functions to
8294 be called at the appropriate time. When you port the compiler to a new
8295 system, you need to specify how to do this.
8297 There are two major ways that GCC currently supports the execution of
8298 initialization and termination functions. Each way has two variants.
8299 Much of the structure is common to all four variations.
8301 @findex __CTOR_LIST__
8302 @findex __DTOR_LIST__
8303 The linker must build two lists of these functions---a list of
8304 initialization functions, called @code{__CTOR_LIST__}, and a list of
8305 termination functions, called @code{__DTOR_LIST__}.
8307 Each list always begins with an ignored function pointer (which may hold
8308 0, @minus{}1, or a count of the function pointers after it, depending on
8309 the environment). This is followed by a series of zero or more function
8310 pointers to constructors (or destructors), followed by a function
8311 pointer containing zero.
8313 Depending on the operating system and its executable file format, either
8314 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8315 time and exit time. Constructors are called in reverse order of the
8316 list; destructors in forward order.
8318 The best way to handle static constructors works only for object file
8319 formats which provide arbitrarily-named sections. A section is set
8320 aside for a list of constructors, and another for a list of destructors.
8321 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8322 object file that defines an initialization function also puts a word in
8323 the constructor section to point to that function. The linker
8324 accumulates all these words into one contiguous @samp{.ctors} section.
8325 Termination functions are handled similarly.
8327 This method will be chosen as the default by @file{target-def.h} if
8328 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8329 support arbitrary sections, but does support special designated
8330 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8331 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8333 When arbitrary sections are available, there are two variants, depending
8334 upon how the code in @file{crtstuff.c} is called. On systems that
8335 support a @dfn{.init} section which is executed at program startup,
8336 parts of @file{crtstuff.c} are compiled into that section. The
8337 program is linked by the @command{gcc} driver like this:
8340 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8343 The prologue of a function (@code{__init}) appears in the @code{.init}
8344 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8345 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8346 files are provided by the operating system or by the GNU C library, but
8347 are provided by GCC for a few targets.
8349 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8350 compiled from @file{crtstuff.c}. They contain, among other things, code
8351 fragments within the @code{.init} and @code{.fini} sections that branch
8352 to routines in the @code{.text} section. The linker will pull all parts
8353 of a section together, which results in a complete @code{__init} function
8354 that invokes the routines we need at startup.
8356 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8359 If no init section is available, when GCC compiles any function called
8360 @code{main} (or more accurately, any function designated as a program
8361 entry point by the language front end calling @code{expand_main_function}),
8362 it inserts a procedure call to @code{__main} as the first executable code
8363 after the function prologue. The @code{__main} function is defined
8364 in @file{libgcc2.c} and runs the global constructors.
8366 In file formats that don't support arbitrary sections, there are again
8367 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8368 and an `a.out' format must be used. In this case,
8369 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8370 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8371 and with the address of the void function containing the initialization
8372 code as its value. The GNU linker recognizes this as a request to add
8373 the value to a @dfn{set}; the values are accumulated, and are eventually
8374 placed in the executable as a vector in the format described above, with
8375 a leading (ignored) count and a trailing zero element.
8376 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8377 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8378 the compilation of @code{main} to call @code{__main} as above, starting
8379 the initialization process.
8381 The last variant uses neither arbitrary sections nor the GNU linker.
8382 This is preferable when you want to do dynamic linking and when using
8383 file formats which the GNU linker does not support, such as `ECOFF'@. In
8384 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8385 termination functions are recognized simply by their names. This requires
8386 an extra program in the linkage step, called @command{collect2}. This program
8387 pretends to be the linker, for use with GCC; it does its job by running
8388 the ordinary linker, but also arranges to include the vectors of
8389 initialization and termination functions. These functions are called
8390 via @code{__main} as described above. In order to use this method,
8391 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8394 The following section describes the specific macros that control and
8395 customize the handling of initialization and termination functions.
8398 @node Macros for Initialization
8399 @subsection Macros Controlling Initialization Routines
8401 Here are the macros that control how the compiler handles initialization
8402 and termination functions:
8404 @defmac INIT_SECTION_ASM_OP
8405 If defined, a C string constant, including spacing, for the assembler
8406 operation to identify the following data as initialization code. If not
8407 defined, GCC will assume such a section does not exist. When you are
8408 using special sections for initialization and termination functions, this
8409 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8410 run the initialization functions.
8413 @defmac HAS_INIT_SECTION
8414 If defined, @code{main} will not call @code{__main} as described above.
8415 This macro should be defined for systems that control start-up code
8416 on a symbol-by-symbol basis, such as OSF/1, and should not
8417 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8420 @defmac LD_INIT_SWITCH
8421 If defined, a C string constant for a switch that tells the linker that
8422 the following symbol is an initialization routine.
8425 @defmac LD_FINI_SWITCH
8426 If defined, a C string constant for a switch that tells the linker that
8427 the following symbol is a finalization routine.
8430 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8431 If defined, a C statement that will write a function that can be
8432 automatically called when a shared library is loaded. The function
8433 should call @var{func}, which takes no arguments. If not defined, and
8434 the object format requires an explicit initialization function, then a
8435 function called @code{_GLOBAL__DI} will be generated.
8437 This function and the following one are used by collect2 when linking a
8438 shared library that needs constructors or destructors, or has DWARF2
8439 exception tables embedded in the code.
8442 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8443 If defined, a C statement that will write a function that can be
8444 automatically called when a shared library is unloaded. The function
8445 should call @var{func}, which takes no arguments. If not defined, and
8446 the object format requires an explicit finalization function, then a
8447 function called @code{_GLOBAL__DD} will be generated.
8450 @defmac INVOKE__main
8451 If defined, @code{main} will call @code{__main} despite the presence of
8452 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8453 where the init section is not actually run automatically, but is still
8454 useful for collecting the lists of constructors and destructors.
8457 @defmac SUPPORTS_INIT_PRIORITY
8458 If nonzero, the C++ @code{init_priority} attribute is supported and the
8459 compiler should emit instructions to control the order of initialization
8460 of objects. If zero, the compiler will issue an error message upon
8461 encountering an @code{init_priority} attribute.
8464 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8465 This value is true if the target supports some ``native'' method of
8466 collecting constructors and destructors to be run at startup and exit.
8467 It is false if we must use @command{collect2}.
8470 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8471 If defined, a function that outputs assembler code to arrange to call
8472 the function referenced by @var{symbol} at initialization time.
8474 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8475 no arguments and with no return value. If the target supports initialization
8476 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8477 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8479 If this macro is not defined by the target, a suitable default will
8480 be chosen if (1) the target supports arbitrary section names, (2) the
8481 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8485 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8486 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8487 functions rather than initialization functions.
8490 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8491 generated for the generated object file will have static linkage.
8493 If your system uses @command{collect2} as the means of processing
8494 constructors, then that program normally uses @command{nm} to scan
8495 an object file for constructor functions to be called.
8497 On certain kinds of systems, you can define this macro to make
8498 @command{collect2} work faster (and, in some cases, make it work at all):
8500 @defmac OBJECT_FORMAT_COFF
8501 Define this macro if the system uses COFF (Common Object File Format)
8502 object files, so that @command{collect2} can assume this format and scan
8503 object files directly for dynamic constructor/destructor functions.
8505 This macro is effective only in a native compiler; @command{collect2} as
8506 part of a cross compiler always uses @command{nm} for the target machine.
8509 @defmac REAL_NM_FILE_NAME
8510 Define this macro as a C string constant containing the file name to use
8511 to execute @command{nm}. The default is to search the path normally for
8516 @command{collect2} calls @command{nm} to scan object files for static
8517 constructors and destructors and LTO info. By default, @option{-n} is
8518 passed. Define @code{NM_FLAGS} to a C string constant if other options
8519 are needed to get the same output format as GNU @command{nm -n}
8523 If your system supports shared libraries and has a program to list the
8524 dynamic dependencies of a given library or executable, you can define
8525 these macros to enable support for running initialization and
8526 termination functions in shared libraries:
8529 Define this macro to a C string constant containing the name of the program
8530 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8533 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8534 Define this macro to be C code that extracts filenames from the output
8535 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8536 of type @code{char *} that points to the beginning of a line of output
8537 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8538 code must advance @var{ptr} to the beginning of the filename on that
8539 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8542 @defmac SHLIB_SUFFIX
8543 Define this macro to a C string constant containing the default shared
8544 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8545 strips version information after this suffix when generating global
8546 constructor and destructor names. This define is only needed on targets
8547 that use @command{collect2} to process constructors and destructors.
8550 @node Instruction Output
8551 @subsection Output of Assembler Instructions
8553 @c prevent bad page break with this line
8554 This describes assembler instruction output.
8556 @defmac REGISTER_NAMES
8557 A C initializer containing the assembler's names for the machine
8558 registers, each one as a C string constant. This is what translates
8559 register numbers in the compiler into assembler language.
8562 @defmac ADDITIONAL_REGISTER_NAMES
8563 If defined, a C initializer for an array of structures containing a name
8564 and a register number. This macro defines additional names for hard
8565 registers, thus allowing the @code{asm} option in declarations to refer
8566 to registers using alternate names.
8569 @defmac OVERLAPPING_REGISTER_NAMES
8570 If defined, a C initializer for an array of structures containing a
8571 name, a register number and a count of the number of consecutive
8572 machine registers the name overlaps. This macro defines additional
8573 names for hard registers, thus allowing the @code{asm} option in
8574 declarations to refer to registers using alternate names. Unlike
8575 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8576 register name implies multiple underlying registers.
8578 This macro should be used when it is important that a clobber in an
8579 @code{asm} statement clobbers all the underlying values implied by the
8580 register name. For example, on ARM, clobbering the double-precision
8581 VFP register ``d0'' implies clobbering both single-precision registers
8585 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8586 Define this macro if you are using an unusual assembler that
8587 requires different names for the machine instructions.
8589 The definition is a C statement or statements which output an
8590 assembler instruction opcode to the stdio stream @var{stream}. The
8591 macro-operand @var{ptr} is a variable of type @code{char *} which
8592 points to the opcode name in its ``internal'' form---the form that is
8593 written in the machine description. The definition should output the
8594 opcode name to @var{stream}, performing any translation you desire, and
8595 increment the variable @var{ptr} to point at the end of the opcode
8596 so that it will not be output twice.
8598 In fact, your macro definition may process less than the entire opcode
8599 name, or more than the opcode name; but if you want to process text
8600 that includes @samp{%}-sequences to substitute operands, you must take
8601 care of the substitution yourself. Just be sure to increment
8602 @var{ptr} over whatever text should not be output normally.
8604 @findex recog_data.operand
8605 If you need to look at the operand values, they can be found as the
8606 elements of @code{recog_data.operand}.
8608 If the macro definition does nothing, the instruction is output
8612 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8613 If defined, a C statement to be executed just prior to the output of
8614 assembler code for @var{insn}, to modify the extracted operands so
8615 they will be output differently.
8617 Here the argument @var{opvec} is the vector containing the operands
8618 extracted from @var{insn}, and @var{noperands} is the number of
8619 elements of the vector which contain meaningful data for this insn.
8620 The contents of this vector are what will be used to convert the insn
8621 template into assembler code, so you can change the assembler output
8622 by changing the contents of the vector.
8624 This macro is useful when various assembler syntaxes share a single
8625 file of instruction patterns; by defining this macro differently, you
8626 can cause a large class of instructions to be output differently (such
8627 as with rearranged operands). Naturally, variations in assembler
8628 syntax affecting individual insn patterns ought to be handled by
8629 writing conditional output routines in those patterns.
8631 If this macro is not defined, it is equivalent to a null statement.
8634 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx @var{insn}, rtx *@var{opvec}, int @var{noperands})
8635 If defined, this target hook is a function which is executed just after the
8636 output of assembler code for @var{insn}, to change the mode of the assembler
8639 Here the argument @var{opvec} is the vector containing the operands
8640 extracted from @var{insn}, and @var{noperands} is the number of
8641 elements of the vector which contain meaningful data for this insn.
