1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Target Description Macros and Functions
9 @cindex machine description macros
10 @cindex target description macros
11 @cindex macros, target description
12 @cindex @file{tm.h} macros
14 In addition to the file @file{@var{machine}.md}, a machine description
15 includes a C header file conventionally given the name
16 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
17 The header file defines numerous macros that convey the information
18 about the target machine that does not fit into the scheme of the
19 @file{.md} file. The file @file{tm.h} should be a link to
20 @file{@var{machine}.h}. The header file @file{config.h} includes
21 @file{tm.h} and most compiler source files include @file{config.h}. The
22 source file defines a variable @code{targetm}, which is a structure
23 containing pointers to functions and data relating to the target
24 machine. @file{@var{machine}.c} should also contain their definitions,
25 if they are not defined elsewhere in GCC, and other functions called
26 through the macros defined in the @file{.h} file.
29 * Target Structure:: The @code{targetm} variable.
30 * Driver:: Controlling how the driver runs the compilation passes.
31 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
32 * Per-Function Data:: Defining data structures for per-function information.
33 * Storage Layout:: Defining sizes and alignments of data.
34 * Type Layout:: Defining sizes and properties of basic user data types.
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
37 * Old Constraints:: The old way to define machine-specific constraints.
38 * Stack and Calling:: Defining which way the stack grows and by how much.
39 * Varargs:: Defining the varargs macros.
40 * Trampolines:: Code set up at run time to enter a nested function.
41 * Library Calls:: Controlling how library routines are implicitly called.
42 * Addressing Modes:: Defining addressing modes valid for memory operands.
43 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
44 * Condition Code:: Defining how insns update the condition code.
45 * Costs:: Defining relative costs of different operations.
46 * Scheduling:: Adjusting the behavior of the instruction scheduler.
47 * Sections:: Dividing storage into text, data, and other sections.
48 * PIC:: Macros for position independent code.
49 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
50 * Debugging Info:: Defining the format of debugging output.
51 * Floating Point:: Handling floating point for cross-compilers.
52 * Mode Switching:: Insertion of mode-switching instructions.
53 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
54 * Emulated TLS:: Emulated TLS support.
55 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
56 * PCH Target:: Validity checking for precompiled headers.
57 * C++ ABI:: Controlling C++ ABI changes.
58 * Named Address Spaces:: Adding support for named address spaces
59 * Misc:: Everything else.
62 @node Target Structure
63 @section The Global @code{targetm} Variable
65 @cindex target functions
67 @deftypevar {struct gcc_target} targetm
68 The target @file{.c} file must define the global @code{targetm} variable
69 which contains pointers to functions and data relating to the target
70 machine. The variable is declared in @file{target.h};
71 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
72 used to initialize the variable, and macros for the default initializers
73 for elements of the structure. The @file{.c} file should override those
74 macros for which the default definition is inappropriate. For example:
77 #include "target-def.h"
79 /* @r{Initialize the GCC target structure.} */
81 #undef TARGET_COMP_TYPE_ATTRIBUTES
82 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
84 struct gcc_target targetm = TARGET_INITIALIZER;
88 Where a macro should be defined in the @file{.c} file in this manner to
89 form part of the @code{targetm} structure, it is documented below as a
90 ``Target Hook'' with a prototype. Many macros will change in future
91 from being defined in the @file{.h} file to being part of the
92 @code{targetm} structure.
94 Similarly, there is a @code{targetcm} variable for hooks that are
95 specific to front ends for C-family languages, documented as ``C
96 Target Hook''. This is declared in @file{c-family/c-target.h}, the
97 initializer @code{TARGETCM_INITIALIZER} in
98 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
99 themselves, they should set @code{target_has_targetcm=yes} in
100 @file{config.gcc}; otherwise a default definition is used.
102 Similarly, there is a @code{targetm_common} variable for hooks that
103 are shared between the compiler driver and the compilers proper,
104 documented as ``Common Target Hook''. This is declared in
105 @file{common/common-target.h}, the initializer
106 @code{TARGETM_COMMON_INITIALIZER} in
107 @file{common/common-target-def.h}. If targets initialize
108 @code{targetm_common} themselves, they should set
109 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
110 default definition is used.
113 @section Controlling the Compilation Driver, @file{gcc}
115 @cindex controlling the compilation driver
117 @c prevent bad page break with this line
118 You can control the compilation driver.
120 @defmac DRIVER_SELF_SPECS
121 A list of specs for the driver itself. It should be a suitable
122 initializer for an array of strings, with no surrounding braces.
124 The driver applies these specs to its own command line between loading
125 default @file{specs} files (but not command-line specified ones) and
126 choosing the multilib directory or running any subcommands. It
127 applies them in the order given, so each spec can depend on the
128 options added by earlier ones. It is also possible to remove options
129 using @samp{%<@var{option}} in the usual way.
131 This macro can be useful when a port has several interdependent target
132 options. It provides a way of standardizing the command line so
133 that the other specs are easier to write.
135 Do not define this macro if it does not need to do anything.
138 @defmac OPTION_DEFAULT_SPECS
139 A list of specs used to support configure-time default options (i.e.@:
140 @option{--with} options) in the driver. It should be a suitable initializer
141 for an array of structures, each containing two strings, without the
142 outermost pair of surrounding braces.
144 The first item in the pair is the name of the default. This must match
145 the code in @file{config.gcc} for the target. The second item is a spec
146 to apply if a default with this name was specified. The string
147 @samp{%(VALUE)} in the spec will be replaced by the value of the default
148 everywhere it occurs.
150 The driver will apply these specs to its own command line between loading
151 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
152 the same mechanism as @code{DRIVER_SELF_SPECS}.
154 Do not define this macro if it does not need to do anything.
158 A C string constant that tells the GCC driver program options to
159 pass to CPP@. It can also specify how to translate options you
160 give to GCC into options for GCC to pass to the CPP@.
162 Do not define this macro if it does not need to do anything.
165 @defmac CPLUSPLUS_CPP_SPEC
166 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
167 than C@. If you do not define this macro, then the value of
168 @code{CPP_SPEC} (if any) will be used instead.
172 A C string constant that tells the GCC driver program options to
173 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
175 It can also specify how to translate options you give to GCC into options
176 for GCC to pass to front ends.
178 Do not define this macro if it does not need to do anything.
182 A C string constant that tells the GCC driver program options to
183 pass to @code{cc1plus}. It can also specify how to translate options you
184 give to GCC into options for GCC to pass to the @code{cc1plus}.
186 Do not define this macro if it does not need to do anything.
187 Note that everything defined in CC1_SPEC is already passed to
188 @code{cc1plus} so there is no need to duplicate the contents of
189 CC1_SPEC in CC1PLUS_SPEC@.
193 A C string constant that tells the GCC driver program options to
194 pass to the assembler. It can also specify how to translate options
195 you give to GCC into options for GCC to pass to the assembler.
196 See the file @file{sun3.h} for an example of this.
198 Do not define this macro if it does not need to do anything.
201 @defmac ASM_FINAL_SPEC
202 A C string constant that tells the GCC driver program how to
203 run any programs which cleanup after the normal assembler.
204 Normally, this is not needed. See the file @file{mips.h} for
207 Do not define this macro if it does not need to do anything.
210 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
211 Define this macro, with no value, if the driver should give the assembler
212 an argument consisting of a single dash, @option{-}, to instruct it to
213 read from its standard input (which will be a pipe connected to the
214 output of the compiler proper). This argument is given after any
215 @option{-o} option specifying the name of the output file.
217 If you do not define this macro, the assembler is assumed to read its
218 standard input if given no non-option arguments. If your assembler
219 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
220 see @file{mips.h} for instance.
224 A C string constant that tells the GCC driver program options to
225 pass to the linker. It can also specify how to translate options you
226 give to GCC into options for GCC to pass to the linker.
228 Do not define this macro if it does not need to do anything.
232 Another C string constant used much like @code{LINK_SPEC}. The difference
233 between the two is that @code{LIB_SPEC} is used at the end of the
234 command given to the linker.
236 If this macro is not defined, a default is provided that
237 loads the standard C library from the usual place. See @file{gcc.c}.
241 Another C string constant that tells the GCC driver program
242 how and when to place a reference to @file{libgcc.a} into the
243 linker command line. This constant is placed both before and after
244 the value of @code{LIB_SPEC}.
246 If this macro is not defined, the GCC driver provides a default that
247 passes the string @option{-lgcc} to the linker.
250 @defmac REAL_LIBGCC_SPEC
251 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
252 @code{LIBGCC_SPEC} is not directly used by the driver program but is
253 instead modified to refer to different versions of @file{libgcc.a}
254 depending on the values of the command line flags @option{-static},
255 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
256 targets where these modifications are inappropriate, define
257 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
258 driver how to place a reference to @file{libgcc} on the link command
259 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
262 @defmac USE_LD_AS_NEEDED
263 A macro that controls the modifications to @code{LIBGCC_SPEC}
264 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
265 generated that uses --as-needed and the shared libgcc in place of the
266 static exception handler library, when linking without any of
267 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
271 If defined, this C string constant is added to @code{LINK_SPEC}.
272 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
273 the modifications to @code{LIBGCC_SPEC} mentioned in
274 @code{REAL_LIBGCC_SPEC}.
277 @defmac STARTFILE_SPEC
278 Another C string constant used much like @code{LINK_SPEC}. The
279 difference between the two is that @code{STARTFILE_SPEC} is used at
280 the very beginning of the command given to the linker.
282 If this macro is not defined, a default is provided that loads the
283 standard C startup file from the usual place. See @file{gcc.c}.
287 Another C string constant used much like @code{LINK_SPEC}. The
288 difference between the two is that @code{ENDFILE_SPEC} is used at
289 the very end of the command given to the linker.
291 Do not define this macro if it does not need to do anything.
294 @defmac THREAD_MODEL_SPEC
295 GCC @code{-v} will print the thread model GCC was configured to use.
296 However, this doesn't work on platforms that are multilibbed on thread
297 models, such as AIX 4.3. On such platforms, define
298 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
299 blanks that names one of the recognized thread models. @code{%*}, the
300 default value of this macro, will expand to the value of
301 @code{thread_file} set in @file{config.gcc}.
304 @defmac SYSROOT_SUFFIX_SPEC
305 Define this macro to add a suffix to the target sysroot when GCC is
306 configured with a sysroot. This will cause GCC to search for usr/lib,
307 et al, within sysroot+suffix.
310 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
311 Define this macro to add a headers_suffix to the target sysroot when
312 GCC is configured with a sysroot. This will cause GCC to pass the
313 updated sysroot+headers_suffix to CPP, causing it to search for
314 usr/include, et al, within sysroot+headers_suffix.
318 Define this macro to provide additional specifications to put in the
319 @file{specs} file that can be used in various specifications like
322 The definition should be an initializer for an array of structures,
323 containing a string constant, that defines the specification name, and a
324 string constant that provides the specification.
326 Do not define this macro if it does not need to do anything.
328 @code{EXTRA_SPECS} is useful when an architecture contains several
329 related targets, which have various @code{@dots{}_SPECS} which are similar
330 to each other, and the maintainer would like one central place to keep
333 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
334 define either @code{_CALL_SYSV} when the System V calling sequence is
335 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
338 The @file{config/rs6000/rs6000.h} target file defines:
341 #define EXTRA_SPECS \
342 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
344 #define CPP_SYS_DEFAULT ""
347 The @file{config/rs6000/sysv.h} target file defines:
351 "%@{posix: -D_POSIX_SOURCE @} \
352 %@{mcall-sysv: -D_CALL_SYSV @} \
353 %@{!mcall-sysv: %(cpp_sysv_default) @} \
354 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
356 #undef CPP_SYSV_DEFAULT
357 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
360 while the @file{config/rs6000/eabiaix.h} target file defines
361 @code{CPP_SYSV_DEFAULT} as:
364 #undef CPP_SYSV_DEFAULT
365 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
369 @defmac LINK_LIBGCC_SPECIAL_1
370 Define this macro if the driver program should find the library
371 @file{libgcc.a}. If you do not define this macro, the driver program will pass
372 the argument @option{-lgcc} to tell the linker to do the search.
375 @defmac LINK_GCC_C_SEQUENCE_SPEC
376 The sequence in which libgcc and libc are specified to the linker.
377 By default this is @code{%G %L %G}.
380 @defmac LINK_COMMAND_SPEC
381 A C string constant giving the complete command line need to execute the
382 linker. When you do this, you will need to update your port each time a
383 change is made to the link command line within @file{gcc.c}. Therefore,
384 define this macro only if you need to completely redefine the command
385 line for invoking the linker and there is no other way to accomplish
386 the effect you need. Overriding this macro may be avoidable by overriding
387 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
390 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
391 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
392 directories from linking commands. Do not give it a nonzero value if
393 removing duplicate search directories changes the linker's semantics.
396 @deftypevr {Common Target Hook} bool TARGET_ALWAYS_STRIP_DOTDOT
397 True if @file{..} components should always be removed from directory names computed relative to GCC's internal directories, false (default) if such components should be preserved and directory names containing them passed to other tools such as the linker.
400 @defmac MULTILIB_DEFAULTS
401 Define this macro as a C expression for the initializer of an array of
402 string to tell the driver program which options are defaults for this
403 target and thus do not need to be handled specially when using
404 @code{MULTILIB_OPTIONS}.
406 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
407 the target makefile fragment or if none of the options listed in
408 @code{MULTILIB_OPTIONS} are set by default.
409 @xref{Target Fragment}.
412 @defmac RELATIVE_PREFIX_NOT_LINKDIR
413 Define this macro to tell @command{gcc} that it should only translate
414 a @option{-B} prefix into a @option{-L} linker option if the prefix
415 indicates an absolute file name.
418 @defmac MD_EXEC_PREFIX
419 If defined, this macro is an additional prefix to try after
420 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
421 when the compiler is built as a cross
422 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
423 to the list of directories used to find the assembler in @file{configure.in}.
426 @defmac STANDARD_STARTFILE_PREFIX
427 Define this macro as a C string constant if you wish to override the
428 standard choice of @code{libdir} as the default prefix to
429 try when searching for startup files such as @file{crt0.o}.
430 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
431 is built as a cross compiler.
434 @defmac STANDARD_STARTFILE_PREFIX_1
435 Define this macro as a C string constant if you wish to override the
436 standard choice of @code{/lib} as a prefix to try after the default prefix
437 when searching for startup files such as @file{crt0.o}.
438 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
439 is built as a cross compiler.
442 @defmac STANDARD_STARTFILE_PREFIX_2
443 Define this macro as a C string constant if you wish to override the
444 standard choice of @code{/lib} as yet another prefix to try after the
445 default prefix when searching for startup files such as @file{crt0.o}.
446 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
447 is built as a cross compiler.
450 @defmac MD_STARTFILE_PREFIX
451 If defined, this macro supplies an additional prefix to try after the
452 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
453 compiler is built as a cross compiler.
456 @defmac MD_STARTFILE_PREFIX_1
457 If defined, this macro supplies yet another prefix to try after the
458 standard prefixes. It is not searched when the compiler is built as a
462 @defmac INIT_ENVIRONMENT
463 Define this macro as a C string constant if you wish to set environment
464 variables for programs called by the driver, such as the assembler and
465 loader. The driver passes the value of this macro to @code{putenv} to
466 initialize the necessary environment variables.
469 @defmac LOCAL_INCLUDE_DIR
470 Define this macro as a C string constant if you wish to override the
471 standard choice of @file{/usr/local/include} as the default prefix to
472 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
473 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
475 Cross compilers do not search either @file{/usr/local/include} or its
479 @defmac SYSTEM_INCLUDE_DIR
480 Define this macro as a C string constant if you wish to specify a
481 system-specific directory to search for header files before the standard
482 directory. @code{SYSTEM_INCLUDE_DIR} comes before
483 @code{STANDARD_INCLUDE_DIR} in the search order.
485 Cross compilers do not use this macro and do not search the directory
489 @defmac STANDARD_INCLUDE_DIR
490 Define this macro as a C string constant if you wish to override the
491 standard choice of @file{/usr/include} as the default prefix to
492 try when searching for header files.
494 Cross compilers ignore this macro and do not search either
495 @file{/usr/include} or its replacement.
498 @defmac STANDARD_INCLUDE_COMPONENT
499 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
500 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
501 If you do not define this macro, no component is used.
504 @defmac INCLUDE_DEFAULTS
505 Define this macro if you wish to override the entire default search path
506 for include files. For a native compiler, the default search path
507 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
508 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
509 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
510 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
511 and specify private search areas for GCC@. The directory
512 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
514 The definition should be an initializer for an array of structures.
515 Each array element should have four elements: the directory name (a
516 string constant), the component name (also a string constant), a flag
517 for C++-only directories,
518 and a flag showing that the includes in the directory don't need to be
519 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
520 the array with a null element.
522 The component name denotes what GNU package the include file is part of,
523 if any, in all uppercase letters. For example, it might be @samp{GCC}
524 or @samp{BINUTILS}. If the package is part of a vendor-supplied
525 operating system, code the component name as @samp{0}.
527 For example, here is the definition used for VAX/VMS:
530 #define INCLUDE_DEFAULTS \
532 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
533 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
534 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
541 Here is the order of prefixes tried for exec files:
545 Any prefixes specified by the user with @option{-B}.
548 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
549 is not set and the compiler has not been installed in the configure-time
550 @var{prefix}, the location in which the compiler has actually been installed.
553 The directories specified by the environment variable @code{COMPILER_PATH}.
556 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
557 in the configured-time @var{prefix}.
560 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
563 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
566 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
570 Here is the order of prefixes tried for startfiles:
574 Any prefixes specified by the user with @option{-B}.
577 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
578 value based on the installed toolchain location.
581 The directories specified by the environment variable @code{LIBRARY_PATH}
582 (or port-specific name; native only, cross compilers do not use this).
585 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
586 in the configured @var{prefix} or this is a native compiler.
589 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
592 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
596 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
597 native compiler, or we have a target system root.
600 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
601 native compiler, or we have a target system root.