8642 The contents of this vector are what was used to convert the insn
8643 template into assembler code, so you can change the assembler mode
8644 by checking the contents of the vector.
8647 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8648 A C compound statement to output to stdio stream @var{stream} the
8649 assembler syntax for an instruction operand @var{x}. @var{x} is an
8652 @var{code} is a value that can be used to specify one of several ways
8653 of printing the operand. It is used when identical operands must be
8654 printed differently depending on the context. @var{code} comes from
8655 the @samp{%} specification that was used to request printing of the
8656 operand. If the specification was just @samp{%@var{digit}} then
8657 @var{code} is 0; if the specification was @samp{%@var{ltr}
8658 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8661 If @var{x} is a register, this macro should print the register's name.
8662 The names can be found in an array @code{reg_names} whose type is
8663 @code{char *[]}. @code{reg_names} is initialized from
8664 @code{REGISTER_NAMES}.
8666 When the machine description has a specification @samp{%@var{punct}}
8667 (a @samp{%} followed by a punctuation character), this macro is called
8668 with a null pointer for @var{x} and the punctuation character for
8672 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8673 A C expression which evaluates to true if @var{code} is a valid
8674 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8675 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8676 punctuation characters (except for the standard one, @samp{%}) are used
8680 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8681 A C compound statement to output to stdio stream @var{stream} the
8682 assembler syntax for an instruction operand that is a memory reference
8683 whose address is @var{x}. @var{x} is an RTL expression.
8685 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8686 On some machines, the syntax for a symbolic address depends on the
8687 section that the address refers to. On these machines, define the hook
8688 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8689 @code{symbol_ref}, and then check for it here. @xref{Assembler
8693 @findex dbr_sequence_length
8694 @defmac DBR_OUTPUT_SEQEND (@var{file})
8695 A C statement, to be executed after all slot-filler instructions have
8696 been output. If necessary, call @code{dbr_sequence_length} to
8697 determine the number of slots filled in a sequence (zero if not
8698 currently outputting a sequence), to decide how many no-ops to output,
8701 Don't define this macro if it has nothing to do, but it is helpful in
8702 reading assembly output if the extent of the delay sequence is made
8703 explicit (e.g.@: with white space).
8706 @findex final_sequence
8707 Note that output routines for instructions with delay slots must be
8708 prepared to deal with not being output as part of a sequence
8709 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8710 found.) The variable @code{final_sequence} is null when not
8711 processing a sequence, otherwise it contains the @code{sequence} rtx
8715 @defmac REGISTER_PREFIX
8716 @defmacx LOCAL_LABEL_PREFIX
8717 @defmacx USER_LABEL_PREFIX
8718 @defmacx IMMEDIATE_PREFIX
8719 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8720 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8721 @file{final.c}). These are useful when a single @file{md} file must
8722 support multiple assembler formats. In that case, the various @file{tm.h}
8723 files can define these macros differently.
8726 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8727 If defined this macro should expand to a series of @code{case}
8728 statements which will be parsed inside the @code{switch} statement of
8729 the @code{asm_fprintf} function. This allows targets to define extra
8730 printf formats which may useful when generating their assembler
8731 statements. Note that uppercase letters are reserved for future
8732 generic extensions to asm_fprintf, and so are not available to target
8733 specific code. The output file is given by the parameter @var{file}.
8734 The varargs input pointer is @var{argptr} and the rest of the format
8735 string, starting the character after the one that is being switched
8736 upon, is pointed to by @var{format}.
8739 @defmac ASSEMBLER_DIALECT
8740 If your target supports multiple dialects of assembler language (such as
8741 different opcodes), define this macro as a C expression that gives the
8742 numeric index of the assembler language dialect to use, with zero as the
8745 If this macro is defined, you may use constructs of the form
8747 @samp{@{option0|option1|option2@dots{}@}}
8750 in the output templates of patterns (@pxref{Output Template}) or in the
8751 first argument of @code{asm_fprintf}. This construct outputs
8752 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8753 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8754 within these strings retain their usual meaning. If there are fewer
8755 alternatives within the braces than the value of
8756 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8758 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8759 @samp{@}} do not have any special meaning when used in templates or
8760 operands to @code{asm_fprintf}.
8762 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8763 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8764 the variations in assembler language syntax with that mechanism. Define
8765 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8766 if the syntax variant are larger and involve such things as different
8767 opcodes or operand order.
8770 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8771 A C expression to output to @var{stream} some assembler code
8772 which will push hard register number @var{regno} onto the stack.
8773 The code need not be optimal, since this macro is used only when
8777 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8778 A C expression to output to @var{stream} some assembler code
8779 which will pop hard register number @var{regno} off of the stack.
8780 The code need not be optimal, since this macro is used only when
8784 @node Dispatch Tables
8785 @subsection Output of Dispatch Tables
8787 @c prevent bad page break with this line
8788 This concerns dispatch tables.
8790 @cindex dispatch table
8791 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8792 A C statement to output to the stdio stream @var{stream} an assembler
8793 pseudo-instruction to generate a difference between two labels.
8794 @var{value} and @var{rel} are the numbers of two internal labels. The
8795 definitions of these labels are output using
8796 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8797 way here. For example,
8800 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8801 @var{value}, @var{rel})
8804 You must provide this macro on machines where the addresses in a
8805 dispatch table are relative to the table's own address. If defined, GCC
8806 will also use this macro on all machines when producing PIC@.
8807 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8808 mode and flags can be read.
8811 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8812 This macro should be provided on machines where the addresses
8813 in a dispatch table are absolute.
8815 The definition should be a C statement to output to the stdio stream
8816 @var{stream} an assembler pseudo-instruction to generate a reference to
8817 a label. @var{value} is the number of an internal label whose
8818 definition is output using @code{(*targetm.asm_out.internal_label)}.
8822 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8826 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8827 Define this if the label before a jump-table needs to be output
8828 specially. The first three arguments are the same as for
8829 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8830 jump-table which follows (a @code{jump_insn} containing an
8831 @code{addr_vec} or @code{addr_diff_vec}).
8833 This feature is used on system V to output a @code{swbeg} statement
8836 If this macro is not defined, these labels are output with
8837 @code{(*targetm.asm_out.internal_label)}.
8840 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8841 Define this if something special must be output at the end of a
8842 jump-table. The definition should be a C statement to be executed
8843 after the assembler code for the table is written. It should write
8844 the appropriate code to stdio stream @var{stream}. The argument
8845 @var{table} is the jump-table insn, and @var{num} is the label-number
8846 of the preceding label.
8848 If this macro is not defined, nothing special is output at the end of
8852 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
8853 This target hook emits a label at the beginning of each FDE@. It
8854 should be defined on targets where FDEs need special labels, and it
8855 should write the appropriate label, for the FDE associated with the
8856 function declaration @var{decl}, to the stdio stream @var{stream}.
8857 The third argument, @var{for_eh}, is a boolean: true if this is for an
8858 exception table. The fourth argument, @var{empty}, is a boolean:
8859 true if this is a placeholder label for an omitted FDE@.
8861 The default is that FDEs are not given nonlocal labels.
8864 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
8865 This target hook emits a label at the beginning of the exception table.
8866 It should be defined on targets where it is desirable for the table
8867 to be broken up according to function.
8869 The default is that no label is emitted.
8872 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
8873 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.
8876 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx @var{insn})
8877 This target hook emits assembly directives required to unwind the
8878 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8879 returns @code{UI_TARGET}.
8882 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8883 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.
8886 @node Exception Region Output
8887 @subsection Assembler Commands for Exception Regions
8889 @c prevent bad page break with this line
8891 This describes commands marking the start and the end of an exception
8894 @defmac EH_FRAME_SECTION_NAME
8895 If defined, a C string constant for the name of the section containing
8896 exception handling frame unwind information. If not defined, GCC will
8897 provide a default definition if the target supports named sections.
8898 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8900 You should define this symbol if your target supports DWARF 2 frame
8901 unwind information and the default definition does not work.
8904 @defmac EH_FRAME_IN_DATA_SECTION
8905 If defined, DWARF 2 frame unwind information will be placed in the
8906 data section even though the target supports named sections. This
8907 might be necessary, for instance, if the system linker does garbage
8908 collection and sections cannot be marked as not to be collected.
8910 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8914 @defmac EH_TABLES_CAN_BE_READ_ONLY
8915 Define this macro to 1 if your target is such that no frame unwind
8916 information encoding used with non-PIC code will ever require a
8917 runtime relocation, but the linker may not support merging read-only
8918 and read-write sections into a single read-write section.
8921 @defmac MASK_RETURN_ADDR
8922 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8923 that it does not contain any extraneous set bits in it.
8926 @defmac DWARF2_UNWIND_INFO
8927 Define this macro to 0 if your target supports DWARF 2 frame unwind
8928 information, but it does not yet work with exception handling.
8929 Otherwise, if your target supports this information (if it defines
8930 @code{INCOMING_RETURN_ADDR_RTX} and either @code{UNALIGNED_INT_ASM_OP}
8931 or @code{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8934 @deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
8935 This hook defines the mechanism that will be used for exception handling
8936 by the target. If the target has ABI specified unwind tables, the hook
8937 should return @code{UI_TARGET}. If the target is to use the
8938 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8939 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8940 information, the hook should return @code{UI_DWARF2}.
8942 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8943 This may end up simplifying other parts of target-specific code. The
8944 default implementation of this hook never returns @code{UI_NONE}.
8946 Note that the value returned by this hook should be constant. It should
8947 not depend on anything except the command-line switches described by
8948 @var{opts}. In particular, the
8949 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8950 macros and builtin functions related to exception handling are set up
8951 depending on this setting.
8953 The default implementation of the hook first honors the
8954 @option{--enable-sjlj-exceptions} configure option, then
8955 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
8956 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
8957 must define this hook so that @var{opts} is used correctly.
8960 @deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8961 This variable should be set to @code{true} if the target ABI requires unwinding
8962 tables even when exceptions are not used. It must not be modified by
8963 command-line option processing.
8966 @defmac DONT_USE_BUILTIN_SETJMP
8967 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8968 should use the @code{setjmp}/@code{longjmp} functions from the C library
8969 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8972 @defmac DWARF_CIE_DATA_ALIGNMENT
8973 This macro need only be defined if the target might save registers in the
8974 function prologue at an offset to the stack pointer that is not aligned to
8975 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8976 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8977 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8978 the target supports DWARF 2 frame unwind information.
8981 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8982 Contains the value true if the target should add a zero word onto the
8983 end of a Dwarf-2 frame info section when used for exception handling.
8984 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8988 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8989 Given a register, this hook should return a parallel of registers to
8990 represent where to find the register pieces. Define this hook if the
8991 register and its mode are represented in Dwarf in non-contiguous
8992 locations, or if the register should be represented in more than one
8993 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8994 If not defined, the default is to return @code{NULL_RTX}.
8997 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8998 If some registers are represented in Dwarf-2 unwind information in
8999 multiple pieces, define this hook to fill in information about the
9000 sizes of those pieces in the table used by the unwinder at runtime.
9001 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
9002 filling in a single size corresponding to each hard register;
9003 @var{address} is the address of the table.
9006 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
9007 This hook is used to output a reference from a frame unwinding table to
9008 the type_info object identified by @var{sym}. It should return @code{true}
9009 if the reference was output. Returning @code{false} will cause the
9010 reference to be output using the normal Dwarf2 routines.
9013 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
9014 This flag should be set to @code{true} on targets that use an ARM EABI
9015 based unwinding library, and @code{false} on other targets. This effects
9016 the format of unwinding tables, and how the unwinder in entered after
9017 running a cleanup. The default is @code{false}.
9020 @node Alignment Output
9021 @subsection Assembler Commands for Alignment
9023 @c prevent bad page break with this line
9024 This describes commands for alignment.
9026 @defmac JUMP_ALIGN (@var{label})
9027 The alignment (log base 2) to put in front of @var{label}, which is
9028 a common destination of jumps and has no fallthru incoming edge.
9030 This macro need not be defined if you don't want any special alignment
9031 to be done at such a time. Most machine descriptions do not currently
9034 Unless it's necessary to inspect the @var{label} parameter, it is better
9035 to set the variable @var{align_jumps} in the target's
9036 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9037 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
9040 @deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx @var{label})
9041 The maximum number of bytes to skip before @var{label} when applying
9042 @code{JUMP_ALIGN}. This works only if
9043 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9046 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
9047 The alignment (log base 2) to put in front of @var{label}, which follows
9050 This macro need not be defined if you don't want any special alignment
9051 to be done at such a time. Most machine descriptions do not currently
9055 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx @var{label})
9056 The maximum number of bytes to skip before @var{label} when applying
9057 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
9058 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9061 @defmac LOOP_ALIGN (@var{label})
9062 The alignment (log base 2) to put in front of @var{label}, which follows
9063 a @code{NOTE_INSN_LOOP_BEG} note.
9065 This macro need not be defined if you don't want any special alignment
9066 to be done at such a time. Most machine descriptions do not currently
9069 Unless it's necessary to inspect the @var{label} parameter, it is better
9070 to set the variable @code{align_loops} in the target's
9071 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9072 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
9075 @deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx @var{label})
9076 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
9077 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
9081 @defmac LABEL_ALIGN (@var{label})
9082 The alignment (log base 2) to put in front of @var{label}.
9083 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
9084 the maximum of the specified values is used.
9086 Unless it's necessary to inspect the @var{label} parameter, it is better
9087 to set the variable @code{align_labels} in the target's
9088 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9089 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9092 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx @var{label})
9093 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
9094 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
9098 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9099 A C statement to output to the stdio stream @var{stream} an assembler
9100 instruction to advance the location counter by @var{nbytes} bytes.
9101 Those bytes should be zero when loaded. @var{nbytes} will be a C
9102 expression of type @code{unsigned HOST_WIDE_INT}.
9105 @defmac ASM_NO_SKIP_IN_TEXT
9106 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9107 text section because it fails to put zeros in the bytes that are skipped.
9108 This is true on many Unix systems, where the pseudo--op to skip bytes
9109 produces no-op instructions rather than zeros when used in the text
9113 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9114 A C statement to output to the stdio stream @var{stream} an assembler
9115 command to advance the location counter to a multiple of 2 to the
9116 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9119 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9120 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9121 for padding, if necessary.
9124 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9125 A C statement to output to the stdio stream @var{stream} an assembler
9126 command to advance the location counter to a multiple of 2 to the
9127 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9128 satisfy the alignment request. @var{power} and @var{max_skip} will be
9129 a C expression of type @code{int}.
9133 @node Debugging Info
9134 @section Controlling Debugging Information Format
9136 @c prevent bad page break with this line
9137 This describes how to specify debugging information.
9140 * All Debuggers:: Macros that affect all debugging formats uniformly.
9141 * DBX Options:: Macros enabling specific options in DBX format.
9142 * DBX Hooks:: Hook macros for varying DBX format.
9143 * File Names and DBX:: Macros controlling output of file names in DBX format.
9144 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9145 * VMS Debug:: Macros for VMS debug format.
9149 @subsection Macros Affecting All Debugging Formats
9151 @c prevent bad page break with this line
9152 These macros affect all debugging formats.
9154 @defmac DBX_REGISTER_NUMBER (@var{regno})
9155 A C expression that returns the DBX register number for the compiler
9156 register number @var{regno}. In the default macro provided, the value
9157 of this expression will be @var{regno} itself. But sometimes there are
9158 some registers that the compiler knows about and DBX does not, or vice
9159 versa. In such cases, some register may need to have one number in the
9160 compiler and another for DBX@.
9162 If two registers have consecutive numbers inside GCC, and they can be
9163 used as a pair to hold a multiword value, then they @emph{must} have
9164 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9165 Otherwise, debuggers will be unable to access such a pair, because they
9166 expect register pairs to be consecutive in their own numbering scheme.
9168 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9169 does not preserve register pairs, then what you must do instead is
9170 redefine the actual register numbering scheme.
9173 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9174 A C expression that returns the integer offset value for an automatic
9175 variable having address @var{x} (an RTL expression). The default
9176 computation assumes that @var{x} is based on the frame-pointer and
9177 gives the offset from the frame-pointer. This is required for targets
9178 that produce debugging output for DBX or COFF-style debugging output
9179 for SDB and allow the frame-pointer to be eliminated when the
9180 @option{-g} options is used.
9183 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9184 A C expression that returns the integer offset value for an argument
9185 having address @var{x} (an RTL expression). The nominal offset is
9189 @defmac PREFERRED_DEBUGGING_TYPE
9190 A C expression that returns the type of debugging output GCC should
9191 produce when the user specifies just @option{-g}. Define
9192 this if you have arranged for GCC to support more than one format of
9193 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9194 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9195 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9197 When the user specifies @option{-ggdb}, GCC normally also uses the
9198 value of this macro to select the debugging output format, but with two
9199 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9200 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9201 defined, GCC uses @code{DBX_DEBUG}.
9203 The value of this macro only affects the default debugging output; the
9204 user can always get a specific type of output by using @option{-gstabs},
9205 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9209 @subsection Specific Options for DBX Output
9211 @c prevent bad page break with this line
9212 These are specific options for DBX output.
9214 @defmac DBX_DEBUGGING_INFO
9215 Define this macro if GCC should produce debugging output for DBX
9216 in response to the @option{-g} option.
9219 @defmac XCOFF_DEBUGGING_INFO
9220 Define this macro if GCC should produce XCOFF format debugging output
9221 in response to the @option{-g} option. This is a variant of DBX format.
9224 @defmac DEFAULT_GDB_EXTENSIONS
9225 Define this macro to control whether GCC should by default generate
9226 GDB's extended version of DBX debugging information (assuming DBX-format
9227 debugging information is enabled at all). If you don't define the
9228 macro, the default is 1: always generate the extended information
9229 if there is any occasion to.
9232 @defmac DEBUG_SYMS_TEXT
9233 Define this macro if all @code{.stabs} commands should be output while
9234 in the text section.
9237 @defmac ASM_STABS_OP
9238 A C string constant, including spacing, naming the assembler pseudo op to
9239 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9240 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9241 applies only to DBX debugging information format.
9244 @defmac ASM_STABD_OP
9245 A C string constant, including spacing, naming the assembler pseudo op to
9246 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9247 value is the current location. If you don't define this macro,
9248 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9252 @defmac ASM_STABN_OP
9253 A C string constant, including spacing, naming the assembler pseudo op to
9254 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9255 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9256 macro applies only to DBX debugging information format.
9259 @defmac DBX_NO_XREFS
9260 Define this macro if DBX on your system does not support the construct
9261 @samp{xs@var{tagname}}. On some systems, this construct is used to
9262 describe a forward reference to a structure named @var{tagname}.
9263 On other systems, this construct is not supported at all.
9266 @defmac DBX_CONTIN_LENGTH
9267 A symbol name in DBX-format debugging information is normally
9268 continued (split into two separate @code{.stabs} directives) when it
9269 exceeds a certain length (by default, 80 characters). On some
9270 operating systems, DBX requires this splitting; on others, splitting
9271 must not be done. You can inhibit splitting by defining this macro
9272 with the value zero. You can override the default splitting-length by
9273 defining this macro as an expression for the length you desire.
9276 @defmac DBX_CONTIN_CHAR
9277 Normally continuation is indicated by adding a @samp{\} character to
9278 the end of a @code{.stabs} string when a continuation follows. To use
9279 a different character instead, define this macro as a character
9280 constant for the character you want to use. Do not define this macro
9281 if backslash is correct for your system.
9284 @defmac DBX_STATIC_STAB_DATA_SECTION
9285 Define this macro if it is necessary to go to the data section before
9286 outputting the @samp{.stabs} pseudo-op for a non-global static
9290 @defmac DBX_TYPE_DECL_STABS_CODE
9291 The value to use in the ``code'' field of the @code{.stabs} directive
9292 for a typedef. The default is @code{N_LSYM}.
9295 @defmac DBX_STATIC_CONST_VAR_CODE
9296 The value to use in the ``code'' field of the @code{.stabs} directive
9297 for a static variable located in the text section. DBX format does not
9298 provide any ``right'' way to do this. The default is @code{N_FUN}.
9301 @defmac DBX_REGPARM_STABS_CODE
9302 The value to use in the ``code'' field of the @code{.stabs} directive
9303 for a parameter passed in registers. DBX format does not provide any
9304 ``right'' way to do this. The default is @code{N_RSYM}.
9307 @defmac DBX_REGPARM_STABS_LETTER
9308 The letter to use in DBX symbol data to identify a symbol as a parameter
9309 passed in registers. DBX format does not customarily provide any way to
9310 do this. The default is @code{'P'}.
9313 @defmac DBX_FUNCTION_FIRST
9314 Define this macro if the DBX information for a function and its
9315 arguments should precede the assembler code for the function. Normally,
9316 in DBX format, the debugging information entirely follows the assembler
9320 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9321 Define this macro, with value 1, if the value of a symbol describing
9322 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9323 relative to the start of the enclosing function. Normally, GCC uses
9324 an absolute address.
9327 @defmac DBX_LINES_FUNCTION_RELATIVE
9328 Define this macro, with value 1, if the value of a symbol indicating
9329 the current line number (@code{N_SLINE}) should be relative to the
9330 start of the enclosing function. Normally, GCC uses an absolute address.
9333 @defmac DBX_USE_BINCL
9334 Define this macro if GCC should generate @code{N_BINCL} and
9335 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9336 macro also directs GCC to output a type number as a pair of a file
9337 number and a type number within the file. Normally, GCC does not
9338 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9339 number for a type number.
9343 @subsection Open-Ended Hooks for DBX Format
9345 @c prevent bad page break with this line
9346 These are hooks for DBX format.
9348 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9349 Define this macro to say how to output to @var{stream} the debugging
9350 information for the start of a scope level for variable names. The
9351 argument @var{name} is the name of an assembler symbol (for use with
9352 @code{assemble_name}) whose value is the address where the scope begins.
9355 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9356 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9359 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9360 Define this macro if the target machine requires special handling to
9361 output an @code{N_FUN} entry for the function @var{decl}.
9364 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9365 A C statement to output DBX debugging information before code for line
9366 number @var{line} of the current source file to the stdio stream
9367 @var{stream}. @var{counter} is the number of time the macro was
9368 invoked, including the current invocation; it is intended to generate
9369 unique labels in the assembly output.
9371 This macro should not be defined if the default output is correct, or
9372 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9375 @defmac NO_DBX_FUNCTION_END
9376 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9377 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9378 On those machines, define this macro to turn this feature off without
9379 disturbing the rest of the gdb extensions.
9382 @defmac NO_DBX_BNSYM_ENSYM
9383 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9384 extension construct. On those machines, define this macro to turn this
9385 feature off without disturbing the rest of the gdb extensions.
9388 @node File Names and DBX
9389 @subsection File Names in DBX Format
9391 @c prevent bad page break with this line
9392 This describes file names in DBX format.
9394 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9395 A C statement to output DBX debugging information to the stdio stream
9396 @var{stream}, which indicates that file @var{name} is the main source
9397 file---the file specified as the input file for compilation.
9398 This macro is called only once, at the beginning of compilation.
9400 This macro need not be defined if the standard form of output
9401 for DBX debugging information is appropriate.
9403 It may be necessary to refer to a label equal to the beginning of the
9404 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9405 to do so. If you do this, you must also set the variable
9406 @var{used_ltext_label_name} to @code{true}.
9409 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9410 Define this macro, with value 1, if GCC should not emit an indication
9411 of the current directory for compilation and current source language at
9412 the beginning of the file.
9415 @defmac NO_DBX_GCC_MARKER
9416 Define this macro, with value 1, if GCC should not emit an indication
9417 that this object file was compiled by GCC@. The default is to emit
9418 an @code{N_OPT} stab at the beginning of every source file, with
9419 @samp{gcc2_compiled.} for the string and value 0.