604 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
605 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
606 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
609 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
610 compiler, or we have a target system root. The default for this macro is
614 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
615 compiler, or we have a target system root. The default for this macro is
619 @node Run-time Target
620 @section Run-time Target Specification
621 @cindex run-time target specification
622 @cindex predefined macros
623 @cindex target specifications
625 @c prevent bad page break with this line
626 Here are run-time target specifications.
628 @defmac TARGET_CPU_CPP_BUILTINS ()
629 This function-like macro expands to a block of code that defines
630 built-in preprocessor macros and assertions for the target CPU, using
631 the functions @code{builtin_define}, @code{builtin_define_std} and
632 @code{builtin_assert}. When the front end
633 calls this macro it provides a trailing semicolon, and since it has
634 finished command line option processing your code can use those
637 @code{builtin_assert} takes a string in the form you pass to the
638 command-line option @option{-A}, such as @code{cpu=mips}, and creates
639 the assertion. @code{builtin_define} takes a string in the form
640 accepted by option @option{-D} and unconditionally defines the macro.
642 @code{builtin_define_std} takes a string representing the name of an
643 object-like macro. If it doesn't lie in the user's namespace,
644 @code{builtin_define_std} defines it unconditionally. Otherwise, it
645 defines a version with two leading underscores, and another version
646 with two leading and trailing underscores, and defines the original
647 only if an ISO standard was not requested on the command line. For
648 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
649 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
650 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
651 defines only @code{_ABI64}.
653 You can also test for the C dialect being compiled. The variable
654 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
655 or @code{clk_objective_c}. Note that if we are preprocessing
656 assembler, this variable will be @code{clk_c} but the function-like
657 macro @code{preprocessing_asm_p()} will return true, so you might want
658 to check for that first. If you need to check for strict ANSI, the
659 variable @code{flag_iso} can be used. The function-like macro
660 @code{preprocessing_trad_p()} can be used to check for traditional
664 @defmac TARGET_OS_CPP_BUILTINS ()
665 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
666 and is used for the target operating system instead.
669 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
670 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
671 and is used for the target object format. @file{elfos.h} uses this
672 macro to define @code{__ELF__}, so you probably do not need to define
676 @deftypevar {extern int} target_flags
677 This variable is declared in @file{options.h}, which is included before
678 any target-specific headers.
681 @deftypevr {Common Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
682 This variable specifies the initial value of @code{target_flags}.
683 Its default setting is 0.
686 @cindex optional hardware or system features
687 @cindex features, optional, in system conventions
689 @deftypefn {Common Target Hook} bool TARGET_HANDLE_OPTION (struct gcc_options *@var{opts}, struct gcc_options *@var{opts_set}, const struct cl_decoded_option *@var{decoded}, location_t @var{loc})
690 This hook is called whenever the user specifies one of the
691 target-specific options described by the @file{.opt} definition files
692 (@pxref{Options}). It has the opportunity to do some option-specific
693 processing and should return true if the option is valid. The default
694 definition does nothing but return true.
696 @var{decoded} specifies the option and its arguments. @var{opts} and
697 @var{opts_set} are the @code{gcc_options} structures to be used for
698 storing option state, and @var{loc} is the location at which the
699 option was passed (@code{UNKNOWN_LOCATION} except for options passed
703 @deftypefn {C Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
704 This target hook is called whenever the user specifies one of the
705 target-specific C language family options described by the @file{.opt}
706 definition files(@pxref{Options}). It has the opportunity to do some
707 option-specific processing and should return true if the option is
708 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
709 default definition does nothing but return false.
711 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
712 options. However, if processing an option requires routines that are
713 only available in the C (and related language) front ends, then you
714 should use @code{TARGET_HANDLE_C_OPTION} instead.
717 @deftypefn {C Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING_OBJECT (tree @var{string})
718 Targets may provide a string object type that can be used within and between C, C++ and their respective Objective-C dialects. A string object might, for example, embed encoding and length information. These objects are considered opaque to the compiler and handled as references. An ideal implementation makes the composition of the string object match that of the Objective-C @code{NSString} (@code{NXString} for GNUStep), allowing efficient interworking between C-only and Objective-C code. If a target implements string objects then this hook should return a reference to such an object constructed from the normal `C' string representation provided in @var{string}. At present, the hook is used by Objective-C only, to obtain a common-format string object when the target provides one.
721 @deftypefn {C Target Hook} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref})
722 If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object.
725 @deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list})
726 If a target implements string objects then this hook should should provide a facility to check the function arguments in @var{args_list} against the format specifiers in @var{format_arg} where the type of @var{format_arg} is one recognized as a valid string reference type.
729 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
730 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
731 but is called when the optimize level is changed via an attribute or
732 pragma or when it is reset at the end of the code affected by the
733 attribute or pragma. It is not called at the beginning of compilation
734 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
735 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
736 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
739 @defmac C_COMMON_OVERRIDE_OPTIONS
740 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
741 but is only used in the C
742 language frontends (C, Objective-C, C++, Objective-C++) and so can be
743 used to alter option flag variables which only exist in those
747 @deftypevr {Common Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
748 Some machines may desire to change what optimizations are performed for
749 various optimization levels. This variable, if defined, describes
750 options to enable at particular sets of optimization levels. These
751 options are processed once
752 just after the optimization level is determined and before the remainder
753 of the command options have been parsed, so may be overridden by other
754 options passed explicitly.
756 This processing is run once at program startup and when the optimization
757 options are changed via @code{#pragma GCC optimize} or by using the
758 @code{optimize} attribute.
761 @deftypefn {Common Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
762 Set target-dependent initial values of fields in @var{opts}.
765 @deftypefn {Common Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
766 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
769 @defmac SWITCHABLE_TARGET
770 Some targets need to switch between substantially different subtargets
771 during compilation. For example, the MIPS target has one subtarget for
772 the traditional MIPS architecture and another for MIPS16. Source code
773 can switch between these two subarchitectures using the @code{mips16}
774 and @code{nomips16} attributes.
776 Such subtargets can differ in things like the set of available
777 registers, the set of available instructions, the costs of various
778 operations, and so on. GCC caches a lot of this type of information
779 in global variables, and recomputing them for each subtarget takes a
780 significant amount of time. The compiler therefore provides a facility
781 for maintaining several versions of the global variables and quickly
782 switching between them; see @file{target-globals.h} for details.
784 Define this macro to 1 if your target needs this facility. The default
788 @node Per-Function Data
789 @section Defining data structures for per-function information.
790 @cindex per-function data
791 @cindex data structures
793 If the target needs to store information on a per-function basis, GCC
794 provides a macro and a couple of variables to allow this. Note, just
795 using statics to store the information is a bad idea, since GCC supports
796 nested functions, so you can be halfway through encoding one function
797 when another one comes along.
799 GCC defines a data structure called @code{struct function} which
800 contains all of the data specific to an individual function. This
801 structure contains a field called @code{machine} whose type is
802 @code{struct machine_function *}, which can be used by targets to point
803 to their own specific data.
805 If a target needs per-function specific data it should define the type
806 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
807 This macro should be used to initialize the function pointer
808 @code{init_machine_status}. This pointer is explained below.
810 One typical use of per-function, target specific data is to create an
811 RTX to hold the register containing the function's return address. This
812 RTX can then be used to implement the @code{__builtin_return_address}
813 function, for level 0.
815 Note---earlier implementations of GCC used a single data area to hold
816 all of the per-function information. Thus when processing of a nested
817 function began the old per-function data had to be pushed onto a
818 stack, and when the processing was finished, it had to be popped off the
819 stack. GCC used to provide function pointers called
820 @code{save_machine_status} and @code{restore_machine_status} to handle
821 the saving and restoring of the target specific information. Since the
822 single data area approach is no longer used, these pointers are no
825 @defmac INIT_EXPANDERS
826 Macro called to initialize any target specific information. This macro
827 is called once per function, before generation of any RTL has begun.
828 The intention of this macro is to allow the initialization of the
829 function pointer @code{init_machine_status}.
832 @deftypevar {void (*)(struct function *)} init_machine_status
833 If this function pointer is non-@code{NULL} it will be called once per
834 function, before function compilation starts, in order to allow the
835 target to perform any target specific initialization of the
836 @code{struct function} structure. It is intended that this would be
837 used to initialize the @code{machine} of that structure.
839 @code{struct machine_function} structures are expected to be freed by GC@.
840 Generally, any memory that they reference must be allocated by using
841 GC allocation, including the structure itself.
845 @section Storage Layout
846 @cindex storage layout
848 Note that the definitions of the macros in this table which are sizes or
849 alignments measured in bits do not need to be constant. They can be C
850 expressions that refer to static variables, such as the @code{target_flags}.
851 @xref{Run-time Target}.
853 @defmac BITS_BIG_ENDIAN
854 Define this macro to have the value 1 if the most significant bit in a
855 byte has the lowest number; otherwise define it to have the value zero.
856 This means that bit-field instructions count from the most significant
857 bit. If the machine has no bit-field instructions, then this must still
858 be defined, but it doesn't matter which value it is defined to. This
859 macro need not be a constant.
861 This macro does not affect the way structure fields are packed into
862 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
865 @defmac BYTES_BIG_ENDIAN
866 Define this macro to have the value 1 if the most significant byte in a
867 word has the lowest number. This macro need not be a constant.
870 @defmac WORDS_BIG_ENDIAN
871 Define this macro to have the value 1 if, in a multiword object, the
872 most significant word has the lowest number. This applies to both
873 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
874 order of words in memory is not the same as the order in registers. This
875 macro need not be a constant.
878 @defmac REG_WORDS_BIG_ENDIAN
879 On some machines, the order of words in a multiword object differs between
880 registers in memory. In such a situation, define this macro to describe
881 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
882 the order of words in memory.
885 @defmac FLOAT_WORDS_BIG_ENDIAN
886 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
887 @code{TFmode} floating point numbers are stored in memory with the word
888 containing the sign bit at the lowest address; otherwise define it to
889 have the value 0. This macro need not be a constant.
891 You need not define this macro if the ordering is the same as for
895 @defmac BITS_PER_UNIT
896 Define this macro to be the number of bits in an addressable storage
897 unit (byte). If you do not define this macro the default is 8.
900 @defmac BITS_PER_WORD
901 Number of bits in a word. If you do not define this macro, the default
902 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
905 @defmac MAX_BITS_PER_WORD
906 Maximum number of bits in a word. If this is undefined, the default is
907 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
908 largest value that @code{BITS_PER_WORD} can have at run-time.
911 @defmac UNITS_PER_WORD
912 Number of storage units in a word; normally the size of a general-purpose
913 register, a power of two from 1 or 8.
916 @defmac MIN_UNITS_PER_WORD
917 Minimum number of units in a word. If this is undefined, the default is
918 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
919 smallest value that @code{UNITS_PER_WORD} can have at run-time.
923 Width of a pointer, in bits. You must specify a value no wider than the
924 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
925 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
926 a value the default is @code{BITS_PER_WORD}.
929 @defmac POINTERS_EXTEND_UNSIGNED
930 A C expression that determines how pointers should be extended from
931 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
932 greater than zero if pointers should be zero-extended, zero if they
933 should be sign-extended, and negative if some other sort of conversion
934 is needed. In the last case, the extension is done by the target's
935 @code{ptr_extend} instruction.
937 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
938 and @code{word_mode} are all the same width.
941 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
942 A macro to update @var{m} and @var{unsignedp} when an object whose type
943 is @var{type} and which has the specified mode and signedness is to be
944 stored in a register. This macro is only called when @var{type} is a
947 On most RISC machines, which only have operations that operate on a full
948 register, define this macro to set @var{m} to @code{word_mode} if
949 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
950 cases, only integer modes should be widened because wider-precision
951 floating-point operations are usually more expensive than their narrower
954 For most machines, the macro definition does not change @var{unsignedp}.
955 However, some machines, have instructions that preferentially handle
956 either signed or unsigned quantities of certain modes. For example, on
957 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
958 sign-extend the result to 64 bits. On such machines, set
959 @var{unsignedp} according to which kind of extension is more efficient.
961 Do not define this macro if it would never modify @var{m}.
964 @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})
965 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
966 function return values. The target hook should return the new mode
967 and possibly change @code{*@var{punsignedp}} if the promotion should
968 change signedness. This function is called only for scalar @emph{or
971 @var{for_return} allows to distinguish the promotion of arguments and
972 return values. If it is @code{1}, a return value is being promoted and
973 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
974 If it is @code{2}, the returned mode should be that of the register in
975 which an incoming parameter is copied, or the outgoing result is computed;
976 then the hook should return the same mode as @code{promote_mode}, though
977 the signedness may be different.
979 @var{type} can be NULL when promoting function arguments of libcalls.
981 The default is to not promote arguments and return values. You can
982 also define the hook to @code{default_promote_function_mode_always_promote}
983 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
986 @defmac PARM_BOUNDARY
987 Normal alignment required for function parameters on the stack, in
988 bits. All stack parameters receive at least this much alignment
989 regardless of data type. On most machines, this is the same as the
993 @defmac STACK_BOUNDARY
994 Define this macro to the minimum alignment enforced by hardware for the
995 stack pointer on this machine. The definition is a C expression for the
996 desired alignment (measured in bits). This value is used as a default
997 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
998 this should be the same as @code{PARM_BOUNDARY}.
1001 @defmac PREFERRED_STACK_BOUNDARY
1002 Define this macro if you wish to preserve a certain alignment for the
1003 stack pointer, greater than what the hardware enforces. The definition
1004 is a C expression for the desired alignment (measured in bits). This
1005 macro must evaluate to a value equal to or larger than
1006 @code{STACK_BOUNDARY}.
1009 @defmac INCOMING_STACK_BOUNDARY
1010 Define this macro if the incoming stack boundary may be different
1011 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
1012 to a value equal to or larger than @code{STACK_BOUNDARY}.
1015 @defmac FUNCTION_BOUNDARY
1016 Alignment required for a function entry point, in bits.
1019 @defmac BIGGEST_ALIGNMENT
1020 Biggest alignment that any data type can require on this machine, in
1021 bits. Note that this is not the biggest alignment that is supported,
1022 just the biggest alignment that, when violated, may cause a fault.
1025 @defmac MALLOC_ABI_ALIGNMENT
1026 Alignment, in bits, a C conformant malloc implementation has to
1027 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1030 @defmac ATTRIBUTE_ALIGNED_VALUE
1031 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1032 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1035 @defmac MINIMUM_ATOMIC_ALIGNMENT
1036 If defined, the smallest alignment, in bits, that can be given to an
1037 object that can be referenced in one operation, without disturbing any
1038 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1039 on machines that don't have byte or half-word store operations.
1042 @defmac BIGGEST_FIELD_ALIGNMENT
1043 Biggest alignment that any structure or union field can require on this
1044 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1045 structure and union fields only, unless the field alignment has been set
1046 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1049 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1050 An expression for the alignment of a structure field @var{field} if the
1051 alignment computed in the usual way (including applying of
1052 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1053 alignment) is @var{computed}. It overrides alignment only if the
1054 field alignment has not been set by the
1055 @code{__attribute__ ((aligned (@var{n})))} construct.
1058 @defmac MAX_STACK_ALIGNMENT
1059 Biggest stack alignment guaranteed by the backend. Use this macro
1060 to specify the maximum alignment of a variable on stack.
1062 If not defined, the default value is @code{STACK_BOUNDARY}.
1064 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1065 @c But the fix for PR 32893 indicates that we can only guarantee
1066 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1067 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1070 @defmac MAX_OFILE_ALIGNMENT
1071 Biggest alignment supported by the object file format of this machine.
1072 Use this macro to limit the alignment which can be specified using the
1073 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1074 the default value is @code{BIGGEST_ALIGNMENT}.
1076 On systems that use ELF, the default (in @file{config/elfos.h}) is
1077 the largest supported 32-bit ELF section alignment representable on
1078 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1079 On 32-bit ELF the largest supported section alignment in bits is
1080 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1083 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1084 If defined, a C expression to compute the alignment for a variable in
1085 the static store. @var{type} is the data type, and @var{basic-align} is
1086 the alignment that the object would ordinarily have. The value of this
1087 macro is used instead of that alignment to align the object.
1089 If this macro is not defined, then @var{basic-align} is used.
1092 One use of this macro is to increase alignment of medium-size data to
1093 make it all fit in fewer cache lines. Another is to cause character
1094 arrays to be word-aligned so that @code{strcpy} calls that copy
1095 constants to character arrays can be done inline.
1098 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1099 If defined, a C expression to compute the alignment given to a constant
1100 that is being placed in memory. @var{constant} is the constant and
1101 @var{basic-align} is the alignment that the object would ordinarily
1102 have. The value of this macro is used instead of that alignment to
1105 If this macro is not defined, then @var{basic-align} is used.
1107 The typical use of this macro is to increase alignment for string
1108 constants to be word aligned so that @code{strcpy} calls that copy
1109 constants can be done inline.
1112 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1113 If defined, a C expression to compute the alignment for a variable in
1114 the local store. @var{type} is the data type, and @var{basic-align} is
1115 the alignment that the object would ordinarily have. The value of this
1116 macro is used instead of that alignment to align the object.
1118 If this macro is not defined, then @var{basic-align} is used.
1120 One use of this macro is to increase alignment of medium-size data to
1121 make it all fit in fewer cache lines.
1123 If the value of this macro has a type, it should be an unsigned type.
1126 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1127 If defined, a C expression to compute the alignment for stack slot.
1128 @var{type} is the data type, @var{mode} is the widest mode available,
1129 and @var{basic-align} is the alignment that the slot would ordinarily
1130 have. The value of this macro is used instead of that alignment to
1133 If this macro is not defined, then @var{basic-align} is used when
1134 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1137 This macro is to set alignment of stack slot to the maximum alignment
1138 of all possible modes which the slot may have.
1140 If the value of this macro has a type, it should be an unsigned type.
1143 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1144 If defined, a C expression to compute the alignment for a local
1145 variable @var{decl}.
1147 If this macro is not defined, then
1148 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1151 One use of this macro is to increase alignment of medium-size data to
1152 make it all fit in fewer cache lines.
1154 If the value of this macro has a type, it should be an unsigned type.
1157 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1158 If defined, a C expression to compute the minimum required alignment
1159 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1160 @var{mode}, assuming normal alignment @var{align}.
1162 If this macro is not defined, then @var{align} will be used.
1165 @defmac EMPTY_FIELD_BOUNDARY
1166 Alignment in bits to be given to a structure bit-field that follows an
1167 empty field such as @code{int : 0;}.
1169 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1172 @defmac STRUCTURE_SIZE_BOUNDARY
1173 Number of bits which any structure or union's size must be a multiple of.
1174 Each structure or union's size is rounded up to a multiple of this.
1176 If you do not define this macro, the default is the same as
1177 @code{BITS_PER_UNIT}.