9422 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9423 A C statement to output DBX debugging information at the end of
9424 compilation of the main source file @var{name}. Output should be
9425 written to the stdio stream @var{stream}.
9427 If you don't define this macro, nothing special is output at the end
9428 of compilation, which is correct for most machines.
9431 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9432 Define this macro @emph{instead of} defining
9433 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9434 the end of compilation is an @code{N_SO} stab with an empty string,
9435 whose value is the highest absolute text address in the file.
9440 @subsection Macros for SDB and DWARF Output
9442 @c prevent bad page break with this line
9443 Here are macros for SDB and DWARF output.
9445 @defmac SDB_DEBUGGING_INFO
9446 Define this macro if GCC should produce COFF-style debugging output
9447 for SDB in response to the @option{-g} option.
9450 @defmac DWARF2_DEBUGGING_INFO
9451 Define this macro if GCC should produce dwarf version 2 format
9452 debugging output in response to the @option{-g} option.
9454 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
9455 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9456 be emitted for each function. Instead of an integer return the enum
9457 value for the @code{DW_CC_} tag.
9460 To support optional call frame debugging information, you must also
9461 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9462 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9463 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9464 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9467 @defmac DWARF2_FRAME_INFO
9468 Define this macro to a nonzero value if GCC should always output
9469 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9470 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9471 exceptions are enabled, GCC will output this information not matter
9472 how you define @code{DWARF2_FRAME_INFO}.
9475 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
9476 This hook defines the mechanism that will be used for describing frame
9477 unwind information to the debugger. Normally the hook will return
9478 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9479 return @code{UI_NONE} otherwise.
9481 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9482 is disabled in order to always output DWARF 2 frame information.
9484 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9485 This will suppress generation of the normal debug frame unwind information.
9488 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9489 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9490 line debug info sections. This will result in much more compact line number
9491 tables, and hence is desirable if it works.
9494 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
9495 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.
9498 @deftypevr {Target Hook} bool TARGET_DELAY_SCHED2
9499 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.
9502 @deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK
9503 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.
9506 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9507 A C statement to issue assembly directives that create a difference
9508 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9511 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9512 A C statement to issue assembly directives that create a difference
9513 between the two given labels in system defined units, e.g. instruction
9514 slots on IA64 VMS, using an integer of the given size.
9517 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9518 A C statement to issue assembly directives that create a
9519 section-relative reference to the given @var{label}, using an integer of the
9520 given @var{size}. The label is known to be defined in the given @var{section}.
9523 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9524 A C statement to issue assembly directives that create a self-relative
9525 reference to the given @var{label}, using an integer of the given @var{size}.
9528 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9529 A C statement to issue assembly directives that create a reference to
9530 the DWARF table identifier @var{label} from the current section. This
9531 is used on some systems to avoid garbage collecting a DWARF table which
9532 is referenced by a function.
9535 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
9536 If defined, this target hook is a function which outputs a DTP-relative
9537 reference to the given TLS symbol of the specified size.
9540 @defmac PUT_SDB_@dots{}
9541 Define these macros to override the assembler syntax for the special
9542 SDB assembler directives. See @file{sdbout.c} for a list of these
9543 macros and their arguments. If the standard syntax is used, you need
9544 not define them yourself.
9548 Some assemblers do not support a semicolon as a delimiter, even between
9549 SDB assembler directives. In that case, define this macro to be the
9550 delimiter to use (usually @samp{\n}). It is not necessary to define
9551 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9555 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9556 Define this macro to allow references to unknown structure,
9557 union, or enumeration tags to be emitted. Standard COFF does not
9558 allow handling of unknown references, MIPS ECOFF has support for
9562 @defmac SDB_ALLOW_FORWARD_REFERENCES
9563 Define this macro to allow references to structure, union, or
9564 enumeration tags that have not yet been seen to be handled. Some
9565 assemblers choke if forward tags are used, while some require it.
9568 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9569 A C statement to output SDB debugging information before code for line
9570 number @var{line} of the current source file to the stdio stream
9571 @var{stream}. The default is to emit an @code{.ln} directive.
9576 @subsection Macros for VMS Debug Format
9578 @c prevent bad page break with this line
9579 Here are macros for VMS debug format.
9581 @defmac VMS_DEBUGGING_INFO
9582 Define this macro if GCC should produce debugging output for VMS
9583 in response to the @option{-g} option. The default behavior for VMS
9584 is to generate minimal debug info for a traceback in the absence of
9585 @option{-g} unless explicitly overridden with @option{-g0}. This
9586 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9587 @code{TARGET_OPTION_OVERRIDE}.
9590 @node Floating Point
9591 @section Cross Compilation and Floating Point
9592 @cindex cross compilation and floating point
9593 @cindex floating point and cross compilation
9595 While all modern machines use twos-complement representation for integers,
9596 there are a variety of representations for floating point numbers. This
9597 means that in a cross-compiler the representation of floating point numbers
9598 in the compiled program may be different from that used in the machine
9599 doing the compilation.
9601 Because different representation systems may offer different amounts of
9602 range and precision, all floating point constants must be represented in
9603 the target machine's format. Therefore, the cross compiler cannot
9604 safely use the host machine's floating point arithmetic; it must emulate
9605 the target's arithmetic. To ensure consistency, GCC always uses
9606 emulation to work with floating point values, even when the host and
9607 target floating point formats are identical.
9609 The following macros are provided by @file{real.h} for the compiler to
9610 use. All parts of the compiler which generate or optimize
9611 floating-point calculations must use these macros. They may evaluate
9612 their operands more than once, so operands must not have side effects.
9614 @defmac REAL_VALUE_TYPE
9615 The C data type to be used to hold a floating point value in the target
9616 machine's format. Typically this is a @code{struct} containing an
9617 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9621 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9622 Compares for equality the two values, @var{x} and @var{y}. If the target
9623 floating point format supports negative zeroes and/or NaNs,
9624 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9625 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9628 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9629 Tests whether @var{x} is less than @var{y}.
9632 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9633 Truncates @var{x} to a signed integer, rounding toward zero.
9636 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9637 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9638 @var{x} is negative, returns zero.
9641 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9642 Converts @var{string} into a floating point number in the target machine's
9643 representation for mode @var{mode}. This routine can handle both
9644 decimal and hexadecimal floating point constants, using the syntax
9645 defined by the C language for both.
9648 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9649 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9652 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9653 Determines whether @var{x} represents infinity (positive or negative).
9656 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9657 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9660 @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})
9661 Calculates an arithmetic operation on the two floating point values
9662 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9665 The operation to be performed is specified by @var{code}. Only the
9666 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9667 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9669 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9670 target's floating point format cannot represent infinity, it will call
9671 @code{abort}. Callers should check for this situation first, using
9672 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9675 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9676 Returns the negative of the floating point value @var{x}.
9679 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9680 Returns the absolute value of @var{x}.
9683 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9684 Truncates the floating point value @var{x} to fit in @var{mode}. The
9685 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9686 appropriate bit pattern to be output as a floating constant whose
9687 precision accords with mode @var{mode}.
9690 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9691 Converts a floating point value @var{x} into a double-precision integer
9692 which is then stored into @var{low} and @var{high}. If the value is not
9693 integral, it is truncated.
9696 @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})
9697 Converts a double-precision integer found in @var{low} and @var{high},
9698 into a floating point value which is then stored into @var{x}. The
9699 value is truncated to fit in mode @var{mode}.
9702 @node Mode Switching
9703 @section Mode Switching Instructions
9704 @cindex mode switching
9705 The following macros control mode switching optimizations:
9707 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9708 Define this macro if the port needs extra instructions inserted for mode
9709 switching in an optimizing compilation.
9711 For an example, the SH4 can perform both single and double precision
9712 floating point operations, but to perform a single precision operation,
9713 the FPSCR PR bit has to be cleared, while for a double precision
9714 operation, this bit has to be set. Changing the PR bit requires a general
9715 purpose register as a scratch register, hence these FPSCR sets have to
9716 be inserted before reload, i.e.@: you can't put this into instruction emitting
9717 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9719 You can have multiple entities that are mode-switched, and select at run time
9720 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9721 return nonzero for any @var{entity} that needs mode-switching.
9722 If you define this macro, you also have to define
9723 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9724 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9725 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9729 @defmac NUM_MODES_FOR_MODE_SWITCHING
9730 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9731 initializer for an array of integers. Each initializer element
9732 N refers to an entity that needs mode switching, and specifies the number
9733 of different modes that might need to be set for this entity.
9734 The position of the initializer in the initializer---starting counting at
9735 zero---determines the integer that is used to refer to the mode-switched
9737 In macros that take mode arguments / yield a mode result, modes are
9738 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9739 switch is needed / supplied.
9742 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9743 @var{entity} is an integer specifying a mode-switched entity. If
9744 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9745 return an integer value not larger than the corresponding element in
9746 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9747 be switched into prior to the execution of @var{insn}.
9750 @defmac MODE_AFTER (@var{mode}, @var{insn})
9751 If this macro is defined, it is evaluated for every @var{insn} during
9752 mode switching. It determines the mode that an insn results in (if
9753 different from the incoming mode).
9756 @defmac MODE_ENTRY (@var{entity})
9757 If this macro is defined, it is evaluated for every @var{entity} that needs
9758 mode switching. It should evaluate to an integer, which is a mode that
9759 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9760 is defined then @code{MODE_EXIT} must be defined.
9763 @defmac MODE_EXIT (@var{entity})
9764 If this macro is defined, it is evaluated for every @var{entity} that needs
9765 mode switching. It should evaluate to an integer, which is a mode that
9766 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9767 is defined then @code{MODE_ENTRY} must be defined.
9770 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9771 This macro specifies the order in which modes for @var{entity} are processed.
9772 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9773 lowest. The value of the macro should be an integer designating a mode
9774 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9775 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9776 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9779 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9780 Generate one or more insns to set @var{entity} to @var{mode}.
9781 @var{hard_reg_live} is the set of hard registers live at the point where
9782 the insn(s) are to be inserted.
9785 @node Target Attributes
9786 @section Defining target-specific uses of @code{__attribute__}
9787 @cindex target attributes
9788 @cindex machine attributes
9789 @cindex attributes, target-specific
9791 Target-specific attributes may be defined for functions, data and types.
9792 These are described using the following target hooks; they also need to
9793 be documented in @file{extend.texi}.
9795 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9796 If defined, this target hook points to an array of @samp{struct
9797 attribute_spec} (defined in @file{tree.h}) specifying the machine
9798 specific attributes for this target and some of the restrictions on the
9799 entities to which these attributes are applied and the arguments they
9803 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
9804 If defined, this target hook is a function which returns true if the
9805 machine-specific attribute named @var{name} expects an identifier
9806 given as its first argument to be passed on as a plain identifier, not
9807 subjected to name lookup. If this is not defined, the default is
9808 false for all machine-specific attributes.
9811 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
9812 If defined, this target hook is a function which returns zero if the attributes on
9813 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9814 and two if they are nearly compatible (which causes a warning to be
9815 generated). If this is not defined, machine-specific attributes are
9816 supposed always to be compatible.
9819 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9820 If defined, this target hook is a function which assigns default attributes to
9821 the newly defined @var{type}.
9824 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9825 Define this target hook if the merging of type attributes needs special
9826 handling. If defined, the result is a list of the combined
9827 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9828 that @code{comptypes} has already been called and returned 1. This
9829 function may call @code{merge_attributes} to handle machine-independent
9833 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9834 Define this target hook if the merging of decl attributes needs special
9835 handling. If defined, the result is a list of the combined
9836 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9837 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9838 when this is needed are when one attribute overrides another, or when an
9839 attribute is nullified by a subsequent definition. This function may
9840 call @code{merge_attributes} to handle machine-independent merging.
9842 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9843 If the only target-specific handling you require is @samp{dllimport}
9844 for Microsoft Windows targets, you should define the macro
9845 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9846 will then define a function called
9847 @code{merge_dllimport_decl_attributes} which can then be defined as
9848 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9849 add @code{handle_dll_attribute} in the attribute table for your port
9850 to perform initial processing of the @samp{dllimport} and
9851 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9852 @file{i386/i386.c}, for example.
9855 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
9856 @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}.
9859 @defmac TARGET_DECLSPEC
9860 Define this macro to a nonzero value if you want to treat
9861 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9862 default, this behavior is enabled only for targets that define
9863 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9864 of @code{__declspec} is via a built-in macro, but you should not rely
9865 on this implementation detail.
9868 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9869 Define this target hook if you want to be able to add attributes to a decl
9870 when it is being created. This is normally useful for back ends which
9871 wish to implement a pragma by using the attributes which correspond to
9872 the pragma's effect. The @var{node} argument is the decl which is being
9873 created. The @var{attr_ptr} argument is a pointer to the attribute list
9874 for this decl. The list itself should not be modified, since it may be
9875 shared with other decls, but attributes may be chained on the head of
9876 the list and @code{*@var{attr_ptr}} modified to point to the new
9877 attributes, or a copy of the list may be made if further changes are
9881 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
9883 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9884 into the current function, despite its having target-specific
9885 attributes, @code{false} otherwise. By default, if a function has a
9886 target specific attribute attached to it, it will not be inlined.
9889 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9890 This hook is called to parse the @code{attribute(option("..."))}, and
9891 it allows the function to set different target machine compile time
9892 options for the current function that might be different than the
9893 options specified on the command line. The hook should return
9894 @code{true} if the options are valid.
9896 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9897 the function declaration to hold a pointer to a target specific
9898 @var{struct cl_target_option} structure.
9901 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr})
9902 This hook is called to save any additional target specific information
9903 in the @var{struct cl_target_option} structure for function specific
9905 @xref{Option file format}.
9908 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct cl_target_option *@var{ptr})
9909 This hook is called to restore any additional target specific
9910 information in the @var{struct cl_target_option} structure for
9911 function specific options.
9914 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
9915 This hook is called to print any additional target specific
9916 information in the @var{struct cl_target_option} structure for
9917 function specific options.
9920 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
9921 This target hook parses the options for @code{#pragma GCC option} to
9922 set the machine specific options for functions that occur later in the
9923 input stream. The options should be the same as handled by the
9924 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9927 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
9928 Sometimes certain combinations of command options do not make sense on
9929 a particular target machine. You can override the hook
9930 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9931 once just after all the command options have been parsed.
9933 Don't use this hook to turn on various extra optimizations for
9934 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9936 If you need to do something whenever the optimization level is
9937 changed via the optimize attribute or pragma, see
9938 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9941 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
9942 This target hook returns @code{false} if the @var{caller} function
9943 cannot inline @var{callee}, based on target specific information. By
9944 default, inlining is not allowed if the callee function has function
9945 specific target options and the caller does not use the same options.
9949 @section Emulating TLS
9950 @cindex Emulated TLS
9952 For targets whose psABI does not provide Thread Local Storage via
9953 specific relocations and instruction sequences, an emulation layer is
9954 used. A set of target hooks allows this emulation layer to be
9955 configured for the requirements of a particular target. For instance
9956 the psABI may in fact specify TLS support in terms of an emulation
9959 The emulation layer works by creating a control object for every TLS
9960 object. To access the TLS object, a lookup function is provided
9961 which, when given the address of the control object, will return the
9962 address of the current thread's instance of the TLS object.
9964 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9965 Contains the name of the helper function that uses a TLS control
9966 object to locate a TLS instance. The default causes libgcc's
9967 emulated TLS helper function to be used.
9970 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9971 Contains the name of the helper function that should be used at
9972 program startup to register TLS objects that are implicitly
9973 initialized to zero. If this is @code{NULL}, all TLS objects will
9974 have explicit initializers. The default causes libgcc's emulated TLS
9975 registration function to be used.
9978 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9979 Contains the name of the section in which TLS control variables should
9980 be placed. The default of @code{NULL} allows these to be placed in
9984 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9985 Contains the name of the section in which TLS initializers should be
9986 placed. The default of @code{NULL} allows these to be placed in any
9990 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9991 Contains the prefix to be prepended to TLS control variable names.
9992 The default of @code{NULL} uses a target-specific prefix.
9995 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9996 Contains the prefix to be prepended to TLS initializer objects. The
9997 default of @code{NULL} uses a target-specific prefix.
10000 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
10001 Specifies a function that generates the FIELD_DECLs for a TLS control
10002 object type. @var{type} is the RECORD_TYPE the fields are for and
10003 @var{name} should be filled with the structure tag, if the default of
10004 @code{__emutls_object} is unsuitable. The default creates a type suitable
10005 for libgcc's emulated TLS function.
10008 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
10009 Specifies a function that generates the CONSTRUCTOR to initialize a
10010 TLS control object. @var{var} is the TLS control object, @var{decl}
10011 is the TLS object and @var{tmpl_addr} is the address of the
10012 initializer. The default initializes libgcc's emulated TLS control object.
10015 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
10016 Specifies whether the alignment of TLS control variable objects is
10017 fixed and should not be increased as some backends may do to optimize
10018 single objects. The default is false.
10021 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
10022 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
10023 may be used to describe emulated TLS control objects.
10026 @node MIPS Coprocessors
10027 @section Defining coprocessor specifics for MIPS targets.
10028 @cindex MIPS coprocessor-definition macros
10030 The MIPS specification allows MIPS implementations to have as many as 4
10031 coprocessors, each with as many as 32 private registers. GCC supports
10032 accessing these registers and transferring values between the registers
10033 and memory using asm-ized variables. For example:
10036 register unsigned int cp0count asm ("c0r1");
10042 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
10043 names may be added as described below, or the default names may be
10044 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
10046 Coprocessor registers are assumed to be epilogue-used; sets to them will
10047 be preserved even if it does not appear that the register is used again
10048 later in the function.
10050 Another note: according to the MIPS spec, coprocessor 1 (if present) is
10051 the FPU@. One accesses COP1 registers through standard mips
10052 floating-point support; they are not included in this mechanism.
10054 There is one macro used in defining the MIPS coprocessor interface which
10055 you may want to override in subtargets; it is described below.
10057 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
10058 A comma-separated list (with leading comma) of pairs describing the
10059 alternate names of coprocessor registers. The format of each entry should be
10061 @{ @var{alternatename}, @var{register_number}@}
10067 @section Parameters for Precompiled Header Validity Checking
10068 @cindex parameters, precompiled headers
10070 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
10071 This hook returns a pointer to the data needed by
10072 @code{TARGET_PCH_VALID_P} and sets
10073 @samp{*@var{sz}} to the size of the data in bytes.
10076 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
10077 This hook checks whether the options used to create a PCH file are
10078 compatible with the current settings. It returns @code{NULL}
10079 if so and a suitable error message if not. Error messages will
10080 be presented to the user and must be localized using @samp{_(@var{msg})}.
10082 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
10083 when the PCH file was created and @var{sz} is the size of that data in bytes.
10084 It's safe to assume that the data was created by the same version of the
10085 compiler, so no format checking is needed.
10087 The default definition of @code{default_pch_valid_p} should be
10088 suitable for most targets.
10091 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
10092 If this hook is nonnull, the default implementation of
10093 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
10094 of @code{target_flags}. @var{pch_flags} specifies the value that
10095 @code{target_flags} had when the PCH file was created. The return
10096 value is the same as for @code{TARGET_PCH_VALID_P}.
10099 @deftypefn {Target Hook} void TARGET_PREPARE_PCH_SAVE (void)
10100 Called before writing out a PCH file. If the target has some
10101 garbage-collected data that needs to be in a particular state on PCH loads,
10102 it can use this hook to enforce that state. Very few targets need
10103 to do anything here.
10107 @section C++ ABI parameters
10108 @cindex parameters, c++ abi
10110 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10111 Define this hook to override the integer type used for guard variables.
10112 These are used to implement one-time construction of static objects. The
10113 default is long_long_integer_type_node.
10116 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10117 This hook determines how guard variables are used. It should return
10118 @code{false} (the default) if the first byte should be used. A return value of
10119 @code{true} indicates that only the least significant bit should be used.
10122 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10123 This hook returns the size of the cookie to use when allocating an array
10124 whose elements have the indicated @var{type}. Assumes that it is already
10125 known that a cookie is needed. The default is
10126 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10127 IA64/Generic C++ ABI@.
10130 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10131 This hook should return @code{true} if the element size should be stored in
10132 array cookies. The default is to return @code{false}.
10135 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10136 If defined by a backend this hook allows the decision made to export
10137 class @var{type} to be overruled. Upon entry @var{import_export}
10138 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10139 to be imported and 0 otherwise. This function should return the
10140 modified value and perform any other actions necessary to support the
10141 backend's targeted operating system.
10144 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10145 This hook should return @code{true} if constructors and destructors return
10146 the address of the object created/destroyed. The default is to return
10150 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10151 This hook returns true if the key method for a class (i.e., the method
10152 which, if defined in the current translation unit, causes the virtual
10153 table to be emitted) may be an inline function. Under the standard
10154 Itanium C++ ABI the key method may be an inline function so long as
10155 the function is not declared inline in the class definition. Under
10156 some variants of the ABI, an inline function can never be the key
10157 method. The default is to return @code{true}.
10160 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10161 @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}.
10164 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10165 This hook returns true (the default) if virtual tables and other
10166 similar implicit class data objects are always COMDAT if they have
10167 external linkage. If this hook returns false, then class data for
10168 classes whose virtual table will be emitted in only one translation
10169 unit will not be COMDAT.
10172 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10173 This hook returns true (the default) if the RTTI information for
10174 the basic types which is defined in the C++ runtime should always
10175 be COMDAT, false if it should not be COMDAT.
10178 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10179 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10180 should be used to register static destructors when @option{-fuse-cxa-atexit}
10181 is in effect. The default is to return false to use @code{__cxa_atexit}.
10184 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10185 This hook returns true if the target @code{atexit} function can be used
10186 in the same manner as @code{__cxa_atexit} to register C++ static
10187 destructors. This requires that @code{atexit}-registered functions in
10188 shared libraries are run in the correct order when the libraries are
10189 unloaded. The default is to return false.
10192 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10193 @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).
10196 @deftypefn {Target Hook} tree TARGET_CXX_DECL_MANGLING_CONTEXT (const_tree @var{decl})
10197 Return target-specific mangling context of @var{decl} or @code{NULL_TREE}.
10200 @node Named Address Spaces
10201 @section Adding support for named address spaces
10202 @cindex named address spaces
10204 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10205 standards committee, @cite{Programming Languages - C - Extensions to
10206 support embedded processors}, specifies a syntax for embedded
10207 processors to specify alternate address spaces. You can configure a
10208 GCC port to support section 5.1 of the draft report to add support for
10209 address spaces other than the default address space. These address
10210 spaces are new keywords that are similar to the @code{volatile} and
10211 @code{const} type attributes.
10213 Pointers to named address spaces can have a different size than
10214 pointers to the generic address space.
10216 For example, the SPU port uses the @code{__ea} address space to refer
10217 to memory in the host processor, rather than memory local to the SPU
10218 processor. Access to memory in the @code{__ea} address space involves
10219 issuing DMA operations to move data between the host processor and the
10220 local processor memory address space. Pointers in the @code{__ea}
10221 address space are either 32 bits or 64 bits based on the
10222 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10225 Internally, address spaces are represented as a small integer in the
10226 range 0 to 15 with address space 0 being reserved for the generic
10229 To register a named address space qualifier keyword with the C front end,
10230 the target may call the @code{c_register_addr_space} routine. For example,
10231 the SPU port uses the following to declare @code{__ea} as the keyword for
10232 named address space #1:
10234 #define ADDR_SPACE_EA 1
10235 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10238 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10239 Define this to return the machine mode to use for pointers to
10240 @var{address_space} if the target supports named address spaces.
10241 The default version of this hook returns @code{ptr_mode} for the
10242 generic address space only.
10245 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10246 Define this to return the machine mode to use for addresses in
10247 @var{address_space} if the target supports named address spaces.