1180 @defmac STRICT_ALIGNMENT
1181 Define this macro to be the value 1 if instructions will fail to work
1182 if given data not on the nominal alignment. If instructions will merely
1183 go slower in that case, define this macro as 0.
1186 @defmac PCC_BITFIELD_TYPE_MATTERS
1187 Define this if you wish to imitate the way many other C compilers handle
1188 alignment of bit-fields and the structures that contain them.
1190 The behavior is that the type written for a named bit-field (@code{int},
1191 @code{short}, or other integer type) imposes an alignment for the entire
1192 structure, as if the structure really did contain an ordinary field of
1193 that type. In addition, the bit-field is placed within the structure so
1194 that it would fit within such a field, not crossing a boundary for it.
1196 Thus, on most machines, a named bit-field whose type is written as
1197 @code{int} would not cross a four-byte boundary, and would force
1198 four-byte alignment for the whole structure. (The alignment used may
1199 not be four bytes; it is controlled by the other alignment parameters.)
1201 An unnamed bit-field will not affect the alignment of the containing
1204 If the macro is defined, its definition should be a C expression;
1205 a nonzero value for the expression enables this behavior.
1207 Note that if this macro is not defined, or its value is zero, some
1208 bit-fields may cross more than one alignment boundary. The compiler can
1209 support such references if there are @samp{insv}, @samp{extv}, and
1210 @samp{extzv} insns that can directly reference memory.
1212 The other known way of making bit-fields work is to define
1213 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1214 Then every structure can be accessed with fullwords.
1216 Unless the machine has bit-field instructions or you define
1217 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1218 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1220 If your aim is to make GCC use the same conventions for laying out
1221 bit-fields as are used by another compiler, here is how to investigate
1222 what the other compiler does. Compile and run this program:
1241 printf ("Size of foo1 is %d\n",
1242 sizeof (struct foo1));
1243 printf ("Size of foo2 is %d\n",
1244 sizeof (struct foo2));
1249 If this prints 2 and 5, then the compiler's behavior is what you would
1250 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1253 @defmac BITFIELD_NBYTES_LIMITED
1254 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1255 to aligning a bit-field within the structure.
1258 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1259 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1260 whether unnamed bitfields affect the alignment of the containing
1261 structure. The hook should return true if the structure should inherit
1262 the alignment requirements of an unnamed bitfield's type.
1265 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1266 This target hook should return @code{true} if accesses to volatile bitfields
1267 should use the narrowest mode possible. It should return @code{false} if
1268 these accesses should use the bitfield container type.
1270 The default is @code{!TARGET_STRICT_ALIGN}.
1273 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1274 Return 1 if a structure or array containing @var{field} should be accessed using
1277 If @var{field} is the only field in the structure, @var{mode} is its
1278 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1279 case where structures of one field would require the structure's mode to
1280 retain the field's mode.
1282 Normally, this is not needed.
1285 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1286 Define this macro as an expression for the alignment of a type (given
1287 by @var{type} as a tree node) if the alignment computed in the usual
1288 way is @var{computed} and the alignment explicitly specified was
1291 The default is to use @var{specified} if it is larger; otherwise, use
1292 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1295 @defmac MAX_FIXED_MODE_SIZE
1296 An integer expression for the size in bits of the largest integer
1297 machine mode that should actually be used. All integer machine modes of
1298 this size or smaller can be used for structures and unions with the
1299 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1300 (DImode)} is assumed.
1303 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1304 If defined, an expression of type @code{enum machine_mode} that
1305 specifies the mode of the save area operand of a
1306 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1307 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1308 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1309 having its mode specified.
1311 You need not define this macro if it always returns @code{Pmode}. You
1312 would most commonly define this macro if the
1313 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1317 @defmac STACK_SIZE_MODE
1318 If defined, an expression of type @code{enum machine_mode} that
1319 specifies the mode of the size increment operand of an
1320 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1322 You need not define this macro if it always returns @code{word_mode}.
1323 You would most commonly define this macro if the @code{allocate_stack}
1324 pattern needs to support both a 32- and a 64-bit mode.
1327 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE (void)
1328 This target hook should return the mode to be used for the return value
1329 of compare instructions expanded to libgcc calls. If not defined
1330 @code{word_mode} is returned which is the right choice for a majority of
1334 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1335 This target hook should return the mode to be used for the shift count operand
1336 of shift instructions expanded to libgcc calls. If not defined
1337 @code{word_mode} is returned which is the right choice for a majority of
1341 @deftypefn {Target Hook} {enum machine_mode} TARGET_UNWIND_WORD_MODE (void)
1342 Return machine mode to be used for @code{_Unwind_Word} type.
1343 The default is to use @code{word_mode}.
1346 @defmac ROUND_TOWARDS_ZERO
1347 If defined, this macro should be true if the prevailing rounding
1348 mode is towards zero.
1350 Defining this macro only affects the way @file{libgcc.a} emulates
1351 floating-point arithmetic.
1353 Not defining this macro is equivalent to returning zero.
1356 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1357 This macro should return true if floats with @var{size}
1358 bits do not have a NaN or infinity representation, but use the largest
1359 exponent for normal numbers instead.
1361 Defining this macro only affects the way @file{libgcc.a} emulates
1362 floating-point arithmetic.
1364 The default definition of this macro returns false for all sizes.
1367 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1368 This target hook returns @code{true} if bit-fields in the given
1369 @var{record_type} are to be laid out following the rules of Microsoft
1370 Visual C/C++, namely: (i) a bit-field won't share the same storage
1371 unit with the previous bit-field if their underlying types have
1372 different sizes, and the bit-field will be aligned to the highest
1373 alignment of the underlying types of itself and of the previous
1374 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1375 the whole enclosing structure, even if it is unnamed; except that
1376 (iii) a zero-sized bit-field will be disregarded unless it follows
1377 another bit-field of nonzero size. If this hook returns @code{true},
1378 other macros that control bit-field layout are ignored.
1380 When a bit-field is inserted into a packed record, the whole size
1381 of the underlying type is used by one or more same-size adjacent
1382 bit-fields (that is, if its long:3, 32 bits is used in the record,
1383 and any additional adjacent long bit-fields are packed into the same
1384 chunk of 32 bits. However, if the size changes, a new field of that
1385 size is allocated). In an unpacked record, this is the same as using
1386 alignment, but not equivalent when packing.
1388 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1389 the latter will take precedence. If @samp{__attribute__((packed))} is
1390 used on a single field when MS bit-fields are in use, it will take
1391 precedence for that field, but the alignment of the rest of the structure
1392 may affect its placement.
1395 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1396 Returns true if the target supports decimal floating point.
1399 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1400 Returns true if the target supports fixed-point arithmetic.
1403 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1404 This hook is called just before expansion into rtl, allowing the target
1405 to perform additional initializations or analysis before the expansion.
1406 For example, the rs6000 port uses it to allocate a scratch stack slot
1407 for use in copying SDmode values between memory and floating point
1408 registers whenever the function being expanded has any SDmode
1412 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1413 This hook allows the backend to perform additional instantiations on rtl
1414 that are not actually in any insns yet, but will be later.
1417 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1418 If your target defines any fundamental types, or any types your target
1419 uses should be mangled differently from the default, define this hook
1420 to return the appropriate encoding for these types as part of a C++
1421 mangled name. The @var{type} argument is the tree structure representing
1422 the type to be mangled. The hook may be applied to trees which are
1423 not target-specific fundamental types; it should return @code{NULL}
1424 for all such types, as well as arguments it does not recognize. If the
1425 return value is not @code{NULL}, it must point to a statically-allocated
1428 Target-specific fundamental types might be new fundamental types or
1429 qualified versions of ordinary fundamental types. Encode new
1430 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1431 is the name used for the type in source code, and @var{n} is the
1432 length of @var{name} in decimal. Encode qualified versions of
1433 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1434 @var{name} is the name used for the type qualifier in source code,
1435 @var{n} is the length of @var{name} as above, and @var{code} is the
1436 code used to represent the unqualified version of this type. (See
1437 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1438 codes.) In both cases the spaces are for clarity; do not include any
1439 spaces in your string.
1441 This hook is applied to types prior to typedef resolution. If the mangled
1442 name for a particular type depends only on that type's main variant, you
1443 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1446 The default version of this hook always returns @code{NULL}, which is
1447 appropriate for a target that does not define any new fundamental
1452 @section Layout of Source Language Data Types
1454 These macros define the sizes and other characteristics of the standard
1455 basic data types used in programs being compiled. Unlike the macros in
1456 the previous section, these apply to specific features of C and related
1457 languages, rather than to fundamental aspects of storage layout.
1459 @defmac INT_TYPE_SIZE
1460 A C expression for the size in bits of the type @code{int} on the
1461 target machine. If you don't define this, the default is one word.
1464 @defmac SHORT_TYPE_SIZE
1465 A C expression for the size in bits of the type @code{short} on the
1466 target machine. If you don't define this, the default is half a word.
1467 (If this would be less than one storage unit, it is rounded up to one
1471 @defmac LONG_TYPE_SIZE
1472 A C expression for the size in bits of the type @code{long} on the
1473 target machine. If you don't define this, the default is one word.
1476 @defmac ADA_LONG_TYPE_SIZE
1477 On some machines, the size used for the Ada equivalent of the type
1478 @code{long} by a native Ada compiler differs from that used by C@. In
1479 that situation, define this macro to be a C expression to be used for
1480 the size of that type. If you don't define this, the default is the
1481 value of @code{LONG_TYPE_SIZE}.
1484 @defmac LONG_LONG_TYPE_SIZE
1485 A C expression for the size in bits of the type @code{long long} on the
1486 target machine. If you don't define this, the default is two
1487 words. If you want to support GNU Ada on your machine, the value of this
1488 macro must be at least 64.
1491 @defmac CHAR_TYPE_SIZE
1492 A C expression for the size in bits of the type @code{char} on the
1493 target machine. If you don't define this, the default is
1494 @code{BITS_PER_UNIT}.
1497 @defmac BOOL_TYPE_SIZE
1498 A C expression for the size in bits of the C++ type @code{bool} and
1499 C99 type @code{_Bool} on the target machine. If you don't define
1500 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1503 @defmac FLOAT_TYPE_SIZE
1504 A C expression for the size in bits of the type @code{float} on the
1505 target machine. If you don't define this, the default is one word.
1508 @defmac DOUBLE_TYPE_SIZE
1509 A C expression for the size in bits of the type @code{double} on the
1510 target machine. If you don't define this, the default is two
1514 @defmac LONG_DOUBLE_TYPE_SIZE
1515 A C expression for the size in bits of the type @code{long double} on
1516 the target machine. If you don't define this, the default is two
1520 @defmac SHORT_FRACT_TYPE_SIZE
1521 A C expression for the size in bits of the type @code{short _Fract} on
1522 the target machine. If you don't define this, the default is
1523 @code{BITS_PER_UNIT}.
1526 @defmac FRACT_TYPE_SIZE
1527 A C expression for the size in bits of the type @code{_Fract} on
1528 the target machine. If you don't define this, the default is
1529 @code{BITS_PER_UNIT * 2}.
1532 @defmac LONG_FRACT_TYPE_SIZE
1533 A C expression for the size in bits of the type @code{long _Fract} on
1534 the target machine. If you don't define this, the default is
1535 @code{BITS_PER_UNIT * 4}.
1538 @defmac LONG_LONG_FRACT_TYPE_SIZE
1539 A C expression for the size in bits of the type @code{long long _Fract} on
1540 the target machine. If you don't define this, the default is
1541 @code{BITS_PER_UNIT * 8}.
1544 @defmac SHORT_ACCUM_TYPE_SIZE
1545 A C expression for the size in bits of the type @code{short _Accum} on
1546 the target machine. If you don't define this, the default is
1547 @code{BITS_PER_UNIT * 2}.
1550 @defmac ACCUM_TYPE_SIZE
1551 A C expression for the size in bits of the type @code{_Accum} on
1552 the target machine. If you don't define this, the default is
1553 @code{BITS_PER_UNIT * 4}.
1556 @defmac LONG_ACCUM_TYPE_SIZE
1557 A C expression for the size in bits of the type @code{long _Accum} on
1558 the target machine. If you don't define this, the default is
1559 @code{BITS_PER_UNIT * 8}.
1562 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1563 A C expression for the size in bits of the type @code{long long _Accum} on
1564 the target machine. If you don't define this, the default is
1565 @code{BITS_PER_UNIT * 16}.
1568 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1569 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1570 if you want routines in @file{libgcc2.a} for a size other than
1571 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1572 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1575 @defmac LIBGCC2_HAS_DF_MODE
1576 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1577 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1578 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1579 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1580 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1584 @defmac LIBGCC2_HAS_XF_MODE
1585 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1586 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1587 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1588 is 80 then the default is 1, otherwise it is 0.
1591 @defmac LIBGCC2_HAS_TF_MODE
1592 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1593 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1594 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1595 is 128 then the default is 1, otherwise it is 0.
1598 @defmac LIBGCC2_GNU_PREFIX
1599 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1600 hook and should be defined if that hook is overriden to be true. It
1601 causes function names in libgcc to be changed to use a @code{__gnu_}
1602 prefix for their name rather than the default @code{__}. A port which
1603 uses this macro should also arrange to use @file{t-gnu-prefix} in
1604 the libgcc @file{config.host}.
1611 Define these macros to be the size in bits of the mantissa of
1612 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1613 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1614 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1615 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1616 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1617 @code{DOUBLE_TYPE_SIZE} or
1618 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1621 @defmac TARGET_FLT_EVAL_METHOD
1622 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1623 assuming, if applicable, that the floating-point control word is in its
1624 default state. If you do not define this macro the value of
1625 @code{FLT_EVAL_METHOD} will be zero.
1628 @defmac WIDEST_HARDWARE_FP_SIZE
1629 A C expression for the size in bits of the widest floating-point format
1630 supported by the hardware. If you define this macro, you must specify a
1631 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1632 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1636 @defmac DEFAULT_SIGNED_CHAR
1637 An expression whose value is 1 or 0, according to whether the type
1638 @code{char} should be signed or unsigned by default. The user can
1639 always override this default with the options @option{-fsigned-char}
1640 and @option{-funsigned-char}.
1643 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1644 This target hook should return true if the compiler should give an
1645 @code{enum} type only as many bytes as it takes to represent the range
1646 of possible values of that type. It should return false if all
1647 @code{enum} types should be allocated like @code{int}.
1649 The default is to return false.
1653 A C expression for a string describing the name of the data type to use
1654 for size values. The typedef name @code{size_t} is defined using the
1655 contents of the string.
1657 The string can contain more than one keyword. If so, separate them with
1658 spaces, and write first any length keyword, then @code{unsigned} if
1659 appropriate, and finally @code{int}. The string must exactly match one
1660 of the data type names defined in the function
1661 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1662 omit @code{int} or change the order---that would cause the compiler to
1665 If you don't define this macro, the default is @code{"long unsigned
1669 @defmac PTRDIFF_TYPE
1670 A C expression for a string describing the name of the data type to use
1671 for the result of subtracting two pointers. The typedef name
1672 @code{ptrdiff_t} is defined using the contents of the string. See
1673 @code{SIZE_TYPE} above for more information.
1675 If you don't define this macro, the default is @code{"long int"}.
1679 A C expression for a string describing the name of the data type to use
1680 for wide characters. The typedef name @code{wchar_t} is defined using
1681 the contents of the string. See @code{SIZE_TYPE} above for more
1684 If you don't define this macro, the default is @code{"int"}.
1687 @defmac WCHAR_TYPE_SIZE
1688 A C expression for the size in bits of the data type for wide
1689 characters. This is used in @code{cpp}, which cannot make use of
1694 A C expression for a string describing the name of the data type to
1695 use for wide characters passed to @code{printf} and returned from
1696 @code{getwc}. The typedef name @code{wint_t} is defined using the
1697 contents of the string. See @code{SIZE_TYPE} above for more
1700 If you don't define this macro, the default is @code{"unsigned int"}.
1704 A C expression for a string describing the name of the data type that
1705 can represent any value of any standard or extended signed integer type.
1706 The typedef name @code{intmax_t} is defined using the contents of the
1707 string. See @code{SIZE_TYPE} above for more information.
1709 If you don't define this macro, the default is the first of
1710 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1711 much precision as @code{long long int}.
1714 @defmac UINTMAX_TYPE
1715 A C expression for a string describing the name of the data type that
1716 can represent any value of any standard or extended unsigned integer
1717 type. The typedef name @code{uintmax_t} is defined using the contents
1718 of the string. See @code{SIZE_TYPE} above for more information.
1720 If you don't define this macro, the default is the first of
1721 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1722 unsigned int"} that has as much precision as @code{long long unsigned
1726 @defmac SIG_ATOMIC_TYPE
1732 @defmacx UINT16_TYPE
1733 @defmacx UINT32_TYPE
1734 @defmacx UINT64_TYPE
1735 @defmacx INT_LEAST8_TYPE
1736 @defmacx INT_LEAST16_TYPE
1737 @defmacx INT_LEAST32_TYPE
1738 @defmacx INT_LEAST64_TYPE
1739 @defmacx UINT_LEAST8_TYPE
1740 @defmacx UINT_LEAST16_TYPE
1741 @defmacx UINT_LEAST32_TYPE
1742 @defmacx UINT_LEAST64_TYPE
1743 @defmacx INT_FAST8_TYPE
1744 @defmacx INT_FAST16_TYPE
1745 @defmacx INT_FAST32_TYPE
1746 @defmacx INT_FAST64_TYPE
1747 @defmacx UINT_FAST8_TYPE
1748 @defmacx UINT_FAST16_TYPE
1749 @defmacx UINT_FAST32_TYPE
1750 @defmacx UINT_FAST64_TYPE
1751 @defmacx INTPTR_TYPE
1752 @defmacx UINTPTR_TYPE
1753 C expressions for the standard types @code{sig_atomic_t},
1754 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1755 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1756 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1757 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1758 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1759 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1760 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1761 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1762 @code{SIZE_TYPE} above for more information.
1764 If any of these macros evaluates to a null pointer, the corresponding
1765 type is not supported; if GCC is configured to provide
1766 @code{<stdint.h>} in such a case, the header provided may not conform
1767 to C99, depending on the type in question. The defaults for all of
1768 these macros are null pointers.