10248 The default version of this hook returns @code{Pmode} for the
10249 generic address space only.
10252 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (enum machine_mode @var{mode}, addr_space_t @var{as})
10253 Define this to return nonzero if the port can handle pointers
10254 with machine mode @var{mode} to address space @var{as}. This target
10255 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10256 except that it includes explicit named address space support. The default
10257 version of this hook returns true for the modes returned by either the
10258 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10259 target hooks for the given address space.
10262 @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})
10263 Define this to return true if @var{exp} is a valid address for mode
10264 @var{mode} in the named address space @var{as}. The @var{strict}
10265 parameter says whether strict addressing is in effect after reload has
10266 finished. This target hook is the same as the
10267 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10268 explicit named address space support.
10271 @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})
10272 Define this to modify an invalid address @var{x} to be a valid address
10273 with mode @var{mode} in the named address space @var{as}. This target
10274 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10275 except that it includes explicit named address space support.
10278 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{subset}, addr_space_t @var{superset})
10279 Define this to return whether the @var{subset} named address space is
10280 contained within the @var{superset} named address space. Pointers to
10281 a named address space that is a subset of another named address space
10282 will be converted automatically without a cast if used together in
10283 arithmetic operations. Pointers to a superset address space can be
10284 converted to pointers to a subset address space via explicit casts.
10287 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10288 Define this to convert the pointer expression represented by the RTL
10289 @var{op} with type @var{from_type} that points to a named address
10290 space to a new pointer expression with type @var{to_type} that points
10291 to a different named address space. When this hook it called, it is
10292 guaranteed that one of the two address spaces is a subset of the other,
10293 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10297 @section Miscellaneous Parameters
10298 @cindex parameters, miscellaneous
10300 @c prevent bad page break with this line
10301 Here are several miscellaneous parameters.
10303 @defmac HAS_LONG_COND_BRANCH
10304 Define this boolean macro to indicate whether or not your architecture
10305 has conditional branches that can span all of memory. It is used in
10306 conjunction with an optimization that partitions hot and cold basic
10307 blocks into separate sections of the executable. If this macro is
10308 set to false, gcc will convert any conditional branches that attempt
10309 to cross between sections into unconditional branches or indirect jumps.
10312 @defmac HAS_LONG_UNCOND_BRANCH
10313 Define this boolean macro to indicate whether or not your architecture
10314 has unconditional branches that can span all of memory. It is used in
10315 conjunction with an optimization that partitions hot and cold basic
10316 blocks into separate sections of the executable. If this macro is
10317 set to false, gcc will convert any unconditional branches that attempt
10318 to cross between sections into indirect jumps.
10321 @defmac CASE_VECTOR_MODE
10322 An alias for a machine mode name. This is the machine mode that
10323 elements of a jump-table should have.
10326 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10327 Optional: return the preferred mode for an @code{addr_diff_vec}
10328 when the minimum and maximum offset are known. If you define this,
10329 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10330 To make this work, you also have to define @code{INSN_ALIGN} and
10331 make the alignment for @code{addr_diff_vec} explicit.
10332 The @var{body} argument is provided so that the offset_unsigned and scale
10333 flags can be updated.
10336 @defmac CASE_VECTOR_PC_RELATIVE
10337 Define this macro to be a C expression to indicate when jump-tables
10338 should contain relative addresses. You need not define this macro if
10339 jump-tables never contain relative addresses, or jump-tables should
10340 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10344 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10345 This function return the smallest number of different values for which it
10346 is best to use a jump-table instead of a tree of conditional branches.
10347 The default is four for machines with a @code{casesi} instruction and
10348 five otherwise. This is best for most machines.
10351 @defmac CASE_USE_BIT_TESTS
10352 Define this macro to be a C expression to indicate whether C switch
10353 statements may be implemented by a sequence of bit tests. This is
10354 advantageous on processors that can efficiently implement left shift
10355 of 1 by the number of bits held in a register, but inappropriate on
10356 targets that would require a loop. By default, this macro returns
10357 @code{true} if the target defines an @code{ashlsi3} pattern, and
10358 @code{false} otherwise.
10361 @defmac WORD_REGISTER_OPERATIONS
10362 Define this macro if operations between registers with integral mode
10363 smaller than a word are always performed on the entire register.
10364 Most RISC machines have this property and most CISC machines do not.
10367 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10368 Define this macro to be a C expression indicating when insns that read
10369 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10370 bits outside of @var{mem_mode} to be either the sign-extension or the
10371 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10372 of @var{mem_mode} for which the
10373 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10374 @code{UNKNOWN} for other modes.
10376 This macro is not called with @var{mem_mode} non-integral or with a width
10377 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10378 value in this case. Do not define this macro if it would always return
10379 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10380 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10382 You may return a non-@code{UNKNOWN} value even if for some hard registers
10383 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10384 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10385 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10386 integral mode larger than this but not larger than @code{word_mode}.
10388 You must return @code{UNKNOWN} if for some hard registers that allow this
10389 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10390 @code{word_mode}, but that they can change to another integral mode that
10391 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10394 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10395 Define this macro if loading short immediate values into registers sign
10399 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
10400 Define this macro if the same instructions that convert a floating
10401 point number to a signed fixed point number also convert validly to an
10405 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
10406 When @option{-ffast-math} is in effect, GCC tries to optimize
10407 divisions by the same divisor, by turning them into multiplications by
10408 the reciprocal. This target hook specifies the minimum number of divisions
10409 that should be there for GCC to perform the optimization for a variable
10410 of mode @var{mode}. The default implementation returns 3 if the machine
10411 has an instruction for the division, and 2 if it does not.
10415 The maximum number of bytes that a single instruction can move quickly
10416 between memory and registers or between two memory locations.
10419 @defmac MAX_MOVE_MAX
10420 The maximum number of bytes that a single instruction can move quickly
10421 between memory and registers or between two memory locations. If this
10422 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10423 constant value that is the largest value that @code{MOVE_MAX} can have
10427 @defmac SHIFT_COUNT_TRUNCATED
10428 A C expression that is nonzero if on this machine the number of bits
10429 actually used for the count of a shift operation is equal to the number
10430 of bits needed to represent the size of the object being shifted. When
10431 this macro is nonzero, the compiler will assume that it is safe to omit
10432 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10433 truncates the count of a shift operation. On machines that have
10434 instructions that act on bit-fields at variable positions, which may
10435 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10436 also enables deletion of truncations of the values that serve as
10437 arguments to bit-field instructions.
10439 If both types of instructions truncate the count (for shifts) and
10440 position (for bit-field operations), or if no variable-position bit-field
10441 instructions exist, you should define this macro.
10443 However, on some machines, such as the 80386 and the 680x0, truncation
10444 only applies to shift operations and not the (real or pretended)
10445 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10446 such machines. Instead, add patterns to the @file{md} file that include
10447 the implied truncation of the shift instructions.
10449 You need not define this macro if it would always have the value of zero.
10452 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10453 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
10454 This function describes how the standard shift patterns for @var{mode}
10455 deal with shifts by negative amounts or by more than the width of the mode.
10456 @xref{shift patterns}.
10458 On many machines, the shift patterns will apply a mask @var{m} to the
10459 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10460 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10461 this is true for mode @var{mode}, the function should return @var{m},
10462 otherwise it should return 0. A return value of 0 indicates that no
10463 particular behavior is guaranteed.
10465 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10466 @emph{not} apply to general shift rtxes; it applies only to instructions
10467 that are generated by the named shift patterns.
10469 The default implementation of this function returns
10470 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10471 and 0 otherwise. This definition is always safe, but if
10472 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10473 nevertheless truncate the shift count, you may get better code
10477 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10478 A C expression which is nonzero if on this machine it is safe to
10479 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10480 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10481 operating on it as if it had only @var{outprec} bits.
10483 On many machines, this expression can be 1.
10485 @c rearranged this, removed the phrase "it is reported that". this was
10486 @c to fix an overfull hbox. --mew 10feb93
10487 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10488 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10489 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10490 such cases may improve things.
10493 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
10494 The representation of an integral mode can be such that the values
10495 are always extended to a wider integral mode. Return
10496 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10497 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10498 otherwise. (Currently, none of the targets use zero-extended
10499 representation this way so unlike @code{LOAD_EXTEND_OP},
10500 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10501 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10502 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10503 widest integral mode and currently we take advantage of this fact.)
10505 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10506 value even if the extension is not performed on certain hard registers
10507 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10508 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10510 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10511 describe two related properties. If you define
10512 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10513 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10516 In order to enforce the representation of @code{mode},
10517 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10521 @defmac STORE_FLAG_VALUE
10522 A C expression describing the value returned by a comparison operator
10523 with an integral mode and stored by a store-flag instruction
10524 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10525 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10526 comparison operators whose results have a @code{MODE_INT} mode.
10528 A value of 1 or @minus{}1 means that the instruction implementing the
10529 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10530 and 0 when the comparison is false. Otherwise, the value indicates
10531 which bits of the result are guaranteed to be 1 when the comparison is
10532 true. This value is interpreted in the mode of the comparison
10533 operation, which is given by the mode of the first operand in the
10534 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10535 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10538 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10539 generate code that depends only on the specified bits. It can also
10540 replace comparison operators with equivalent operations if they cause
10541 the required bits to be set, even if the remaining bits are undefined.
10542 For example, on a machine whose comparison operators return an
10543 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10544 @samp{0x80000000}, saying that just the sign bit is relevant, the
10548 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10552 can be converted to
10555 (ashift:SI @var{x} (const_int @var{n}))
10559 where @var{n} is the appropriate shift count to move the bit being
10560 tested into the sign bit.
10562 There is no way to describe a machine that always sets the low-order bit
10563 for a true value, but does not guarantee the value of any other bits,
10564 but we do not know of any machine that has such an instruction. If you
10565 are trying to port GCC to such a machine, include an instruction to
10566 perform a logical-and of the result with 1 in the pattern for the
10567 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10569 Often, a machine will have multiple instructions that obtain a value
10570 from a comparison (or the condition codes). Here are rules to guide the
10571 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10576 Use the shortest sequence that yields a valid definition for
10577 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10578 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10579 comparison operators to do so because there may be opportunities to
10580 combine the normalization with other operations.
10583 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10584 slightly preferred on machines with expensive jumps and 1 preferred on
10588 As a second choice, choose a value of @samp{0x80000001} if instructions
10589 exist that set both the sign and low-order bits but do not define the
10593 Otherwise, use a value of @samp{0x80000000}.
10596 Many machines can produce both the value chosen for
10597 @code{STORE_FLAG_VALUE} and its negation in the same number of
10598 instructions. On those machines, you should also define a pattern for
10599 those cases, e.g., one matching
10602 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10605 Some machines can also perform @code{and} or @code{plus} operations on
10606 condition code values with less instructions than the corresponding
10607 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10608 machines, define the appropriate patterns. Use the names @code{incscc}
10609 and @code{decscc}, respectively, for the patterns which perform
10610 @code{plus} or @code{minus} operations on condition code values. See
10611 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10612 find such instruction sequences on other machines.
10614 If this macro is not defined, the default value, 1, is used. You need
10615 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10616 instructions, or if the value generated by these instructions is 1.
10619 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10620 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10621 returned when comparison operators with floating-point results are true.
10622 Define this macro on machines that have comparison operations that return
10623 floating-point values. If there are no such operations, do not define
10627 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10628 A C expression that gives a rtx representing the nonzero true element
10629 for vector comparisons. The returned rtx should be valid for the inner
10630 mode of @var{mode} which is guaranteed to be a vector mode. Define
10631 this macro on machines that have vector comparison operations that
10632 return a vector result. If there are no such operations, do not define
10633 this macro. Typically, this macro is defined as @code{const1_rtx} or
10634 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10635 the compiler optimizing such vector comparison operations for the
10639 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10640 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10641 A C expression that indicates whether the architecture defines a value
10642 for @code{clz} or @code{ctz} with a zero operand.
10643 A result of @code{0} indicates the value is undefined.
10644 If the value is defined for only the RTL expression, the macro should
10645 evaluate to @code{1}; if the value applies also to the corresponding optab
10646 entry (which is normally the case if it expands directly into
10647 the corresponding RTL), then the macro should evaluate to @code{2}.
10648 In the cases where the value is defined, @var{value} should be set to
10651 If this macro is not defined, the value of @code{clz} or
10652 @code{ctz} at zero is assumed to be undefined.
10654 This macro must be defined if the target's expansion for @code{ffs}
10655 relies on a particular value to get correct results. Otherwise it
10656 is not necessary, though it may be used to optimize some corner cases, and
10657 to provide a default expansion for the @code{ffs} optab.
10659 Note that regardless of this macro the ``definedness'' of @code{clz}
10660 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10661 visible to the user. Thus one may be free to adjust the value at will
10662 to match the target expansion of these operations without fear of
10667 An alias for the machine mode for pointers. On most machines, define
10668 this to be the integer mode corresponding to the width of a hardware
10669 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10670 On some machines you must define this to be one of the partial integer
10671 modes, such as @code{PSImode}.
10673 The width of @code{Pmode} must be at least as large as the value of
10674 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10675 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10679 @defmac FUNCTION_MODE
10680 An alias for the machine mode used for memory references to functions
10681 being called, in @code{call} RTL expressions. On most CISC machines,
10682 where an instruction can begin at any byte address, this should be
10683 @code{QImode}. On most RISC machines, where all instructions have fixed
10684 size and alignment, this should be a mode with the same size and alignment
10685 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10688 @defmac STDC_0_IN_SYSTEM_HEADERS
10689 In normal operation, the preprocessor expands @code{__STDC__} to the
10690 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10691 hosts, like Solaris, the system compiler uses a different convention,
10692 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10693 strict conformance to the C Standard.
10695 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10696 convention when processing system header files, but when processing user
10697 files @code{__STDC__} will always expand to 1.
10700 @defmac NO_IMPLICIT_EXTERN_C
10701 Define this macro if the system header files support C++ as well as C@.
10702 This macro inhibits the usual method of using system header files in
10703 C++, which is to pretend that the file's contents are enclosed in
10704 @samp{extern "C" @{@dots{}@}}.
10709 @defmac REGISTER_TARGET_PRAGMAS ()
10710 Define this macro if you want to implement any target-specific pragmas.
10711 If defined, it is a C expression which makes a series of calls to
10712 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10713 for each pragma. The macro may also do any
10714 setup required for the pragmas.
10716 The primary reason to define this macro is to provide compatibility with
10717 other compilers for the same target. In general, we discourage
10718 definition of target-specific pragmas for GCC@.
10720 If the pragma can be implemented by attributes then you should consider
10721 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10723 Preprocessor macros that appear on pragma lines are not expanded. All
10724 @samp{#pragma} directives that do not match any registered pragma are
10725 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10728 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10729 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10731 Each call to @code{c_register_pragma} or
10732 @code{c_register_pragma_with_expansion} establishes one pragma. The
10733 @var{callback} routine will be called when the preprocessor encounters a
10737 #pragma [@var{space}] @var{name} @dots{}
10740 @var{space} is the case-sensitive namespace of the pragma, or
10741 @code{NULL} to put the pragma in the global namespace. The callback
10742 routine receives @var{pfile} as its first argument, which can be passed
10743 on to cpplib's functions if necessary. You can lex tokens after the
10744 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10745 callback will be silently ignored. The end of the line is indicated by
10746 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10747 arguments of pragmas registered with
10748 @code{c_register_pragma_with_expansion} but not on the arguments of
10749 pragmas registered with @code{c_register_pragma}.
10751 Note that the use of @code{pragma_lex} is specific to the C and C++
10752 compilers. It will not work in the Java or Fortran compilers, or any
10753 other language compilers for that matter. Thus if @code{pragma_lex} is going
10754 to be called from target-specific code, it must only be done so when
10755 building the C and C++ compilers. This can be done by defining the
10756 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10757 target entry in the @file{config.gcc} file. These variables should name
10758 the target-specific, language-specific object file which contains the
10759 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10760 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10761 how to build this object file.
10764 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10765 Define this macro if macros should be expanded in the
10766 arguments of @samp{#pragma pack}.
10769 @deftypevr {Target Hook} bool TARGET_HANDLE_PRAGMA_EXTERN_PREFIX
10770 True if @code{#pragma extern_prefix} is to be supported.
10773 @defmac TARGET_DEFAULT_PACK_STRUCT
10774 If your target requires a structure packing default other than 0 (meaning
10775 the machine default), define this macro to the necessary value (in bytes).
10776 This must be a value that would also be valid to use with
10777 @samp{#pragma pack()} (that is, a small power of two).
10780 @defmac DOLLARS_IN_IDENTIFIERS
10781 Define this macro to control use of the character @samp{$} in
10782 identifier names for the C family of languages. 0 means @samp{$} is
10783 not allowed by default; 1 means it is allowed. 1 is the default;
10784 there is no need to define this macro in that case.
10787 @defmac NO_DOLLAR_IN_LABEL
10788 Define this macro if the assembler does not accept the character
10789 @samp{$} in label names. By default constructors and destructors in
10790 G++ have @samp{$} in the identifiers. If this macro is defined,
10791 @samp{.} is used instead.
10794 @defmac NO_DOT_IN_LABEL
10795 Define this macro if the assembler does not accept the character
10796 @samp{.} in label names. By default constructors and destructors in G++
10797 have names that use @samp{.}. If this macro is defined, these names
10798 are rewritten to avoid @samp{.}.
10801 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10802 Define this macro as a C expression that is nonzero if it is safe for the
10803 delay slot scheduler to place instructions in the delay slot of @var{insn},
10804 even if they appear to use a resource set or clobbered in @var{insn}.
10805 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10806 every @code{call_insn} has this behavior. On machines where some @code{insn}
10807 or @code{jump_insn} is really a function call and hence has this behavior,
10808 you should define this macro.
10810 You need not define this macro if it would always return zero.
10813 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10814 Define this macro as a C expression that is nonzero if it is safe for the
10815 delay slot scheduler to place instructions in the delay slot of @var{insn},
10816 even if they appear to set or clobber a resource referenced in @var{insn}.
10817 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10818 some @code{insn} or @code{jump_insn} is really a function call and its operands
10819 are registers whose use is actually in the subroutine it calls, you should
10820 define this macro. Doing so allows the delay slot scheduler to move
10821 instructions which copy arguments into the argument registers into the delay
10822 slot of @var{insn}.
10824 You need not define this macro if it would always return zero.
10827 @defmac MULTIPLE_SYMBOL_SPACES
10828 Define this macro as a C expression that is nonzero if, in some cases,
10829 global symbols from one translation unit may not be bound to undefined
10830 symbols in another translation unit without user intervention. For
10831 instance, under Microsoft Windows symbols must be explicitly imported
10832 from shared libraries (DLLs).
10834 You need not define this macro if it would always evaluate to zero.
10837 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10838 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10839 any hard regs the port wishes to automatically clobber for an asm.
10840 It should return the result of the last @code{tree_cons} used to add a
10841 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10842 corresponding parameters to the asm and may be inspected to avoid
10843 clobbering a register that is an input or output of the asm. You can use
10844 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10845 for overlap with regards to asm-declared registers.
10848 @defmac MATH_LIBRARY
10849 Define this macro as a C string constant for the linker argument to link
10850 in the system math library, minus the initial @samp{"-l"}, or
10851 @samp{""} if the target does not have a
10852 separate math library.
10854 You need only define this macro if the default of @samp{"m"} is wrong.
10857 @defmac LIBRARY_PATH_ENV
10858 Define this macro as a C string constant for the environment variable that
10859 specifies where the linker should look for libraries.
10861 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10865 @defmac TARGET_POSIX_IO
10866 Define this macro if the target supports the following POSIX@ file
10867 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10868 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10869 to use file locking when exiting a program, which avoids race conditions
10870 if the program has forked. It will also create directories at run-time
10871 for cross-profiling.
10874 @defmac MAX_CONDITIONAL_EXECUTE
10876 A C expression for the maximum number of instructions to execute via
10877 conditional execution instructions instead of a branch. A value of
10878 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10879 1 if it does use cc0.
10882 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10883 Used if the target needs to perform machine-dependent modifications on the
10884 conditionals used for turning basic blocks into conditionally executed code.
10885 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10886 contains information about the currently processed blocks. @var{true_expr}
10887 and @var{false_expr} are the tests that are used for converting the
10888 then-block and the else-block, respectively. Set either @var{true_expr} or
10889 @var{false_expr} to a null pointer if the tests cannot be converted.
10892 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10893 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10894 if-statements into conditions combined by @code{and} and @code{or} operations.
10895 @var{bb} contains the basic block that contains the test that is currently
10896 being processed and about to be turned into a condition.
10899 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10900 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10901 be converted to conditional execution format. @var{ce_info} points to
10902 a data structure, @code{struct ce_if_block}, which contains information
10903 about the currently processed blocks.
10906 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10907 A C expression to perform any final machine dependent modifications in
10908 converting code to conditional execution. The involved basic blocks
10909 can be found in the @code{struct ce_if_block} structure that is pointed
10910 to by @var{ce_info}.
10913 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10914 A C expression to cancel any machine dependent modifications in
10915 converting code to conditional execution. The involved basic blocks
10916 can be found in the @code{struct ce_if_block} structure that is pointed
10917 to by @var{ce_info}.
10920 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10921 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10922 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10925 @defmac IFCVT_EXTRA_FIELDS
10926 If defined, it should expand to a set of field declarations that will be
10927 added to the @code{struct ce_if_block} structure. These should be initialized
10928 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10931 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
10932 If non-null, this hook performs a target-specific pass over the
10933 instruction stream. The compiler will run it at all optimization levels,
10934 just before the point at which it normally does delayed-branch scheduling.
10936 The exact purpose of the hook varies from target to target. Some use
10937 it to do transformations that are necessary for correctness, such as
10938 laying out in-function constant pools or avoiding hardware hazards.
10939 Others use it as an opportunity to do some machine-dependent optimizations.
10941 You need not implement the hook if it has nothing to do. The default
10942 definition is null.
10945 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
10946 Define this hook if you have any machine-specific built-in functions
10947 that need to be defined. It should be a function that performs the
10950 Machine specific built-in functions can be useful to expand special machine
10951 instructions that would otherwise not normally be generated because
10952 they have no equivalent in the source language (for example, SIMD vector
10953 instructions or prefetch instructions).
10955 To create a built-in function, call the function
10956 @code{lang_hooks.builtin_function}
10957 which is defined by the language front end. You can use any type nodes set
10958 up by @code{build_common_tree_nodes};
10959 only language front ends that use those two functions will call
10960 @samp{TARGET_INIT_BUILTINS}.
10963 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
10964 Define this hook if you have any machine-specific built-in functions
10965 that need to be defined. It should be a function that returns the
10966 builtin function declaration for the builtin function code @var{code}.
10967 If there is no such builtin and it cannot be initialized at this time
10968 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10969 If @var{code} is out of range the function should return
10970 @code{error_mark_node}.
10973 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10975 Expand a call to a machine specific built-in function that was set up by
10976 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10977 function call; the result should go to @var{target} if that is
10978 convenient, and have mode @var{mode} if that is convenient.
10979 @var{subtarget} may be used as the target for computing one of
10980 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10981 ignored. This function should return the result of the call to the
10985 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
10986 Select a replacement for a machine specific built-in function that
10987 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10988 @emph{before} regular type checking, and so allows the target to
10989 implement a crude form of function overloading. @var{fndecl} is the
10990 declaration of the built-in function. @var{arglist} is the list of
10991 arguments passed to the built-in function. The result is a
10992 complete expression that implements the operation, usually
10993 another @code{CALL_EXPR}.
10994 @var{arglist} really has type @samp{VEC(tree,gc)*}
10997 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
10998 Fold a call to a machine specific built-in function that was set up by
10999 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
11000 built-in function. @var{n_args} is the number of arguments passed to
11001 the function; the arguments themselves are pointed to by @var{argp}.