1771 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1772 The C++ compiler represents a pointer-to-member-function with a struct
1779 ptrdiff_t vtable_index;
1786 The C++ compiler must use one bit to indicate whether the function that
1787 will be called through a pointer-to-member-function is virtual.
1788 Normally, we assume that the low-order bit of a function pointer must
1789 always be zero. Then, by ensuring that the vtable_index is odd, we can
1790 distinguish which variant of the union is in use. But, on some
1791 platforms function pointers can be odd, and so this doesn't work. In
1792 that case, we use the low-order bit of the @code{delta} field, and shift
1793 the remainder of the @code{delta} field to the left.
1795 GCC will automatically make the right selection about where to store
1796 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1797 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1798 set such that functions always start at even addresses, but the lowest
1799 bit of pointers to functions indicate whether the function at that
1800 address is in ARM or Thumb mode. If this is the case of your
1801 architecture, you should define this macro to
1802 @code{ptrmemfunc_vbit_in_delta}.
1804 In general, you should not have to define this macro. On architectures
1805 in which function addresses are always even, according to
1806 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1807 @code{ptrmemfunc_vbit_in_pfn}.
1810 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1811 Normally, the C++ compiler uses function pointers in vtables. This
1812 macro allows the target to change to use ``function descriptors''
1813 instead. Function descriptors are found on targets for whom a
1814 function pointer is actually a small data structure. Normally the
1815 data structure consists of the actual code address plus a data
1816 pointer to which the function's data is relative.
1818 If vtables are used, the value of this macro should be the number
1819 of words that the function descriptor occupies.
1822 @defmac TARGET_VTABLE_ENTRY_ALIGN
1823 By default, the vtable entries are void pointers, the so the alignment
1824 is the same as pointer alignment. The value of this macro specifies
1825 the alignment of the vtable entry in bits. It should be defined only
1826 when special alignment is necessary. */
1829 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1830 There are a few non-descriptor entries in the vtable at offsets below
1831 zero. If these entries must be padded (say, to preserve the alignment
1832 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1833 of words in each data entry.
1837 @section Register Usage
1838 @cindex register usage
1840 This section explains how to describe what registers the target machine
1841 has, and how (in general) they can be used.
1843 The description of which registers a specific instruction can use is
1844 done with register classes; see @ref{Register Classes}. For information
1845 on using registers to access a stack frame, see @ref{Frame Registers}.
1846 For passing values in registers, see @ref{Register Arguments}.
1847 For returning values in registers, see @ref{Scalar Return}.
1850 * Register Basics:: Number and kinds of registers.
1851 * Allocation Order:: Order in which registers are allocated.
1852 * Values in Registers:: What kinds of values each reg can hold.
1853 * Leaf Functions:: Renumbering registers for leaf functions.
1854 * Stack Registers:: Handling a register stack such as 80387.
1857 @node Register Basics
1858 @subsection Basic Characteristics of Registers
1860 @c prevent bad page break with this line
1861 Registers have various characteristics.
1863 @defmac FIRST_PSEUDO_REGISTER
1864 Number of hardware registers known to the compiler. They receive
1865 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1866 pseudo register's number really is assigned the number
1867 @code{FIRST_PSEUDO_REGISTER}.
1870 @defmac FIXED_REGISTERS
1871 @cindex fixed register
1872 An initializer that says which registers are used for fixed purposes
1873 all throughout the compiled code and are therefore not available for
1874 general allocation. These would include the stack pointer, the frame
1875 pointer (except on machines where that can be used as a general
1876 register when no frame pointer is needed), the program counter on
1877 machines where that is considered one of the addressable registers,
1878 and any other numbered register with a standard use.
1880 This information is expressed as a sequence of numbers, separated by
1881 commas and surrounded by braces. The @var{n}th number is 1 if
1882 register @var{n} is fixed, 0 otherwise.
1884 The table initialized from this macro, and the table initialized by
1885 the following one, may be overridden at run time either automatically,
1886 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1887 the user with the command options @option{-ffixed-@var{reg}},
1888 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1891 @defmac CALL_USED_REGISTERS
1892 @cindex call-used register
1893 @cindex call-clobbered register
1894 @cindex call-saved register
1895 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1896 clobbered (in general) by function calls as well as for fixed
1897 registers. This macro therefore identifies the registers that are not
1898 available for general allocation of values that must live across
1901 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1902 automatically saves it on function entry and restores it on function
1903 exit, if the register is used within the function.
1906 @defmac CALL_REALLY_USED_REGISTERS
1907 @cindex call-used register
1908 @cindex call-clobbered register
1909 @cindex call-saved register
1910 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1911 that the entire set of @code{FIXED_REGISTERS} be included.
1912 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1913 This macro is optional. If not specified, it defaults to the value
1914 of @code{CALL_USED_REGISTERS}.
1917 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1918 @cindex call-used register
1919 @cindex call-clobbered register
1920 @cindex call-saved register
1921 A C expression that is nonzero if it is not permissible to store a
1922 value of mode @var{mode} in hard register number @var{regno} across a
1923 call without some part of it being clobbered. For most machines this
1924 macro need not be defined. It is only required for machines that do not
1925 preserve the entire contents of a register across a call.
1929 @findex call_used_regs
1932 @findex reg_class_contents
1933 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1934 This hook may conditionally modify five variables
1935 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1936 @code{reg_names}, and @code{reg_class_contents}, to take into account
1937 any dependence of these register sets on target flags. The first three
1938 of these are of type @code{char []} (interpreted as Boolean vectors).
1939 @code{global_regs} is a @code{const char *[]}, and
1940 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1941 called, @code{fixed_regs}, @code{call_used_regs},
1942 @code{reg_class_contents}, and @code{reg_names} have been initialized
1943 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1944 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1945 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1946 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1947 command options have been applied.
1949 @cindex disabling certain registers
1950 @cindex controlling register usage
1951 If the usage of an entire class of registers depends on the target
1952 flags, you may indicate this to GCC by using this macro to modify
1953 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1954 registers in the classes which should not be used by GCC@. Also define
1955 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1956 to return @code{NO_REGS} if it
1957 is called with a letter for a class that shouldn't be used.
1959 (However, if this class is not included in @code{GENERAL_REGS} and all
1960 of the insn patterns whose constraints permit this class are
1961 controlled by target switches, then GCC will automatically avoid using
1962 these registers when the target switches are opposed to them.)
1965 @defmac INCOMING_REGNO (@var{out})
1966 Define this macro if the target machine has register windows. This C
1967 expression returns the register number as seen by the called function
1968 corresponding to the register number @var{out} as seen by the calling
1969 function. Return @var{out} if register number @var{out} is not an
1973 @defmac OUTGOING_REGNO (@var{in})
1974 Define this macro if the target machine has register windows. This C
1975 expression returns the register number as seen by the calling function
1976 corresponding to the register number @var{in} as seen by the called
1977 function. Return @var{in} if register number @var{in} is not an inbound
1981 @defmac LOCAL_REGNO (@var{regno})
1982 Define this macro if the target machine has register windows. This C
1983 expression returns true if the register is call-saved but is in the
1984 register window. Unlike most call-saved registers, such registers
1985 need not be explicitly restored on function exit or during non-local
1990 If the program counter has a register number, define this as that
1991 register number. Otherwise, do not define it.
1994 @node Allocation Order
1995 @subsection Order of Allocation of Registers
1996 @cindex order of register allocation
1997 @cindex register allocation order
1999 @c prevent bad page break with this line
2000 Registers are allocated in order.
2002 @defmac REG_ALLOC_ORDER
2003 If defined, an initializer for a vector of integers, containing the
2004 numbers of hard registers in the order in which GCC should prefer
2005 to use them (from most preferred to least).
2007 If this macro is not defined, registers are used lowest numbered first
2008 (all else being equal).
2010 One use of this macro is on machines where the highest numbered
2011 registers must always be saved and the save-multiple-registers
2012 instruction supports only sequences of consecutive registers. On such
2013 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2014 the highest numbered allocable register first.
2017 @defmac ADJUST_REG_ALLOC_ORDER
2018 A C statement (sans semicolon) to choose the order in which to allocate
2019 hard registers for pseudo-registers local to a basic block.
2021 Store the desired register order in the array @code{reg_alloc_order}.
2022 Element 0 should be the register to allocate first; element 1, the next
2023 register; and so on.
2025 The macro body should not assume anything about the contents of
2026 @code{reg_alloc_order} before execution of the macro.
2028 On most machines, it is not necessary to define this macro.
2031 @defmac HONOR_REG_ALLOC_ORDER
2032 Normally, IRA tries to estimate the costs for saving a register in the
2033 prologue and restoring it in the epilogue. This discourages it from
2034 using call-saved registers. If a machine wants to ensure that IRA
2035 allocates registers in the order given by REG_ALLOC_ORDER even if some
2036 call-saved registers appear earlier than call-used ones, this macro
2040 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2041 In some case register allocation order is not enough for the
2042 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2043 If this macro is defined, it should return a floating point value
2044 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2045 be increased by approximately the pseudo's usage frequency times the
2046 value returned by this macro. Not defining this macro is equivalent
2047 to having it always return @code{0.0}.
2049 On most machines, it is not necessary to define this macro.
2052 @node Values in Registers
2053 @subsection How Values Fit in Registers
2055 This section discusses the macros that describe which kinds of values
2056 (specifically, which machine modes) each register can hold, and how many
2057 consecutive registers are needed for a given mode.
2059 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2060 A C expression for the number of consecutive hard registers, starting
2061 at register number @var{regno}, required to hold a value of mode
2062 @var{mode}. This macro must never return zero, even if a register
2063 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2064 and/or CANNOT_CHANGE_MODE_CLASS instead.
2066 On a machine where all registers are exactly one word, a suitable
2067 definition of this macro is
2070 #define HARD_REGNO_NREGS(REGNO, MODE) \
2071 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2076 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2077 A C expression that is nonzero if a value of mode @var{mode}, stored
2078 in memory, ends with padding that causes it to take up more space than
2079 in registers starting at register number @var{regno} (as determined by
2080 multiplying GCC's notion of the size of the register when containing
2081 this mode by the number of registers returned by
2082 @code{HARD_REGNO_NREGS}). By default this is zero.
2084 For example, if a floating-point value is stored in three 32-bit
2085 registers but takes up 128 bits in memory, then this would be
2088 This macros only needs to be defined if there are cases where
2089 @code{subreg_get_info}
2090 would otherwise wrongly determine that a @code{subreg} can be
2091 represented by an offset to the register number, when in fact such a
2092 @code{subreg} would contain some of the padding not stored in
2093 registers and so not be representable.
2096 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2097 For values of @var{regno} and @var{mode} for which
2098 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2099 returning the greater number of registers required to hold the value
2100 including any padding. In the example above, the value would be four.
2103 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2104 Define this macro if the natural size of registers that hold values
2105 of mode @var{mode} is not the word size. It is a C expression that
2106 should give the natural size in bytes for the specified mode. It is
2107 used by the register allocator to try to optimize its results. This
2108 happens for example on SPARC 64-bit where the natural size of
2109 floating-point registers is still 32-bit.
2112 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2113 A C expression that is nonzero if it is permissible to store a value
2114 of mode @var{mode} in hard register number @var{regno} (or in several
2115 registers starting with that one). For a machine where all registers
2116 are equivalent, a suitable definition is
2119 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2122 You need not include code to check for the numbers of fixed registers,
2123 because the allocation mechanism considers them to be always occupied.
2125 @cindex register pairs
2126 On some machines, double-precision values must be kept in even/odd
2127 register pairs. You can implement that by defining this macro to reject
2128 odd register numbers for such modes.
2130 The minimum requirement for a mode to be OK in a register is that the
2131 @samp{mov@var{mode}} instruction pattern support moves between the
2132 register and other hard register in the same class and that moving a
2133 value into the register and back out not alter it.
2135 Since the same instruction used to move @code{word_mode} will work for
2136 all narrower integer modes, it is not necessary on any machine for
2137 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2138 you define patterns @samp{movhi}, etc., to take advantage of this. This
2139 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2140 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2143 Many machines have special registers for floating point arithmetic.
2144 Often people assume that floating point machine modes are allowed only
2145 in floating point registers. This is not true. Any registers that
2146 can hold integers can safely @emph{hold} a floating point machine
2147 mode, whether or not floating arithmetic can be done on it in those
2148 registers. Integer move instructions can be used to move the values.
2150 On some machines, though, the converse is true: fixed-point machine
2151 modes may not go in floating registers. This is true if the floating
2152 registers normalize any value stored in them, because storing a
2153 non-floating value there would garble it. In this case,
2154 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2155 floating registers. But if the floating registers do not automatically
2156 normalize, if you can store any bit pattern in one and retrieve it
2157 unchanged without a trap, then any machine mode may go in a floating
2158 register, so you can define this macro to say so.
2160 The primary significance of special floating registers is rather that
2161 they are the registers acceptable in floating point arithmetic
2162 instructions. However, this is of no concern to
2163 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2164 constraints for those instructions.
2166 On some machines, the floating registers are especially slow to access,
2167 so that it is better to store a value in a stack frame than in such a
2168 register if floating point arithmetic is not being done. As long as the
2169 floating registers are not in class @code{GENERAL_REGS}, they will not
2170 be used unless some pattern's constraint asks for one.
2173 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2174 A C expression that is nonzero if it is OK to rename a hard register
2175 @var{from} to another hard register @var{to}.
2177 One common use of this macro is to prevent renaming of a register to
2178 another register that is not saved by a prologue in an interrupt
2181 The default is always nonzero.
2184 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2185 A C expression that is nonzero if a value of mode
2186 @var{mode1} is accessible in mode @var{mode2} without copying.
2188 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2189 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2190 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2191 should be nonzero. If they differ for any @var{r}, you should define
2192 this macro to return zero unless some other mechanism ensures the
2193 accessibility of the value in a narrower mode.
2195 You should define this macro to return nonzero in as many cases as
2196 possible since doing so will allow GCC to perform better register
2200 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2201 This target hook should return @code{true} if it is OK to use a hard register
2202 @var{regno} as scratch reg in peephole2.
2204 One common use of this macro is to prevent using of a register that
2205 is not saved by a prologue in an interrupt handler.
2207 The default version of this hook always returns @code{true}.
2210 @defmac AVOID_CCMODE_COPIES
2211 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2212 registers. You should only define this macro if support for copying to/from
2213 @code{CCmode} is incomplete.
2216 @node Leaf Functions
2217 @subsection Handling Leaf Functions
2219 @cindex leaf functions
2220 @cindex functions, leaf
2221 On some machines, a leaf function (i.e., one which makes no calls) can run
2222 more efficiently if it does not make its own register window. Often this
2223 means it is required to receive its arguments in the registers where they
2224 are passed by the caller, instead of the registers where they would
2227 The special treatment for leaf functions generally applies only when
2228 other conditions are met; for example, often they may use only those
2229 registers for its own variables and temporaries. We use the term ``leaf
2230 function'' to mean a function that is suitable for this special
2231 handling, so that functions with no calls are not necessarily ``leaf
2234 GCC assigns register numbers before it knows whether the function is
2235 suitable for leaf function treatment. So it needs to renumber the
2236 registers in order to output a leaf function. The following macros
2239 @defmac LEAF_REGISTERS
2240 Name of a char vector, indexed by hard register number, which
2241 contains 1 for a register that is allowable in a candidate for leaf
2244 If leaf function treatment involves renumbering the registers, then the
2245 registers marked here should be the ones before renumbering---those that
2246 GCC would ordinarily allocate. The registers which will actually be
2247 used in the assembler code, after renumbering, should not be marked with 1
2250 Define this macro only if the target machine offers a way to optimize
2251 the treatment of leaf functions.
2254 @defmac LEAF_REG_REMAP (@var{regno})
2255 A C expression whose value is the register number to which @var{regno}
2256 should be renumbered, when a function is treated as a leaf function.
2258 If @var{regno} is a register number which should not appear in a leaf
2259 function before renumbering, then the expression should yield @minus{}1, which
2260 will cause the compiler to abort.
2262 Define this macro only if the target machine offers a way to optimize the
2263 treatment of leaf functions, and registers need to be renumbered to do
2267 @findex current_function_is_leaf
2268 @findex current_function_uses_only_leaf_regs
2269 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2270 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2271 specially. They can test the C variable @code{current_function_is_leaf}
2272 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2273 set prior to local register allocation and is valid for the remaining
2274 compiler passes. They can also test the C variable
2275 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2276 functions which only use leaf registers.
2277 @code{current_function_uses_only_leaf_regs} is valid after all passes
2278 that modify the instructions have been run and is only useful if
2279 @code{LEAF_REGISTERS} is defined.
2280 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2281 @c of the next paragraph?! --mew 2feb93
2283 @node Stack Registers
2284 @subsection Registers That Form a Stack
2286 There are special features to handle computers where some of the
2287 ``registers'' form a stack. Stack registers are normally written by
2288 pushing onto the stack, and are numbered relative to the top of the
2291 Currently, GCC can only handle one group of stack-like registers, and
2292 they must be consecutively numbered. Furthermore, the existing
2293 support for stack-like registers is specific to the 80387 floating
2294 point coprocessor. If you have a new architecture that uses
2295 stack-like registers, you will need to do substantial work on
2296 @file{reg-stack.c} and write your machine description to cooperate
2297 with it, as well as defining these macros.
2300 Define this if the machine has any stack-like registers.
2303 @defmac STACK_REG_COVER_CLASS
2304 This is a cover class containing the stack registers. Define this if
2305 the machine has any stack-like registers.
2308 @defmac FIRST_STACK_REG
2309 The number of the first stack-like register. This one is the top
2313 @defmac LAST_STACK_REG
2314 The number of the last stack-like register. This one is the bottom of
2318 @node Register Classes
2319 @section Register Classes
2320 @cindex register class definitions
2321 @cindex class definitions, register
2323 On many machines, the numbered registers are not all equivalent.
2324 For example, certain registers may not be allowed for indexed addressing;
2325 certain registers may not be allowed in some instructions. These machine
2326 restrictions are described to the compiler using @dfn{register classes}.
2328 You define a number of register classes, giving each one a name and saying
2329 which of the registers belong to it. Then you can specify register classes
2330 that are allowed as operands to particular instruction patterns.
2334 In general, each register will belong to several classes. In fact, one
2335 class must be named @code{ALL_REGS} and contain all the registers. Another
2336 class must be named @code{NO_REGS} and contain no registers. Often the
2337 union of two classes will be another class; however, this is not required.