11002 The result is another tree containing a simplified expression for the
11003 call's result. If @var{ignore} is true the value will be ignored.
11006 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const_rtx @var{insn})
11008 Take an instruction in @var{insn} and return NULL if it is valid within a
11009 low-overhead loop, otherwise return a string explaining why doloop
11010 could not be applied.
11012 Many targets use special registers for low-overhead looping. For any
11013 instruction that clobbers these this function should return a string indicating
11014 the reason why the doloop could not be applied.
11015 By default, the RTL loop optimizer does not use a present doloop pattern for
11016 loops containing function calls or branch on table instructions.
11019 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
11021 Take a branch insn in @var{branch1} and another in @var{branch2}.
11022 Return true if redirecting @var{branch1} to the destination of
11023 @var{branch2} is possible.
11025 On some targets, branches may have a limited range. Optimizing the
11026 filling of delay slots can result in branches being redirected, and this
11027 may in turn cause a branch offset to overflow.
11030 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
11031 This target hook returns @code{true} if @var{x} is considered to be commutative.
11032 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
11033 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
11034 of the enclosing rtl, if known, otherwise it is UNKNOWN.
11037 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
11039 When the initial value of a hard register has been copied in a pseudo
11040 register, it is often not necessary to actually allocate another register
11041 to this pseudo register, because the original hard register or a stack slot
11042 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
11043 is called at the start of register allocation once for each hard register
11044 that had its initial value copied by using
11045 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
11046 Possible values are @code{NULL_RTX}, if you don't want
11047 to do any special allocation, a @code{REG} rtx---that would typically be
11048 the hard register itself, if it is known not to be clobbered---or a
11050 If you are returning a @code{MEM}, this is only a hint for the allocator;
11051 it might decide to use another register anyways.
11052 You may use @code{current_function_leaf_function} in the hook, functions
11053 that use @code{REG_N_SETS}, to determine if the hard
11054 register in question will not be clobbered.
11055 The default value of this hook is @code{NULL}, which disables any special
11059 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
11060 This target hook returns nonzero if @var{x}, an @code{unspec} or
11061 @code{unspec_volatile} operation, might cause a trap. Targets can use
11062 this hook to enhance precision of analysis for @code{unspec} and
11063 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
11064 to analyze inner elements of @var{x} in which case @var{flags} should be
11068 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
11069 The compiler invokes this hook whenever it changes its current function
11070 context (@code{cfun}). You can define this function if
11071 the back end needs to perform any initialization or reset actions on a
11072 per-function basis. For example, it may be used to implement function
11073 attributes that affect register usage or code generation patterns.
11074 The argument @var{decl} is the declaration for the new function context,
11075 and may be null to indicate that the compiler has left a function context
11076 and is returning to processing at the top level.
11077 The default hook function does nothing.
11079 GCC sets @code{cfun} to a dummy function context during initialization of
11080 some parts of the back end. The hook function is not invoked in this
11081 situation; you need not worry about the hook being invoked recursively,
11082 or when the back end is in a partially-initialized state.
11083 @code{cfun} might be @code{NULL} to indicate processing at top level,
11084 outside of any function scope.
11087 @defmac TARGET_OBJECT_SUFFIX
11088 Define this macro to be a C string representing the suffix for object
11089 files on your target machine. If you do not define this macro, GCC will
11090 use @samp{.o} as the suffix for object files.
11093 @defmac TARGET_EXECUTABLE_SUFFIX
11094 Define this macro to be a C string representing the suffix to be
11095 automatically added to executable files on your target machine. If you
11096 do not define this macro, GCC will use the null string as the suffix for
11100 @defmac COLLECT_EXPORT_LIST
11101 If defined, @code{collect2} will scan the individual object files
11102 specified on its command line and create an export list for the linker.
11103 Define this macro for systems like AIX, where the linker discards
11104 object files that are not referenced from @code{main} and uses export
11108 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11109 Define this macro to a C expression representing a variant of the
11110 method call @var{mdecl}, if Java Native Interface (JNI) methods
11111 must be invoked differently from other methods on your target.
11112 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11113 the @code{stdcall} calling convention and this macro is then
11114 defined as this expression:
11117 build_type_attribute_variant (@var{mdecl},
11119 (get_identifier ("stdcall"),
11124 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11125 This target hook returns @code{true} past the point in which new jump
11126 instructions could be created. On machines that require a register for
11127 every jump such as the SHmedia ISA of SH5, this point would typically be
11128 reload, so this target hook should be defined to a function such as:
11132 cannot_modify_jumps_past_reload_p ()
11134 return (reload_completed || reload_in_progress);
11139 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11140 This target hook returns a register class for which branch target register
11141 optimizations should be applied. All registers in this class should be
11142 usable interchangeably. After reload, registers in this class will be
11143 re-allocated and loads will be hoisted out of loops and be subjected
11144 to inter-block scheduling.
11147 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11148 Branch target register optimization will by default exclude callee-saved
11150 that are not already live during the current function; if this target hook
11151 returns true, they will be included. The target code must than make sure
11152 that all target registers in the class returned by
11153 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11154 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11155 epilogues have already been generated. Note, even if you only return
11156 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11157 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11158 to reserve space for caller-saved target registers.
11161 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11162 This target hook returns true if the target supports conditional execution.
11163 This target hook is required only when the target has several different
11164 modes and they have different conditional execution capability, such as ARM.
11167 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop})
11168 This target hook returns a new value for the number of times @var{loop}
11169 should be unrolled. The parameter @var{nunroll} is the number of times
11170 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11171 the loop, which is going to be checked for unrolling. This target hook
11172 is required only when the target has special constraints like maximum
11173 number of memory accesses.
11176 @defmac POWI_MAX_MULTS
11177 If defined, this macro is interpreted as a signed integer C expression
11178 that specifies the maximum number of floating point multiplications
11179 that should be emitted when expanding exponentiation by an integer
11180 constant inline. When this value is defined, exponentiation requiring
11181 more than this number of multiplications is implemented by calling the
11182 system library's @code{pow}, @code{powf} or @code{powl} routines.
11183 The default value places no upper bound on the multiplication count.
11186 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11187 This target hook should register any extra include files for the
11188 target. The parameter @var{stdinc} indicates if normal include files
11189 are present. The parameter @var{sysroot} is the system root directory.
11190 The parameter @var{iprefix} is the prefix for the gcc directory.
11193 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11194 This target hook should register any extra include files for the
11195 target before any standard headers. The parameter @var{stdinc}
11196 indicates if normal include files are present. The parameter
11197 @var{sysroot} is the system root directory. The parameter
11198 @var{iprefix} is the prefix for the gcc directory.
11201 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11202 This target hook should register special include paths for the target.
11203 The parameter @var{path} is the include to register. On Darwin
11204 systems, this is used for Framework includes, which have semantics
11205 that are different from @option{-I}.
11208 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11209 This target macro returns @code{true} if it is safe to use a local alias
11210 for a virtual function @var{fndecl} when constructing thunks,
11211 @code{false} otherwise. By default, the macro returns @code{true} for all
11212 functions, if a target supports aliases (i.e.@: defines
11213 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11216 @defmac TARGET_FORMAT_TYPES
11217 If defined, this macro is the name of a global variable containing
11218 target-specific format checking information for the @option{-Wformat}
11219 option. The default is to have no target-specific format checks.
11222 @defmac TARGET_N_FORMAT_TYPES
11223 If defined, this macro is the number of entries in
11224 @code{TARGET_FORMAT_TYPES}.
11227 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11228 If defined, this macro is the name of a global variable containing
11229 target-specific format overrides for the @option{-Wformat} option. The
11230 default is to have no target-specific format overrides. If defined,
11231 @code{TARGET_FORMAT_TYPES} must be defined, too.
11234 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11235 If defined, this macro specifies the number of entries in
11236 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11239 @defmac TARGET_OVERRIDES_FORMAT_INIT
11240 If defined, this macro specifies the optional initialization
11241 routine for target specific customizations of the system printf
11242 and scanf formatter settings.
11245 @deftypevr {Target Hook} bool TARGET_RELAXED_ORDERING
11246 If set to @code{true}, means that the target's memory model does not
11247 guarantee that loads which do not depend on one another will access
11248 main memory in the order of the instruction stream; if ordering is
11249 important, an explicit memory barrier must be used. This is true of
11250 many recent processors which implement a policy of ``relaxed,''
11251 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11252 and ia64. The default is @code{false}.
11255 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11256 If defined, this macro returns the diagnostic message when it is
11257 illegal to pass argument @var{val} to function @var{funcdecl}
11258 with prototype @var{typelist}.
11261 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11262 If defined, this macro returns the diagnostic message when it is
11263 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11264 if validity should be determined by the front end.
11267 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11268 If defined, this macro returns the diagnostic message when it is
11269 invalid to apply operation @var{op} (where unary plus is denoted by
11270 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11271 if validity should be determined by the front end.
11274 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11275 If defined, this macro returns the diagnostic message when it is
11276 invalid to apply operation @var{op} to operands of types @var{type1}
11277 and @var{type2}, or @code{NULL} if validity should be determined by
11281 @deftypefn {Target Hook} {const char *} TARGET_INVALID_PARAMETER_TYPE (const_tree @var{type})
11282 If defined, this macro returns the diagnostic message when it is
11283 invalid for functions to include parameters of type @var{type},
11284 or @code{NULL} if validity should be determined by
11285 the front end. This is currently used only by the C and C++ front ends.
11288 @deftypefn {Target Hook} {const char *} TARGET_INVALID_RETURN_TYPE (const_tree @var{type})
11289 If defined, this macro returns the diagnostic message when it is
11290 invalid for functions to have return type @var{type},
11291 or @code{NULL} if validity should be determined by
11292 the front end. This is currently used only by the C and C++ front ends.
11295 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11296 If defined, this target hook returns the type to which values of
11297 @var{type} should be promoted when they appear in expressions,
11298 analogous to the integer promotions, or @code{NULL_TREE} to use the
11299 front end's normal promotion rules. This hook is useful when there are
11300 target-specific types with special promotion rules.
11301 This is currently used only by the C and C++ front ends.
11304 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11305 If defined, this hook returns the result of converting @var{expr} to
11306 @var{type}. It should return the converted expression,
11307 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11308 This hook is useful when there are target-specific types with special
11310 This is currently used only by the C and C++ front ends.
11313 @defmac TARGET_USE_JCR_SECTION
11314 This macro determines whether to use the JCR section to register Java
11315 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11316 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11320 This macro determines the size of the objective C jump buffer for the
11321 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11324 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11325 Define this macro if any target-specific attributes need to be attached
11326 to the functions in @file{libgcc} that provide low-level support for
11327 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11328 and the associated definitions of those functions.
11331 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11332 Define this macro to update the current function stack boundary if
11336 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11337 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11338 different argument pointer register is needed to access the function's
11339 argument list due to stack realignment. Return @code{NULL} if no DRAP
11343 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11344 When optimization is disabled, this hook indicates whether or not
11345 arguments should be allocated to stack slots. Normally, GCC allocates
11346 stacks slots for arguments when not optimizing in order to make
11347 debugging easier. However, when a function is declared with
11348 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11349 cannot safely move arguments from the registers in which they are passed
11350 to the stack. Therefore, this hook should return true in general, but
11351 false for naked functions. The default implementation always returns true.
11354 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11355 On some architectures it can take multiple instructions to synthesize
11356 a constant. If there is another constant already in a register that
11357 is close enough in value then it is preferable that the new constant
11358 is computed from this register using immediate addition or
11359 subtraction. We accomplish this through CSE. Besides the value of
11360 the constant we also add a lower and an upper constant anchor to the
11361 available expressions. These are then queried when encountering new
11362 constants. The anchors are computed by rounding the constant up and
11363 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11364 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11365 accepted by immediate-add plus one. We currently assume that the
11366 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11367 MIPS, where add-immediate takes a 16-bit signed value,
11368 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11369 is zero, which disables this optimization. @end deftypevr
11371 @deftypevr {Target Hook} {unsigned char} TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
11372 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}.