2339 @findex GENERAL_REGS
2340 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2341 terribly special about the name, but the operand constraint letters
2342 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2343 the same as @code{ALL_REGS}, just define it as a macro which expands
2346 Order the classes so that if class @var{x} is contained in class @var{y}
2347 then @var{x} has a lower class number than @var{y}.
2349 The way classes other than @code{GENERAL_REGS} are specified in operand
2350 constraints is through machine-dependent operand constraint letters.
2351 You can define such letters to correspond to various classes, then use
2352 them in operand constraints.
2354 You must define the narrowest register classes for allocatable
2355 registers, so that each class either has no subclasses, or that for
2356 some mode, the move cost between registers within the class is
2357 cheaper than moving a register in the class to or from memory
2360 You should define a class for the union of two classes whenever some
2361 instruction allows both classes. For example, if an instruction allows
2362 either a floating point (coprocessor) register or a general register for a
2363 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2364 which includes both of them. Otherwise you will get suboptimal code,
2365 or even internal compiler errors when reload cannot find a register in the
2366 class computed via @code{reg_class_subunion}.
2368 You must also specify certain redundant information about the register
2369 classes: for each class, which classes contain it and which ones are
2370 contained in it; for each pair of classes, the largest class contained
2373 When a value occupying several consecutive registers is expected in a
2374 certain class, all the registers used must belong to that class.
2375 Therefore, register classes cannot be used to enforce a requirement for
2376 a register pair to start with an even-numbered register. The way to
2377 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2379 Register classes used for input-operands of bitwise-and or shift
2380 instructions have a special requirement: each such class must have, for
2381 each fixed-point machine mode, a subclass whose registers can transfer that
2382 mode to or from memory. For example, on some machines, the operations for
2383 single-byte values (@code{QImode}) are limited to certain registers. When
2384 this is so, each register class that is used in a bitwise-and or shift
2385 instruction must have a subclass consisting of registers from which
2386 single-byte values can be loaded or stored. This is so that
2387 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2389 @deftp {Data type} {enum reg_class}
2390 An enumerated type that must be defined with all the register class names
2391 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2392 must be the last register class, followed by one more enumerated value,
2393 @code{LIM_REG_CLASSES}, which is not a register class but rather
2394 tells how many classes there are.
2396 Each register class has a number, which is the value of casting
2397 the class name to type @code{int}. The number serves as an index
2398 in many of the tables described below.
2401 @defmac N_REG_CLASSES
2402 The number of distinct register classes, defined as follows:
2405 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2409 @defmac REG_CLASS_NAMES
2410 An initializer containing the names of the register classes as C string
2411 constants. These names are used in writing some of the debugging dumps.
2414 @defmac REG_CLASS_CONTENTS
2415 An initializer containing the contents of the register classes, as integers
2416 which are bit masks. The @var{n}th integer specifies the contents of class
2417 @var{n}. The way the integer @var{mask} is interpreted is that
2418 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2420 When the machine has more than 32 registers, an integer does not suffice.
2421 Then the integers are replaced by sub-initializers, braced groupings containing
2422 several integers. Each sub-initializer must be suitable as an initializer
2423 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2424 In this situation, the first integer in each sub-initializer corresponds to
2425 registers 0 through 31, the second integer to registers 32 through 63, and
2429 @defmac REGNO_REG_CLASS (@var{regno})
2430 A C expression whose value is a register class containing hard register
2431 @var{regno}. In general there is more than one such class; choose a class
2432 which is @dfn{minimal}, meaning that no smaller class also contains the
2436 @defmac BASE_REG_CLASS
2437 A macro whose definition is the name of the class to which a valid
2438 base register must belong. A base register is one used in an address
2439 which is the register value plus a displacement.
2442 @defmac MODE_BASE_REG_CLASS (@var{mode})
2443 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2444 the selection of a base register in a mode dependent manner. If
2445 @var{mode} is VOIDmode then it should return the same value as
2446 @code{BASE_REG_CLASS}.
2449 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2450 A C expression whose value is the register class to which a valid
2451 base register must belong in order to be used in a base plus index
2452 register address. You should define this macro if base plus index
2453 addresses have different requirements than other base register uses.
2456 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2457 A C expression whose value is the register class to which a valid
2458 base register must belong. @var{outer_code} and @var{index_code} define the
2459 context in which the base register occurs. @var{outer_code} is the code of
2460 the immediately enclosing expression (@code{MEM} for the top level of an
2461 address, @code{ADDRESS} for something that occurs in an
2462 @code{address_operand}). @var{index_code} is the code of the corresponding
2463 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2466 @defmac INDEX_REG_CLASS
2467 A macro whose definition is the name of the class to which a valid
2468 index register must belong. An index register is one used in an
2469 address where its value is either multiplied by a scale factor or
2470 added to another register (as well as added to a displacement).
2473 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2474 A C expression which is nonzero if register number @var{num} is
2475 suitable for use as a base register in operand addresses.
2478 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2479 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2480 that expression may examine the mode of the memory reference in
2481 @var{mode}. You should define this macro if the mode of the memory
2482 reference affects whether a register may be used as a base register. If
2483 you define this macro, the compiler will use it instead of
2484 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2485 addresses that appear outside a @code{MEM}, i.e., as an
2486 @code{address_operand}.
2489 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2490 A C expression which is nonzero if register number @var{num} is suitable for
2491 use as a base register in base plus index operand addresses, accessing
2492 memory in mode @var{mode}. It may be either a suitable hard register or a
2493 pseudo register that has been allocated such a hard register. You should
2494 define this macro if base plus index addresses have different requirements
2495 than other base register uses.
2497 Use of this macro is deprecated; please use the more general
2498 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2501 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2502 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2503 that that expression may examine the context in which the register
2504 appears in the memory reference. @var{outer_code} is the code of the
2505 immediately enclosing expression (@code{MEM} if at the top level of the
2506 address, @code{ADDRESS} for something that occurs in an
2507 @code{address_operand}). @var{index_code} is the code of the
2508 corresponding index expression if @var{outer_code} is @code{PLUS};
2509 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2510 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2513 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2514 A C expression which is nonzero if register number @var{num} is
2515 suitable for use as an index register in operand addresses. It may be
2516 either a suitable hard register or a pseudo register that has been
2517 allocated such a hard register.
2519 The difference between an index register and a base register is that
2520 the index register may be scaled. If an address involves the sum of
2521 two registers, neither one of them scaled, then either one may be
2522 labeled the ``base'' and the other the ``index''; but whichever
2523 labeling is used must fit the machine's constraints of which registers
2524 may serve in each capacity. The compiler will try both labelings,
2525 looking for one that is valid, and will reload one or both registers
2526 only if neither labeling works.
2529 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2530 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.
2533 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2534 A target hook that places additional restrictions on the register class
2535 to use when it is necessary to copy value @var{x} into a register in class
2536 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2537 another, smaller class.
2539 The default version of this hook always returns value of @code{rclass} argument.
2541 Sometimes returning a more restrictive class makes better code. For
2542 example, on the 68000, when @var{x} is an integer constant that is in range
2543 for a @samp{moveq} instruction, the value of this macro is always
2544 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2545 Requiring a data register guarantees that a @samp{moveq} will be used.
2547 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2548 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2549 loaded into some register class. By returning @code{NO_REGS} you can
2550 force @var{x} into a memory location. For example, rs6000 can load
2551 immediate values into general-purpose registers, but does not have an
2552 instruction for loading an immediate value into a floating-point
2553 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2554 @var{x} is a floating-point constant. If the constant can't be loaded
2555 into any kind of register, code generation will be better if
2556 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2557 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2559 If an insn has pseudos in it after register allocation, reload will go
2560 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2561 to find the best one. Returning @code{NO_REGS}, in this case, makes
2562 reload add a @code{!} in front of the constraint: the x86 back-end uses
2563 this feature to discourage usage of 387 registers when math is done in
2564 the SSE registers (and vice versa).
2567 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2568 A C expression that places additional restrictions on the register class
2569 to use when it is necessary to copy value @var{x} into a register in class
2570 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2571 another, smaller class. On many machines, the following definition is
2575 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2578 Sometimes returning a more restrictive class makes better code. For
2579 example, on the 68000, when @var{x} is an integer constant that is in range
2580 for a @samp{moveq} instruction, the value of this macro is always
2581 @code{DATA_REGS} as long as @var{class} includes the data registers.
2582 Requiring a data register guarantees that a @samp{moveq} will be used.
2584 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2585 @var{class} is if @var{x} is a legitimate constant which cannot be
2586 loaded into some register class. By returning @code{NO_REGS} you can
2587 force @var{x} into a memory location. For example, rs6000 can load
2588 immediate values into general-purpose registers, but does not have an
2589 instruction for loading an immediate value into a floating-point
2590 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2591 @var{x} is a floating-point constant. If the constant can't be loaded
2592 into any kind of register, code generation will be better if
2593 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2594 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2596 If an insn has pseudos in it after register allocation, reload will go
2597 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2598 to find the best one. Returning @code{NO_REGS}, in this case, makes
2599 reload add a @code{!} in front of the constraint: the x86 back-end uses
2600 this feature to discourage usage of 387 registers when math is done in
2601 the SSE registers (and vice versa).
2604 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2605 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2606 input reloads. If you don't define this macro, the default is to use
2607 @var{class}, unchanged.
2609 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2610 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2613 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2614 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2617 The default version of this hook always returns value of @code{rclass}
2620 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2621 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2624 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2625 A C expression that places additional restrictions on the register class
2626 to use when it is necessary to be able to hold a value of mode
2627 @var{mode} in a reload register for which class @var{class} would
2630 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2631 there are certain modes that simply can't go in certain reload classes.
2633 The value is a register class; perhaps @var{class}, or perhaps another,
2636 Don't define this macro unless the target machine has limitations which
2637 require the macro to do something nontrivial.
2640 @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})
2641 Many machines have some registers that cannot be copied directly to or
2642 from memory or even from other types of registers. An example is the
2643 @samp{MQ} register, which on most machines, can only be copied to or
2644 from general registers, but not memory. Below, we shall be using the
2645 term 'intermediate register' when a move operation cannot be performed
2646 directly, but has to be done by copying the source into the intermediate
2647 register first, and then copying the intermediate register to the
2648 destination. An intermediate register always has the same mode as
2649 source and destination. Since it holds the actual value being copied,
2650 reload might apply optimizations to re-use an intermediate register
2651 and eliding the copy from the source when it can determine that the
2652 intermediate register still holds the required value.
2654 Another kind of secondary reload is required on some machines which
2655 allow copying all registers to and from memory, but require a scratch
2656 register for stores to some memory locations (e.g., those with symbolic
2657 address on the RT, and those with certain symbolic address on the SPARC
2658 when compiling PIC)@. Scratch registers need not have the same mode
2659 as the value being copied, and usually hold a different value than
2660 that being copied. Special patterns in the md file are needed to
2661 describe how the copy is performed with the help of the scratch register;
2662 these patterns also describe the number, register class(es) and mode(s)
2663 of the scratch register(s).
2665 In some cases, both an intermediate and a scratch register are required.
2667 For input reloads, this target hook is called with nonzero @var{in_p},
2668 and @var{x} is an rtx that needs to be copied to a register of class
2669 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2670 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2671 needs to be copied to rtx @var{x} in @var{reload_mode}.
2673 If copying a register of @var{reload_class} from/to @var{x} requires
2674 an intermediate register, the hook @code{secondary_reload} should
2675 return the register class required for this intermediate register.
2676 If no intermediate register is required, it should return NO_REGS.
2677 If more than one intermediate register is required, describe the one
2678 that is closest in the copy chain to the reload register.
2680 If scratch registers are needed, you also have to describe how to
2681 perform the copy from/to the reload register to/from this
2682 closest intermediate register. Or if no intermediate register is
2683 required, but still a scratch register is needed, describe the
2684 copy from/to the reload register to/from the reload operand @var{x}.
2686 You do this by setting @code{sri->icode} to the instruction code of a pattern
2687 in the md file which performs the move. Operands 0 and 1 are the output
2688 and input of this copy, respectively. Operands from operand 2 onward are
2689 for scratch operands. These scratch operands must have a mode, and a
2690 single-register-class
2691 @c [later: or memory]
2694 When an intermediate register is used, the @code{secondary_reload}
2695 hook will be called again to determine how to copy the intermediate
2696 register to/from the reload operand @var{x}, so your hook must also
2697 have code to handle the register class of the intermediate operand.
2699 @c [For later: maybe we'll allow multi-alternative reload patterns -
2700 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2701 @c and match the constraints of input and output to determine the required
2702 @c alternative. A restriction would be that constraints used to match
2703 @c against reloads registers would have to be written as register class
2704 @c constraints, or we need a new target macro / hook that tells us if an
2705 @c arbitrary constraint can match an unknown register of a given class.
2706 @c Such a macro / hook would also be useful in other places.]
2709 @var{x} might be a pseudo-register or a @code{subreg} of a
2710 pseudo-register, which could either be in a hard register or in memory.
2711 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2712 in memory and the hard register number if it is in a register.
2714 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2715 currently not supported. For the time being, you will have to continue
2716 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2718 @code{copy_cost} also uses this target hook to find out how values are
2719 copied. If you want it to include some extra cost for the need to allocate
2720 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2721 Or if two dependent moves are supposed to have a lower cost than the sum
2722 of the individual moves due to expected fortuitous scheduling and/or special
2723 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2726 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2727 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2728 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2729 These macros are obsolete, new ports should use the target hook
2730 @code{TARGET_SECONDARY_RELOAD} instead.
2732 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2733 target hook. Older ports still define these macros to indicate to the
2734 reload phase that it may
2735 need to allocate at least one register for a reload in addition to the
2736 register to contain the data. Specifically, if copying @var{x} to a
2737 register @var{class} in @var{mode} requires an intermediate register,
2738 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2739 largest register class all of whose registers can be used as
2740 intermediate registers or scratch registers.
2742 If copying a register @var{class} in @var{mode} to @var{x} requires an
2743 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2744 was supposed to be defined be defined to return the largest register
2745 class required. If the
2746 requirements for input and output reloads were the same, the macro
2747 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2750 The values returned by these macros are often @code{GENERAL_REGS}.
2751 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2752 can be directly copied to or from a register of @var{class} in
2753 @var{mode} without requiring a scratch register. Do not define this
2754 macro if it would always return @code{NO_REGS}.
2756 If a scratch register is required (either with or without an
2757 intermediate register), you were supposed to define patterns for
2758 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2759 (@pxref{Standard Names}. These patterns, which were normally
2760 implemented with a @code{define_expand}, should be similar to the
2761 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2764 These patterns need constraints for the reload register and scratch
2766 contain a single register class. If the original reload register (whose
2767 class is @var{class}) can meet the constraint given in the pattern, the
2768 value returned by these macros is used for the class of the scratch
2769 register. Otherwise, two additional reload registers are required.
2770 Their classes are obtained from the constraints in the insn pattern.
2772 @var{x} might be a pseudo-register or a @code{subreg} of a
2773 pseudo-register, which could either be in a hard register or in memory.
2774 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2775 in memory and the hard register number if it is in a register.
2777 These macros should not be used in the case where a particular class of
2778 registers can only be copied to memory and not to another class of
2779 registers. In that case, secondary reload registers are not needed and
2780 would not be helpful. Instead, a stack location must be used to perform
2781 the copy and the @code{mov@var{m}} pattern should use memory as an
2782 intermediate storage. This case often occurs between floating-point and
2786 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2787 Certain machines have the property that some registers cannot be copied
2788 to some other registers without using memory. Define this macro on
2789 those machines to be a C expression that is nonzero if objects of mode
2790 @var{m} in registers of @var{class1} can only be copied to registers of
2791 class @var{class2} by storing a register of @var{class1} into memory
2792 and loading that memory location into a register of @var{class2}.
2794 Do not define this macro if its value would always be zero.
2797 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2798 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2799 allocates a stack slot for a memory location needed for register copies.
2800 If this macro is defined, the compiler instead uses the memory location
2801 defined by this macro.
2803 Do not define this macro if you do not define
2804 @code{SECONDARY_MEMORY_NEEDED}.
2807 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2808 When the compiler needs a secondary memory location to copy between two
2809 registers of mode @var{mode}, it normally allocates sufficient memory to
2810 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2811 load operations in a mode that many bits wide and whose class is the
2812 same as that of @var{mode}.
2814 This is right thing to do on most machines because it ensures that all
2815 bits of the register are copied and prevents accesses to the registers
2816 in a narrower mode, which some machines prohibit for floating-point
2819 However, this default behavior is not correct on some machines, such as
2820 the DEC Alpha, that store short integers in floating-point registers
2821 differently than in integer registers. On those machines, the default
2822 widening will not work correctly and you must define this macro to
2823 suppress that widening in some cases. See the file @file{alpha.h} for
2826 Do not define this macro if you do not define
2827 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2828 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2831 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2832 A target hook which returns @code{true} if pseudos that have been assigned
2833 to registers of class @var{rclass} would likely be spilled because
2834 registers of @var{rclass} are needed for spill registers.
2836 The default version of this target hook returns @code{true} if @var{rclass}
2837 has exactly one register and @code{false} otherwise. On most machines, this
2838 default should be used. Only use this target hook to some other expression
2839 if pseudos allocated by @file{local-alloc.c} end up in memory because their
2840 hard registers were needed for spill registers. If this target hook returns
2841 @code{false} for those classes, those pseudos will only be allocated by
2842 @file{global.c}, which knows how to reallocate the pseudo to another
2843 register. If there would not be another register available for reallocation,
2844 you should not change the implementation of this target hook since
2845 the only effect of such implementation would be to slow down register
2849 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2850 A C expression for the maximum number of consecutive registers
2851 of class @var{class} needed to hold a value of mode @var{mode}.
2853 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2854 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2855 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2856 @var{mode})} for all @var{regno} values in the class @var{class}.
2858 This macro helps control the handling of multiple-word values
2862 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2863 If defined, a C expression that returns nonzero for a @var{class} for which
2864 a change from mode @var{from} to mode @var{to} is invalid.
2866 For the example, loading 32-bit integer or floating-point objects into
2867 floating-point registers on the Alpha extends them to 64 bits.
2868 Therefore loading a 64-bit object and then storing it as a 32-bit object
2869 does not store the low-order 32 bits, as would be the case for a normal
2870 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2874 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2875 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2876 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2880 @node Old Constraints
2881 @section Obsolete Macros for Defining Constraints
2882 @cindex defining constraints, obsolete method
2883 @cindex constraints, defining, obsolete method
2885 Machine-specific constraints can be defined with these macros instead
2886 of the machine description constructs described in @ref{Define
2887 Constraints}. This mechanism is obsolete. New ports should not use
2888 it; old ports should convert to the new mechanism.
2890 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2891 For the constraint at the start of @var{str}, which starts with the letter
2892 @var{c}, return the length. This allows you to have register class /
2893 constant / extra constraints that are longer than a single letter;
2894 you don't need to define this macro if you can do with single-letter
2895 constraints only. The definition of this macro should use
2896 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2897 to handle specially.
2898 There are some sanity checks in genoutput.c that check the constraint lengths
2899 for the md file, so you can also use this macro to help you while you are
2900 transitioning from a byzantine single-letter-constraint scheme: when you
2901 return a negative length for a constraint you want to re-use, genoutput
2902 will complain about every instance where it is used in the md file.
2905 @defmac REG_CLASS_FROM_LETTER (@var{char})
2906 A C expression which defines the machine-dependent operand constraint
2907 letters for register classes. If @var{char} is such a letter, the
2908 value should be the register class corresponding to it. Otherwise,
2909 the value should be @code{NO_REGS}. The register letter @samp{r},
2910 corresponding to class @code{GENERAL_REGS}, will not be passed
2911 to this macro; you do not need to handle it.
2914 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2915 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2916 passed in @var{str}, so that you can use suffixes to distinguish between
2920 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2921 A C expression that defines the machine-dependent operand constraint
2922 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2923 particular ranges of integer values. If @var{c} is one of those
2924 letters, the expression should check that @var{value}, an integer, is in
2925 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2926 not one of those letters, the value should be 0 regardless of
2930 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2931 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2932 string passed in @var{str}, so that you can use suffixes to distinguish
2933 between different variants.
2936 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2937 A C expression that defines the machine-dependent operand constraint
2938 letters that specify particular ranges of @code{const_double} values
2939 (@samp{G} or @samp{H}).
2941 If @var{c} is one of those letters, the expression should check that
2942 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2943 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2944 letters, the value should be 0 regardless of @var{value}.
2946 @code{const_double} is used for all floating-point constants and for
2947 @code{DImode} fixed-point constants. A given letter can accept either
2948 or both kinds of values. It can use @code{GET_MODE} to distinguish
2949 between these kinds.
2952 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2953 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2954 string passed in @var{str}, so that you can use suffixes to distinguish
2955 between different variants.
2958 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2959 A C expression that defines the optional machine-dependent constraint
2960 letters that can be used to segregate specific types of operands, usually
2961 memory references, for the target machine. Any letter that is not
2962 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2963 @code{REG_CLASS_FROM_CONSTRAINT}
2964 may be used. Normally this macro will not be defined.
2966 If it is required for a particular target machine, it should return 1
2967 if @var{value} corresponds to the operand type represented by the
2968 constraint letter @var{c}. If @var{c} is not defined as an extra
2969 constraint, the value returned should be 0 regardless of @var{value}.
2971 For example, on the ROMP, load instructions cannot have their output
2972 in r0 if the memory reference contains a symbolic address. Constraint
2973 letter @samp{Q} is defined as representing a memory address that does
2974 @emph{not} contain a symbolic address. An alternative is specified with
2975 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2976 alternative specifies @samp{m} on the input and a register class that
2977 does not include r0 on the output.
2980 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2981 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2982 in @var{str}, so that you can use suffixes to distinguish between different
2986 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2987 A C expression that defines the optional machine-dependent constraint
2988 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2989 be treated like memory constraints by the reload pass.
2991 It should return 1 if the operand type represented by the constraint
2992 at the start of @var{str}, the first letter of which is the letter @var{c},
2993 comprises a subset of all memory references including
2994 all those whose address is simply a base register. This allows the reload
2995 pass to reload an operand, if it does not directly correspond to the operand
2996 type of @var{c}, by copying its address into a base register.
2998 For example, on the S/390, some instructions do not accept arbitrary
2999 memory references, but only those that do not make use of an index
3000 register. The constraint letter @samp{Q} is defined via
3001 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
3002 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
3003 a @samp{Q} constraint can handle any memory operand, because the
3004 reload pass knows it can be reloaded by copying the memory address
3005 into a base register if required. This is analogous to the way
3006 an @samp{o} constraint can handle any memory operand.
3009 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
3010 A C expression that defines the optional machine-dependent constraint
3011 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
3012 @code{EXTRA_CONSTRAINT_STR}, that should
3013 be treated like address constraints by the reload pass.
3015 It should return 1 if the operand type represented by the constraint
3016 at the start of @var{str}, which starts with the letter @var{c}, comprises
3017 a subset of all memory addresses including
3018 all those that consist of just a base register. This allows the reload
3019 pass to reload an operand, if it does not directly correspond to the operand
3020 type of @var{str}, by copying it into a base register.
3022 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3023 be used with the @code{address_operand} predicate. It is treated
3024 analogously to the @samp{p} constraint.
3027 @node Stack and Calling
3028 @section Stack Layout and Calling Conventions
3029 @cindex calling conventions
3031 @c prevent bad page break with this line
3032 This describes the stack layout and calling conventions.
3036 * Exception Handling::
3041 * Register Arguments::
3043 * Aggregate Return::
3048 * Stack Smashing Protection::
3052 @subsection Basic Stack Layout
3053 @cindex stack frame layout
3054 @cindex frame layout
3056 @c prevent bad page break with this line
3057 Here is the basic stack layout.
3059 @defmac STACK_GROWS_DOWNWARD
3060 Define this macro if pushing a word onto the stack moves the stack
3061 pointer to a smaller address.
3063 When we say, ``define this macro if @dots{}'', it means that the
3064 compiler checks this macro only with @code{#ifdef} so the precise
3065 definition used does not matter.
3068 @defmac STACK_PUSH_CODE
3069 This macro defines the operation used when something is pushed
3070 on the stack. In RTL, a push operation will be
3071 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3073 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3074 and @code{POST_INC}. Which of these is correct depends on
3075 the stack direction and on whether the stack pointer points
3076 to the last item on the stack or whether it points to the
3077 space for the next item on the stack.
3079 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3080 defined, which is almost always right, and @code{PRE_INC} otherwise,
3081 which is often wrong.
3084 @defmac FRAME_GROWS_DOWNWARD
3085 Define this macro to nonzero value if the addresses of local variable slots
3086 are at negative offsets from the frame pointer.
3089 @defmac ARGS_GROW_DOWNWARD
3090 Define this macro if successive arguments to a function occupy decreasing
3091 addresses on the stack.
3094 @defmac STARTING_FRAME_OFFSET
3095 Offset from the frame pointer to the first local variable slot to be allocated.
3097 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3098 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3099 Otherwise, it is found by adding the length of the first slot to the
3100 value @code{STARTING_FRAME_OFFSET}.
3101 @c i'm not sure if the above is still correct.. had to change it to get
3102 @c rid of an overfull. --mew 2feb93
3105 @defmac STACK_ALIGNMENT_NEEDED
3106 Define to zero to disable final alignment of the stack during reload.
3107 The nonzero default for this macro is suitable for most ports.
3109 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3110 is a register save block following the local block that doesn't require
3111 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3112 stack alignment and do it in the backend.
3115 @defmac STACK_POINTER_OFFSET
3116 Offset from the stack pointer register to the first location at which
3117 outgoing arguments are placed. If not specified, the default value of
3118 zero is used. This is the proper value for most machines.
3120 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3121 the first location at which outgoing arguments are placed.
3124 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3125 Offset from the argument pointer register to the first argument's
3126 address. On some machines it may depend on the data type of the
3129 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3130 the first argument's address.
3133 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3134 Offset from the stack pointer register to an item dynamically allocated
3135 on the stack, e.g., by @code{alloca}.
3137 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3138 length of the outgoing arguments. The default is correct for most
3139 machines. See @file{function.c} for details.
3142 @defmac INITIAL_FRAME_ADDRESS_RTX
3143 A C expression whose value is RTL representing the address of the initial
3144 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3145 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3146 default value will be used. Define this macro in order to make frame pointer
3147 elimination work in the presence of @code{__builtin_frame_address (count)} and
3148 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3151 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3152 A C expression whose value is RTL representing the address in a stack
3153 frame where the pointer to the caller's frame is stored. Assume that
3154 @var{frameaddr} is an RTL expression for the address of the stack frame
3157 If you don't define this macro, the default is to return the value
3158 of @var{frameaddr}---that is, the stack frame address is also the
3159 address of the stack word that points to the previous frame.
3162 @defmac SETUP_FRAME_ADDRESSES
3163 If defined, a C expression that produces the machine-specific code to
3164 setup the stack so that arbitrary frames can be accessed. For example,
3165 on the SPARC, we must flush all of the register windows to the stack
3166 before we can access arbitrary stack frames. You will seldom need to
3170 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3171 This target hook should return an rtx that is used to store
3172 the address of the current frame into the built in @code{setjmp} buffer.
3173 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3174 machines. One reason you may need to define this target hook is if
3175 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3178 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3179 A C expression whose value is RTL representing the value of the frame
3180 address for the current frame. @var{frameaddr} is the frame pointer
3181 of the current frame. This is used for __builtin_frame_address.
3182 You need only define this macro if the frame address is not the same
3183 as the frame pointer. Most machines do not need to define it.
3186 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3187 A C expression whose value is RTL representing the value of the return
3188 address for the frame @var{count} steps up from the current frame, after
3189 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3190 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3191 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3193 The value of the expression must always be the correct address when
3194 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3195 determine the return address of other frames.
3198 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3199 Define this if the return address of a particular stack frame is accessed
3200 from the frame pointer of the previous stack frame.
3203 @defmac INCOMING_RETURN_ADDR_RTX
3204 A C expression whose value is RTL representing the location of the
3205 incoming return address at the beginning of any function, before the
3206 prologue. This RTL is either a @code{REG}, indicating that the return
3207 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3210 You only need to define this macro if you want to support call frame
3211 debugging information like that provided by DWARF 2.
3213 If this RTL is a @code{REG}, you should also define
3214 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3217 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3218 A C expression whose value is an integer giving a DWARF 2 column
3219 number that may be used as an alternative return column. The column
3220 must not correspond to any gcc hard register (that is, it must not
3221 be in the range of @code{DWARF_FRAME_REGNUM}).
3223 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3224 general register, but an alternative column needs to be used for signal
3225 frames. Some targets have also used different frame return columns
3229 @defmac DWARF_ZERO_REG
3230 A C expression whose value is an integer giving a DWARF 2 register
3231 number that is considered to always have the value zero. This should
3232 only be defined if the target has an architected zero register, and
3233 someone decided it was a good idea to use that register number to
3234 terminate the stack backtrace. New ports should avoid this.
3237 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3238 This target hook allows the backend to emit frame-related insns that
3239 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3240 info engine will invoke it on insns of the form
3242 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3246 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3248 to let the backend emit the call frame instructions. @var{label} is
3249 the CFI label attached to the insn, @var{pattern} is the pattern of
3250 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3253 @defmac INCOMING_FRAME_SP_OFFSET
3254 A C expression whose value is an integer giving the offset, in bytes,
3255 from the value of the stack pointer register to the top of the stack
3256 frame at the beginning of any function, before the prologue. The top of
3257 the frame is defined to be the value of the stack pointer in the
3258 previous frame, just before the call instruction.
3260 You only need to define this macro if you want to support call frame
3261 debugging information like that provided by DWARF 2.
3264 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3265 A C expression whose value is an integer giving the offset, in bytes,
3266 from the argument pointer to the canonical frame address (cfa). The
3267 final value should coincide with that calculated by
3268 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3269 during virtual register instantiation.
3271 The default value for this macro is
3272 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3273 which is correct for most machines; in general, the arguments are found
3274 immediately before the stack frame. Note that this is not the case on
3275 some targets that save registers into the caller's frame, such as SPARC
3276 and rs6000, and so such targets need to define this macro.
3278 You only need to define this macro if the default is incorrect, and you
3279 want to support call frame debugging information like that provided by
3283 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3284 If defined, a C expression whose value is an integer giving the offset
3285 in bytes from the frame pointer to the canonical frame address (cfa).
3286 The final value should coincide with that calculated by
3287 @code{INCOMING_FRAME_SP_OFFSET}.
3289 Normally the CFA is calculated as an offset from the argument pointer,
3290 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3291 variable due to the ABI, this may not be possible. If this macro is
3292 defined, it implies that the virtual register instantiation should be
3293 based on the frame pointer instead of the argument pointer. Only one
3294 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3298 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3299 If defined, a C expression whose value is an integer giving the offset
3300 in bytes from the canonical frame address (cfa) to the frame base used
3301 in DWARF 2 debug information. The default is zero. A different value
3302 may reduce the size of debug information on some ports.
3305 @node Exception Handling
3306 @subsection Exception Handling Support
3307 @cindex exception handling
3309 @defmac EH_RETURN_DATA_REGNO (@var{N})
3310 A C expression whose value is the @var{N}th register number used for
3311 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3312 @var{N} registers are usable.
3314 The exception handling library routines communicate with the exception
3315 handlers via a set of agreed upon registers. Ideally these registers
3316 should be call-clobbered; it is possible to use call-saved registers,
3317 but may negatively impact code size. The target must support at least
3318 2 data registers, but should define 4 if there are enough free registers.
3320 You must define this macro if you want to support call frame exception
3321 handling like that provided by DWARF 2.
3324 @defmac EH_RETURN_STACKADJ_RTX
3325 A C expression whose value is RTL representing a location in which
3326 to store a stack adjustment to be applied before function return.
3327 This is used to unwind the stack to an exception handler's call frame.
3328 It will be assigned zero on code paths that return normally.
3330 Typically this is a call-clobbered hard register that is otherwise
3331 untouched by the epilogue, but could also be a stack slot.
3333 Do not define this macro if the stack pointer is saved and restored
3334 by the regular prolog and epilog code in the call frame itself; in
3335 this case, the exception handling library routines will update the
3336 stack location to be restored in place. Otherwise, you must define
3337 this macro if you want to support call frame exception handling like
3338 that provided by DWARF 2.
3341 @defmac EH_RETURN_HANDLER_RTX
3342 A C expression whose value is RTL representing a location in which
3343 to store the address of an exception handler to which we should
3344 return. It will not be assigned on code paths that return normally.
3346 Typically this is the location in the call frame at which the normal
3347 return address is stored. For targets that return by popping an
3348 address off the stack, this might be a memory address just below
3349 the @emph{target} call frame rather than inside the current call
3350 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3351 been assigned, so it may be used to calculate the location of the
3354 Some targets have more complex requirements than storing to an
3355 address calculable during initial code generation. In that case
3356 the @code{eh_return} instruction pattern should be used instead.
3358 If you want to support call frame exception handling, you must
3359 define either this macro or the @code{eh_return} instruction pattern.
3362 @defmac RETURN_ADDR_OFFSET
3363 If defined, an integer-valued C expression for which rtl will be generated
3364 to add it to the exception handler address before it is searched in the
3365 exception handling tables, and to subtract it again from the address before
3366 using it to return to the exception handler.
3369 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3370 This macro chooses the encoding of pointers embedded in the exception
3371 handling sections. If at all possible, this should be defined such
3372 that the exception handling section will not require dynamic relocations,
3373 and so may be read-only.
3375 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3376 @var{global} is true if the symbol may be affected by dynamic relocations.
3377 The macro should return a combination of the @code{DW_EH_PE_*} defines
3378 as found in @file{dwarf2.h}.
3380 If this macro is not defined, pointers will not be encoded but
3381 represented directly.
3384 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3385 This macro allows the target to emit whatever special magic is required
3386 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3387 Generic code takes care of pc-relative and indirect encodings; this must
3388 be defined if the target uses text-relative or data-relative encodings.
3390 This is a C statement that branches to @var{done} if the format was
3391 handled. @var{encoding} is the format chosen, @var{size} is the number
3392 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3396 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3397 This macro allows the target to add CPU and operating system specific
3398 code to the call-frame unwinder for use when there is no unwind data
3399 available. The most common reason to implement this macro is to unwind
3400 through signal frames.
3402 This macro is called from @code{uw_frame_state_for} in
3403 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3404 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3405 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3406 for the address of the code being executed and @code{context->cfa} for
3407 the stack pointer value. If the frame can be decoded, the register
3408 save addresses should be updated in @var{fs} and the macro should
3409 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3410 the macro should evaluate to @code{_URC_END_OF_STACK}.
3412 For proper signal handling in Java this macro is accompanied by
3413 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3416 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3417 This macro allows the target to add operating system specific code to the
3418 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3419 usually used for signal or interrupt frames.
3421 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3422 @var{context} is an @code{_Unwind_Context};
3423 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3424 for the abi and context in the @code{.unwabi} directive. If the
3425 @code{.unwabi} directive can be handled, the register save addresses should
3426 be updated in @var{fs}.
3429 @defmac TARGET_USES_WEAK_UNWIND_INFO
3430 A C expression that evaluates to true if the target requires unwind
3431 info to be given comdat linkage. Define it to be @code{1} if comdat
3432 linkage is necessary. The default is @code{0}.
3435 @node Stack Checking
3436 @subsection Specifying How Stack Checking is Done
3438 GCC will check that stack references are within the boundaries of the
3439 stack, if the option @option{-fstack-check} is specified, in one of
3444 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3445 will assume that you have arranged for full stack checking to be done
3446 at appropriate places in the configuration files. GCC will not do
3447 other special processing.
3450 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3451 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3452 that you have arranged for static stack checking (checking of the
3453 static stack frame of functions) to be done at appropriate places
3454 in the configuration files. GCC will only emit code to do dynamic
3455 stack checking (checking on dynamic stack allocations) using the third
3459 If neither of the above are true, GCC will generate code to periodically
3460 ``probe'' the stack pointer using the values of the macros defined below.
3463 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3464 GCC will change its allocation strategy for large objects if the option
3465 @option{-fstack-check} is specified: they will always be allocated
3466 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3468 @defmac STACK_CHECK_BUILTIN
3469 A nonzero value if stack checking is done by the configuration files in a
3470 machine-dependent manner. You should define this macro if stack checking
3471 is required by the ABI of your machine or if you would like to do stack
3472 checking in some more efficient way than the generic approach. The default
3473 value of this macro is zero.
3476 @defmac STACK_CHECK_STATIC_BUILTIN
3477 A nonzero value if static stack checking is done by the configuration files
3478 in a machine-dependent manner. You should define this macro if you would
3479 like to do static stack checking in some more efficient way than the generic
3480 approach. The default value of this macro is zero.
3483 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3484 An integer specifying the interval at which GCC must generate stack probe
3485 instructions, defined as 2 raised to this integer. You will normally
3486 define this macro so that the interval be no larger than the size of
3487 the ``guard pages'' at the end of a stack area. The default value
3488 of 12 (4096-byte interval) is suitable for most systems.
3491 @defmac STACK_CHECK_MOVING_SP
3492 An integer which is nonzero if GCC should move the stack pointer page by page
3493 when doing probes. This can be necessary on systems where the stack pointer
3494 contains the bottom address of the memory area accessible to the executing
3495 thread at any point in time. In this situation an alternate signal stack
3496 is required in order to be able to recover from a stack overflow. The
3497 default value of this macro is zero.
3500 @defmac STACK_CHECK_PROTECT
3501 The number of bytes of stack needed to recover from a stack overflow, for
3502 languages where such a recovery is supported. The default value of 75 words
3503 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3504 8192 bytes with other exception handling mechanisms should be adequate for
3508 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3509 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3510 in the opposite case.
3512 @defmac STACK_CHECK_MAX_FRAME_SIZE
3513 The maximum size of a stack frame, in bytes. GCC will generate probe
3514 instructions in non-leaf functions to ensure at least this many bytes of
3515 stack are available. If a stack frame is larger than this size, stack
3516 checking will not be reliable and GCC will issue a warning. The
3517 default is chosen so that GCC only generates one instruction on most
3518 systems. You should normally not change the default value of this macro.
3521 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3522 GCC uses this value to generate the above warning message. It
3523 represents the amount of fixed frame used by a function, not including
3524 space for any callee-saved registers, temporaries and user variables.
3525 You need only specify an upper bound for this amount and will normally
3526 use the default of four words.
3529 @defmac STACK_CHECK_MAX_VAR_SIZE
3530 The maximum size, in bytes, of an object that GCC will place in the
3531 fixed area of the stack frame when the user specifies
3532 @option{-fstack-check}.
3533 GCC computed the default from the values of the above macros and you will
3534 normally not need to override that default.
3538 @node Frame Registers
3539 @subsection Registers That Address the Stack Frame
3541 @c prevent bad page break with this line
3542 This discusses registers that address the stack frame.
3544 @defmac STACK_POINTER_REGNUM
3545 The register number of the stack pointer register, which must also be a
3546 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3547 the hardware determines which register this is.
3550 @defmac FRAME_POINTER_REGNUM
3551 The register number of the frame pointer register, which is used to
3552 access automatic variables in the stack frame. On some machines, the
3553 hardware determines which register this is. On other machines, you can
3554 choose any register you wish for this purpose.
3557 @defmac HARD_FRAME_POINTER_REGNUM
3558 On some machines the offset between the frame pointer and starting
3559 offset of the automatic variables is not known until after register
3560 allocation has been done (for example, because the saved registers are
3561 between these two locations). On those machines, define
3562 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3563 be used internally until the offset is known, and define
3564 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3565 used for the frame pointer.
3567 You should define this macro only in the very rare circumstances when it
3568 is not possible to calculate the offset between the frame pointer and
3569 the automatic variables until after register allocation has been
3570 completed. When this macro is defined, you must also indicate in your
3571 definition of @code{ELIMINABLE_REGS} how to eliminate
3572 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3573 or @code{STACK_POINTER_REGNUM}.
3575 Do not define this macro if it would be the same as
3576 @code{FRAME_POINTER_REGNUM}.
3579 @defmac ARG_POINTER_REGNUM
3580 The register number of the arg pointer register, which is used to access
3581 the function's argument list. On some machines, this is the same as the
3582 frame pointer register. On some machines, the hardware determines which
3583 register this is. On other machines, you can choose any register you
3584 wish for this purpose. If this is not the same register as the frame
3585 pointer register, then you must mark it as a fixed register according to
3586 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3587 (@pxref{Elimination}).
3590 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3591 Define this to a preprocessor constant that is nonzero if
3592 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3593 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3594 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3595 definition is not suitable for use in preprocessor conditionals.
3598 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3599 Define this to a preprocessor constant that is nonzero if
3600 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3601 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3602 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3603 definition is not suitable for use in preprocessor conditionals.
3606 @defmac RETURN_ADDRESS_POINTER_REGNUM
3607 The register number of the return address pointer register, which is used to
3608 access the current function's return address from the stack. On some
3609 machines, the return address is not at a fixed offset from the frame
3610 pointer or stack pointer or argument pointer. This register can be defined
3611 to point to the return address on the stack, and then be converted by
3612 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3614 Do not define this macro unless there is no other way to get the return
3615 address from the stack.
3618 @defmac STATIC_CHAIN_REGNUM
3619 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3620 Register numbers used for passing a function's static chain pointer. If
3621 register windows are used, the register number as seen by the called
3622 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3623 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3624 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3627 The static chain register need not be a fixed register.
3629 If the static chain is passed in memory, these macros should not be
3630 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3633 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl}, bool @var{incoming_p})
3634 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3635 targets that may use different static chain locations for different
3636 nested functions. This may be required if the target has function
3637 attributes that affect the calling conventions of the function and
3638 those calling conventions use different static chain locations.
3640 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3642 If the static chain is passed in memory, this hook should be used to
3643 provide rtx giving @code{mem} expressions that denote where they are stored.
3644 Often the @code{mem} expression as seen by the caller will be at an offset
3645 from the stack pointer and the @code{mem} expression as seen by the callee
3646 will be at an offset from the frame pointer.
3647 @findex stack_pointer_rtx
3648 @findex frame_pointer_rtx
3649 @findex arg_pointer_rtx
3650 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3651 @code{arg_pointer_rtx} will have been initialized and should be used
3652 to refer to those items.
3655 @defmac DWARF_FRAME_REGISTERS
3656 This macro specifies the maximum number of hard registers that can be
3657 saved in a call frame. This is used to size data structures used in
3658 DWARF2 exception handling.
3660 Prior to GCC 3.0, this macro was needed in order to establish a stable
3661 exception handling ABI in the face of adding new hard registers for ISA
3662 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3663 in the number of hard registers. Nevertheless, this macro can still be
3664 used to reduce the runtime memory requirements of the exception handling
3665 routines, which can be substantial if the ISA contains a lot of
3666 registers that are not call-saved.
3668 If this macro is not defined, it defaults to
3669 @code{FIRST_PSEUDO_REGISTER}.
3672 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3674 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3675 for backward compatibility in pre GCC 3.0 compiled code.
3677 If this macro is not defined, it defaults to
3678 @code{DWARF_FRAME_REGISTERS}.
3681 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3683 Define this macro if the target's representation for dwarf registers
3684 is different than the internal representation for unwind column.
3685 Given a dwarf register, this macro should return the internal unwind
3686 column number to use instead.
3688 See the PowerPC's SPE target for an example.
3691 @defmac DWARF_FRAME_REGNUM (@var{regno})
3693 Define this macro if the target's representation for dwarf registers
3694 used in .eh_frame or .debug_frame is different from that used in other
3695 debug info sections. Given a GCC hard register number, this macro
3696 should return the .eh_frame register number. The default is
3697 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3701 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3703 Define this macro to map register numbers held in the call frame info
3704 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3705 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3706 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3707 return @code{@var{regno}}.
3712 @subsection Eliminating Frame Pointer and Arg Pointer
3714 @c prevent bad page break with this line
3715 This is about eliminating the frame pointer and arg pointer.
3717 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3718 This target hook should return @code{true} if a function must have and use
3719 a frame pointer. This target hook is called in the reload pass. If its return
3720 value is @code{true} the function will have a frame pointer.
3722 This target hook can in principle examine the current function and decide
3723 according to the facts, but on most machines the constant @code{false} or the
3724 constant @code{true} suffices. Use @code{false} when the machine allows code
3725 to be generated with no frame pointer, and doing so saves some time or space.
3726 Use @code{true} when there is no possible advantage to avoiding a frame
3729 In certain cases, the compiler does not know how to produce valid code
3730 without a frame pointer. The compiler recognizes those cases and
3731 automatically gives the function a frame pointer regardless of what
3732 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3735 In a function that does not require a frame pointer, the frame pointer
3736 register can be allocated for ordinary usage, unless you mark it as a
3737 fixed register. See @code{FIXED_REGISTERS} for more information.
3739 Default return value is @code{false}.
3742 @findex get_frame_size
3743 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3744 A C statement to store in the variable @var{depth-var} the difference
3745 between the frame pointer and the stack pointer values immediately after
3746 the function prologue. The value would be computed from information
3747 such as the result of @code{get_frame_size ()} and the tables of
3748 registers @code{regs_ever_live} and @code{call_used_regs}.
3750 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3751 need not be defined. Otherwise, it must be defined even if
3752 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3753 case, you may set @var{depth-var} to anything.
3756 @defmac ELIMINABLE_REGS
3757 If defined, this macro specifies a table of register pairs used to
3758 eliminate unneeded registers that point into the stack frame. If it is not
3759 defined, the only elimination attempted by the compiler is to replace
3760 references to the frame pointer with references to the stack pointer.
3762 The definition of this macro is a list of structure initializations, each
3763 of which specifies an original and replacement register.
3765 On some machines, the position of the argument pointer is not known until
3766 the compilation is completed. In such a case, a separate hard register
3767 must be used for the argument pointer. This register can be eliminated by
3768 replacing it with either the frame pointer or the argument pointer,
3769 depending on whether or not the frame pointer has been eliminated.
3771 In this case, you might specify:
3773 #define ELIMINABLE_REGS \
3774 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3775 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3776 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3779 Note that the elimination of the argument pointer with the stack pointer is
3780 specified first since that is the preferred elimination.
3783 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3784 This target hook should returns @code{true} if the compiler is allowed to
3785 try to replace register number @var{from_reg} with register number
3786 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3787 is defined, and will usually be @code{true}, since most of the cases
3788 preventing register elimination are things that the compiler already
3791 Default return value is @code{true}.
3794 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3795 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3796 specifies the initial difference between the specified pair of
3797 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3801 @node Stack Arguments
3802 @subsection Passing Function Arguments on the Stack
3803 @cindex arguments on stack
3804 @cindex stack arguments
3806 The macros in this section control how arguments are passed
3807 on the stack. See the following section for other macros that
3808 control passing certain arguments in registers.
3810 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3811 This target hook returns @code{true} if an argument declared in a
3812 prototype as an integral type smaller than @code{int} should actually be
3813 passed as an @code{int}. In addition to avoiding errors in certain
3814 cases of mismatch, it also makes for better code on certain machines.
3815 The default is to not promote prototypes.
3819 A C expression. If nonzero, push insns will be used to pass
3821 If the target machine does not have a push instruction, set it to zero.
3822 That directs GCC to use an alternate strategy: to
3823 allocate the entire argument block and then store the arguments into
3824 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3827 @defmac PUSH_ARGS_REVERSED
3828 A C expression. If nonzero, function arguments will be evaluated from
3829 last to first, rather than from first to last. If this macro is not
3830 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3831 and args grow in opposite directions, and 0 otherwise.
3834 @defmac PUSH_ROUNDING (@var{npushed})
3835 A C expression that is the number of bytes actually pushed onto the
3836 stack when an instruction attempts to push @var{npushed} bytes.
3838 On some machines, the definition
3841 #define PUSH_ROUNDING(BYTES) (BYTES)
3845 will suffice. But on other machines, instructions that appear
3846 to push one byte actually push two bytes in an attempt to maintain
3847 alignment. Then the definition should be
3850 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3853 If the value of this macro has a type, it should be an unsigned type.
3856 @findex current_function_outgoing_args_size
3857 @defmac ACCUMULATE_OUTGOING_ARGS
3858 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3859 will be computed and placed into the variable
3860 @code{current_function_outgoing_args_size}. No space will be pushed
3861 onto the stack for each call; instead, the function prologue should
3862 increase the stack frame size by this amount.
3864 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3868 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3869 Define this macro if functions should assume that stack space has been
3870 allocated for arguments even when their values are passed in
3873 The value of this macro is the size, in bytes, of the area reserved for
3874 arguments passed in registers for the function represented by @var{fndecl},
3875 which can be zero if GCC is calling a library function.
3876 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3879 This space can be allocated by the caller, or be a part of the
3880 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3883 @c above is overfull. not sure what to do. --mew 5feb93 did
3884 @c something, not sure if it looks good. --mew 10feb93
3886 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3887 Define this to a nonzero value if it is the responsibility of the
3888 caller to allocate the area reserved for arguments passed in registers
3889 when calling a function of @var{fntype}. @var{fntype} may be NULL
3890 if the function called is a library function.
3892 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3893 whether the space for these arguments counts in the value of
3894 @code{current_function_outgoing_args_size}.
3897 @defmac STACK_PARMS_IN_REG_PARM_AREA
3898 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3899 stack parameters don't skip the area specified by it.
3900 @c i changed this, makes more sens and it should have taken care of the
3901 @c overfull.. not as specific, tho. --mew 5feb93
3903 Normally, when a parameter is not passed in registers, it is placed on the
3904 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3905 suppresses this behavior and causes the parameter to be passed on the
3906 stack in its natural location.
3909 @deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size})
3910 This target hook returns the number of bytes of its own arguments that
3911 a function pops on returning, or 0 if the function pops no arguments
3912 and the caller must therefore pop them all after the function returns.
3914 @var{fundecl} is a C variable whose value is a tree node that describes
3915 the function in question. Normally it is a node of type
3916 @code{FUNCTION_DECL} that describes the declaration of the function.
3917 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3919 @var{funtype} is a C variable whose value is a tree node that
3920 describes the function in question. Normally it is a node of type
3921 @code{FUNCTION_TYPE} that describes the data type of the function.
3922 From this it is possible to obtain the data types of the value and
3923 arguments (if known).
3925 When a call to a library function is being considered, @var{fundecl}
3926 will contain an identifier node for the library function. Thus, if
3927 you need to distinguish among various library functions, you can do so
3928 by their names. Note that ``library function'' in this context means
3929 a function used to perform arithmetic, whose name is known specially
3930 in the compiler and was not mentioned in the C code being compiled.
3932 @var{size} is the number of bytes of arguments passed on the
3933 stack. If a variable number of bytes is passed, it is zero, and
3934 argument popping will always be the responsibility of the calling function.
3936 On the VAX, all functions always pop their arguments, so the definition
3937 of this macro is @var{size}. On the 68000, using the standard
3938 calling convention, no functions pop their arguments, so the value of
3939 the macro is always 0 in this case. But an alternative calling
3940 convention is available in which functions that take a fixed number of
3941 arguments pop them but other functions (such as @code{printf}) pop
3942 nothing (the caller pops all). When this convention is in use,
3943 @var{funtype} is examined to determine whether a function takes a fixed
3944 number of arguments.
3947 @defmac CALL_POPS_ARGS (@var{cum})
3948 A C expression that should indicate the number of bytes a call sequence
3949 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3950 when compiling a function call.
3952 @var{cum} is the variable in which all arguments to the called function
3953 have been accumulated.
3955 On certain architectures, such as the SH5, a call trampoline is used
3956 that pops certain registers off the stack, depending on the arguments
3957 that have been passed to the function. Since this is a property of the
3958 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3962 @node Register Arguments
3963 @subsection Passing Arguments in Registers
3964 @cindex arguments in registers
3965 @cindex registers arguments
3967 This section describes the macros which let you control how various
3968 types of arguments are passed in registers or how they are arranged in
3971 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3972 Return an RTX indicating whether a function argument is passed in a
3973 register and if so, which register.
3975 The arguments are @var{ca}, which summarizes all the previous
3976 arguments; @var{mode}, the machine mode of the argument; @var{type},
3977 the data type of the argument as a tree node or 0 if that is not known
3978 (which happens for C support library functions); and @var{named},
3979 which is @code{true} for an ordinary argument and @code{false} for
3980 nameless arguments that correspond to @samp{@dots{}} in the called
3981 function's prototype. @var{type} can be an incomplete type if a
3982 syntax error has previously occurred.
3984 The return value is usually either a @code{reg} RTX for the hard
3985 register in which to pass the argument, or zero to pass the argument
3988 The value of the expression can also be a @code{parallel} RTX@. This is
3989 used when an argument is passed in multiple locations. The mode of the
3990 @code{parallel} should be the mode of the entire argument. The
3991 @code{parallel} holds any number of @code{expr_list} pairs; each one
3992 describes where part of the argument is passed. In each
3993 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3994 register in which to pass this part of the argument, and the mode of the
3995 register RTX indicates how large this part of the argument is. The
3996 second operand of the @code{expr_list} is a @code{const_int} which gives
3997 the offset in bytes into the entire argument of where this part starts.
3998 As a special exception the first @code{expr_list} in the @code{parallel}
3999 RTX may have a first operand of zero. This indicates that the entire
4000 argument is also stored on the stack.
4002 The last time this hook is called, it is called with @code{MODE ==
4003 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
4004 pattern as operands 2 and 3 respectively.
4006 @cindex @file{stdarg.h} and register arguments
4007 The usual way to make the ISO library @file{stdarg.h} work on a
4008 machine where some arguments are usually passed in registers, is to
4009 cause nameless arguments to be passed on the stack instead. This is
4010 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
4011 @var{named} is @code{false}.
4013 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
4014 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
4015 You may use the hook @code{targetm.calls.must_pass_in_stack}
4016 in the definition of this macro to determine if this argument is of a
4017 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4018 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
4019 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4020 defined, the argument will be computed in the stack and then loaded into
4024 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, const_tree @var{type})
4025 This target hook should return @code{true} if we should not pass @var{type}
4026 solely in registers. The file @file{expr.h} defines a
4027 definition that is usually appropriate, refer to @file{expr.h} for additional
4031 @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})
4032 Define this hook if the target machine has ``register windows'', so
4033 that the register in which a function sees an arguments is not
4034 necessarily the same as the one in which the caller passed the
4037 For such machines, @code{TARGET_FUNCTION_ARG} computes the register in
4038 which the caller passes the value, and
4039 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4040 fashion to tell the function being called where the arguments will
4043 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4044 @code{TARGET_FUNCTION_ARG} serves both purposes.
4047 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4048 This target hook returns the number of bytes at the beginning of an
4049 argument that must be put in registers. The value must be zero for
4050 arguments that are passed entirely in registers or that are entirely
4051 pushed on the stack.
4053 On some machines, certain arguments must be passed partially in
4054 registers and partially in memory. On these machines, typically the
4055 first few words of arguments are passed in registers, and the rest
4056 on the stack. If a multi-word argument (a @code{double} or a
4057 structure) crosses that boundary, its first few words must be passed
4058 in registers and the rest must be pushed. This macro tells the
4059 compiler when this occurs, and how many bytes should go in registers.
4061 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4062 register to be used by the caller for this argument; likewise
4063 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4066 @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})
4067 This target hook should return @code{true} if an argument at the
4068 position indicated by @var{cum} should be passed by reference. This
4069 predicate is queried after target independent reasons for being
4070 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4072 If the hook returns true, a copy of that argument is made in memory and a
4073 pointer to the argument is passed instead of the argument itself.
4074 The pointer is passed in whatever way is appropriate for passing a pointer
4078 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4079 The function argument described by the parameters to this hook is
4080 known to be passed by reference. The hook should return true if the
4081 function argument should be copied by the callee instead of copied
4084 For any argument for which the hook returns true, if it can be
4085 determined that the argument is not modified, then a copy need
4088 The default version of this hook always returns false.
4091 @defmac CUMULATIVE_ARGS
4092 A C type for declaring a variable that is used as the first argument
4093 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4094 target machines, the type @code{int} suffices and can hold the number
4095 of bytes of argument so far.
4097 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4098 arguments that have been passed on the stack. The compiler has other
4099 variables to keep track of that. For target machines on which all
4100 arguments are passed on the stack, there is no need to store anything in
4101 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4102 should not be empty, so use @code{int}.
4105 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4106 If defined, this macro is called before generating any code for a
4107 function, but after the @var{cfun} descriptor for the function has been
4108 created. The back end may use this macro to update @var{cfun} to
4109 reflect an ABI other than that which would normally be used by default.
4110 If the compiler is generating code for a compiler-generated function,
4111 @var{fndecl} may be @code{NULL}.
4114 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4115 A C statement (sans semicolon) for initializing the variable
4116 @var{cum} for the state at the beginning of the argument list. The
4117 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4118 is the tree node for the data type of the function which will receive
4119 the args, or 0 if the args are to a compiler support library function.
4120 For direct calls that are not libcalls, @var{fndecl} contain the
4121 declaration node of the function. @var{fndecl} is also set when
4122 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4123 being compiled. @var{n_named_args} is set to the number of named
4124 arguments, including a structure return address if it is passed as a
4125 parameter, when making a call. When processing incoming arguments,
4126 @var{n_named_args} is set to @minus{}1.
4128 When processing a call to a compiler support library function,
4129 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4130 contains the name of the function, as a string. @var{libname} is 0 when
4131 an ordinary C function call is being processed. Thus, each time this
4132 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4133 never both of them at once.
4136 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4137 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4138 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4139 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4140 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4141 0)} is used instead.
4144 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4145 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4146 finding the arguments for the function being compiled. If this macro is
4147 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4149 The value passed for @var{libname} is always 0, since library routines
4150 with special calling conventions are never compiled with GCC@. The
4151 argument @var{libname} exists for symmetry with
4152 @code{INIT_CUMULATIVE_ARGS}.
4153 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4154 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4157 @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})
4158 This hook updates the summarizer variable pointed to by @var{ca} to
4159 advance past an argument in the argument list. The values @var{mode},
4160 @var{type} and @var{named} describe that argument. Once this is done,
4161 the variable @var{cum} is suitable for analyzing the @emph{following}
4162 argument with @code{TARGET_FUNCTION_ARG}, etc.
4164 This hook need not do anything if the argument in question was passed
4165 on the stack. The compiler knows how to track the amount of stack space
4166 used for arguments without any special help.
4169 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4170 If defined, a C expression that is the number of bytes to add to the
4171 offset of the argument passed in memory. This is needed for the SPU,
4172 which passes @code{char} and @code{short} arguments in the preferred
4173 slot that is in the middle of the quad word instead of starting at the
4177 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4178 If defined, a C expression which determines whether, and in which direction,
4179 to pad out an argument with extra space. The value should be of type
4180 @code{enum direction}: either @code{upward} to pad above the argument,
4181 @code{downward} to pad below, or @code{none} to inhibit padding.
4183 The @emph{amount} of padding is not controlled by this macro, but by the
4184 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
4185 always just enough to reach the next multiple of that boundary.
4187 This macro has a default definition which is right for most systems.
4188 For little-endian machines, the default is to pad upward. For
4189 big-endian machines, the default is to pad downward for an argument of
4190 constant size shorter than an @code{int}, and upward otherwise.
4193 @defmac PAD_VARARGS_DOWN
4194 If defined, a C expression which determines whether the default
4195 implementation of va_arg will attempt to pad down before reading the
4196 next argument, if that argument is smaller than its aligned space as
4197 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4198 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4201 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4202 Specify padding for the last element of a block move between registers and
4203 memory. @var{first} is nonzero if this is the only element. Defining this
4204 macro allows better control of register function parameters on big-endian
4205 machines, without using @code{PARALLEL} rtl. In particular,
4206 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4207 registers, as there is no longer a "wrong" part of a register; For example,
4208 a three byte aggregate may be passed in the high part of a register if so
4212 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4213 This hook returns the alignment boundary, in bits, of an argument
4214 with the specified mode and type. The default hook returns
4215 @code{PARM_BOUNDARY} for all arguments.
4218 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_ROUND_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4219 Normally, the size of an argument is rounded up to @code{PARM_BOUNDARY},
4220 which is the default value for this hook. You can define this hook to
4221 return a different value if an argument size must be rounded to a larger
4225 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4226 A C expression that is nonzero if @var{regno} is the number of a hard
4227 register in which function arguments are sometimes passed. This does
4228 @emph{not} include implicit arguments such as the static chain and
4229 the structure-value address. On many machines, no registers can be
4230 used for this purpose since all function arguments are pushed on the
4234 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4235 This hook should return true if parameter of type @var{type} are passed
4236 as two scalar parameters. By default, GCC will attempt to pack complex
4237 arguments into the target's word size. Some ABIs require complex arguments
4238 to be split and treated as their individual components. For example, on
4239 AIX64, complex floats should be passed in a pair of floating point
4240 registers, even though a complex float would fit in one 64-bit floating
4243 The default value of this hook is @code{NULL}, which is treated as always
4247 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4248 This hook returns a type node for @code{va_list} for the target.
4249 The default version of the hook returns @code{void*}.
4252 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4253 This target hook is used in function @code{c_common_nodes_and_builtins}
4254 to iterate through the target specific builtin types for va_list. The
4255 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4256 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4258 The arguments @var{pname} and @var{ptree} are used to store the result of
4259 this macro and are set to the name of the va_list builtin type and its
4261 If the return value of this macro is zero, then there is no more element.
4262 Otherwise the @var{IDX} should be increased for the next call of this
4263 macro to iterate through all types.
4266 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4267 This hook returns the va_list type of the calling convention specified by
4269 The default version of this hook returns @code{va_list_type_node}.
4272 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4273 This hook returns the va_list type of the calling convention specified by the
4274 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4278 @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})
4279 This hook performs target-specific gimplification of
4280 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4281 arguments to @code{va_arg}; the latter two are as in
4282 @code{gimplify.c:gimplify_expr}.
4285 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4286 Define this to return nonzero if the port can handle pointers
4287 with machine mode @var{mode}. The default version of this
4288 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4291 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (struct ao_ref_s *@var{ref})
4292 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.
4295 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4296 Define this to return nonzero if the port is prepared to handle
4297 insns involving scalar mode @var{mode}. For a scalar mode to be
4298 considered supported, all the basic arithmetic and comparisons
4301 The default version of this hook returns true for any mode
4302 required to handle the basic C types (as defined by the port).
4303 Included here are the double-word arithmetic supported by the
4304 code in @file{optabs.c}.
4307 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4308 Define this to return nonzero if the port is prepared to handle
4309 insns involving vector mode @var{mode}. At the very least, it
4310 must have move patterns for this mode.
4313 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (enum machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4314 Return true if GCC should try to use a scalar mode to store an array
4315 of @var{nelems} elements, given that each element has mode @var{mode}.
4316 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4317 and allows GCC to use any defined integer mode.
4319 One use of this hook is to support vector load and store operations
4320 that operate on several homogeneous vectors. For example, ARM NEON
4321 has operations like:
4324 int8x8x3_t vld3_s8 (const int8_t *)
4327 where the return type is defined as:
4330 typedef struct int8x8x3_t
4336 If this hook allows @code{val} to have a scalar mode, then
4337 @code{int8x8x3_t} can have the same mode. GCC can then store
4338 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4341 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (enum machine_mode @var{mode})
4342 Define this to return nonzero for machine modes for which the port has
4343 small register classes. If this target hook returns nonzero for a given
4344 @var{mode}, the compiler will try to minimize the lifetime of registers
4345 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4346 In this case, the hook is expected to return nonzero if it returns nonzero
4349 On some machines, it is risky to let hard registers live across arbitrary
4350 insns. Typically, these machines have instructions that require values
4351 to be in specific registers (like an accumulator), and reload will fail
4352 if the required hard register is used for another purpose across such an
4355 Passes before reload do not know which hard registers will be used
4356 in an instruction, but the machine modes of the registers set or used in
4357 the instruction are already known. And for some machines, register
4358 classes are small for, say, integer registers but not for floating point
4359 registers. For example, the AMD x86-64 architecture requires specific
4360 registers for the legacy x86 integer instructions, but there are many
4361 SSE registers for floating point operations. On such targets, a good
4362 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4363 machine modes but zero for the SSE register classes.
4365 The default version of this hook returns false for any mode. It is always
4366 safe to redefine this hook to return with a nonzero value. But if you
4367 unnecessarily define it, you will reduce the amount of optimizations
4368 that can be performed in some cases. If you do not define this hook
4369 to return a nonzero value when it is required, the compiler will run out
4370 of spill registers and print a fatal error message.
4373 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
4374 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.
4378 @subsection How Scalar Function Values Are Returned
4379 @cindex return values in registers
4380 @cindex values, returned by functions
4381 @cindex scalars, returned as values
4383 This section discusses the macros that control returning scalars as
4384 values---values that can fit in registers.
4386 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4388 Define this to return an RTX representing the place where a function
4389 returns or receives a value of data type @var{ret_type}, a tree node
4390 representing a data type. @var{fn_decl_or_type} is a tree node
4391 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4392 function being called. If @var{outgoing} is false, the hook should
4393 compute the register in which the caller will see the return value.
4394 Otherwise, the hook should return an RTX representing the place where
4395 a function returns a value.
4397 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4398 (Actually, on most machines, scalar values are returned in the same
4399 place regardless of mode.) The value of the expression is usually a
4400 @code{reg} RTX for the hard register where the return value is stored.
4401 The value can also be a @code{parallel} RTX, if the return value is in
4402 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4403 @code{parallel} form. Note that the callee will populate every
4404 location specified in the @code{parallel}, but if the first element of
4405 the @code{parallel} contains the whole return value, callers will use
4406 that element as the canonical location and ignore the others. The m68k
4407 port uses this type of @code{parallel} to return pointers in both
4408 @samp{%a0} (the canonical location) and @samp{%d0}.
4410 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4411 the same promotion rules specified in @code{PROMOTE_MODE} if
4412 @var{valtype} is a scalar type.
4414 If the precise function being called is known, @var{func} is a tree
4415 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4416 pointer. This makes it possible to use a different value-returning
4417 convention for specific functions when all their calls are
4420 Some target machines have ``register windows'' so that the register in
4421 which a function returns its value is not the same as the one in which
4422 the caller sees the value. For such machines, you should return
4423 different RTX depending on @var{outgoing}.
4425 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4426 aggregate data types, because these are returned in another way. See
4427 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4430 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4431 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4432 a new target instead.
4435 @defmac LIBCALL_VALUE (@var{mode})
4436 A C expression to create an RTX representing the place where a library
4437 function returns a value of mode @var{mode}.
4439 Note that ``library function'' in this context means a compiler
4440 support routine, used to perform arithmetic, whose name is known
4441 specially by the compiler and was not mentioned in the C code being
4445 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (enum machine_mode @var{mode}, const_rtx @var{fun})
4446 Define this hook if the back-end needs to know the name of the libcall
4447 function in order to determine where the result should be returned.
4449 The mode of the result is given by @var{mode} and the name of the called
4450 library function is given by @var{fun}. The hook should return an RTX
4451 representing the place where the library function result will be returned.
4453 If this hook is not defined, then LIBCALL_VALUE will be used.
4456 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4457 A C expression that is nonzero if @var{regno} is the number of a hard
4458 register in which the values of called function may come back.
4460 A register whose use for returning values is limited to serving as the
4461 second of a pair (for a value of type @code{double}, say) need not be
4462 recognized by this macro. So for most machines, this definition
4466 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4469 If the machine has register windows, so that the caller and the called
4470 function use different registers for the return value, this macro
4471 should recognize only the caller's register numbers.
4473 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4474 for a new target instead.
4477 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4478 A target hook that return @code{true} if @var{regno} is the number of a hard
4479 register in which the values of called function may come back.
4481 A register whose use for returning values is limited to serving as the
4482 second of a pair (for a value of type @code{double}, say) need not be
4483 recognized by this target hook.
4485 If the machine has register windows, so that the caller and the called
4486 function use different registers for the return value, this target hook
4487 should recognize only the caller's register numbers.
4489 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4492 @defmac APPLY_RESULT_SIZE
4493 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4494 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4495 saving and restoring an arbitrary return value.
4498 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4499 This hook should return true if values of type @var{type} are returned
4500 at the most significant end of a register (in other words, if they are
4501 padded at the least significant end). You can assume that @var{type}
4502 is returned in a register; the caller is required to check this.
4504 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4505 be able to hold the complete return value. For example, if a 1-, 2-
4506 or 3-byte structure is returned at the most significant end of a
4507 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4511 @node Aggregate Return
4512 @subsection How Large Values Are Returned
4513 @cindex aggregates as return values
4514 @cindex large return values
4515 @cindex returning aggregate values
4516 @cindex structure value address
4518 When a function value's mode is @code{BLKmode} (and in some other
4519 cases), the value is not returned according to
4520 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4521 caller passes the address of a block of memory in which the value
4522 should be stored. This address is called the @dfn{structure value
4525 This section describes how to control returning structure values in
4528 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4529 This target hook should return a nonzero value to say to return the
4530 function value in memory, just as large structures are always returned.
4531 Here @var{type} will be the data type of the value, and @var{fntype}
4532 will be the type of the function doing the returning, or @code{NULL} for
4535 Note that values of mode @code{BLKmode} must be explicitly handled
4536 by this function. Also, the option @option{-fpcc-struct-return}
4537 takes effect regardless of this macro. On most systems, it is
4538 possible to leave the hook undefined; this causes a default
4539 definition to be used, whose value is the constant 1 for @code{BLKmode}
4540 values, and 0 otherwise.
4542 Do not use this hook to indicate that structures and unions should always
4543 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4547 @defmac DEFAULT_PCC_STRUCT_RETURN
4548 Define this macro to be 1 if all structure and union return values must be
4549 in memory. Since this results in slower code, this should be defined
4550 only if needed for compatibility with other compilers or with an ABI@.
4551 If you define this macro to be 0, then the conventions used for structure
4552 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4555 If not defined, this defaults to the value 1.
4558 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4559 This target hook should return the location of the structure value
4560 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4561 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4562 be @code{NULL}, for libcalls. You do not need to define this target
4563 hook if the address is always passed as an ``invisible'' first
4566 On some architectures the place where the structure value address
4567 is found by the called function is not the same place that the
4568 caller put it. This can be due to register windows, or it could
4569 be because the function prologue moves it to a different place.
4570 @var{incoming} is @code{1} or @code{2} when the location is needed in
4571 the context of the called function, and @code{0} in the context of
4574 If @var{incoming} is nonzero and the address is to be found on the
4575 stack, return a @code{mem} which refers to the frame pointer. If
4576 @var{incoming} is @code{2}, the result is being used to fetch the
4577 structure value address at the beginning of a function. If you need
4578 to emit adjusting code, you should do it at this point.
4581 @defmac PCC_STATIC_STRUCT_RETURN
4582 Define this macro if the usual system convention on the target machine
4583 for returning structures and unions is for the called function to return
4584 the address of a static variable containing the value.
4586 Do not define this if the usual system convention is for the caller to
4587 pass an address to the subroutine.
4589 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4590 nothing when you use @option{-freg-struct-return} mode.
4593 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4594 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.
4597 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_ARG_MODE (int @var{regno})
4598 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.
4602 @subsection Caller-Saves Register Allocation
4604 If you enable it, GCC can save registers around function calls. This
4605 makes it possible to use call-clobbered registers to hold variables that
4606 must live across calls.
4608 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4609 A C expression to determine whether it is worthwhile to consider placing
4610 a pseudo-register in a call-clobbered hard register and saving and
4611 restoring it around each function call. The expression should be 1 when
4612 this is worth doing, and 0 otherwise.
4614 If you don't define this macro, a default is used which is good on most
4615 machines: @code{4 * @var{calls} < @var{refs}}.
4618 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4619 A C expression specifying which mode is required for saving @var{nregs}
4620 of a pseudo-register in call-clobbered hard register @var{regno}. If
4621 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4622 returned. For most machines this macro need not be defined since GCC
4623 will select the smallest suitable mode.
4626 @node Function Entry
4627 @subsection Function Entry and Exit
4628 @cindex function entry and exit
4632 This section describes the macros that output function entry
4633 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4635 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4636 If defined, a function that outputs the assembler code for entry to a
4637 function. The prologue is responsible for setting up the stack frame,
4638 initializing the frame pointer register, saving registers that must be
4639 saved, and allocating @var{size} additional bytes of storage for the
4640 local variables. @var{size} is an integer. @var{file} is a stdio
4641 stream to which the assembler code should be output.
4643 The label for the beginning of the function need not be output by this
4644 macro. That has already been done when the macro is run.
4646 @findex regs_ever_live
4647 To determine which registers to save, the macro can refer to the array
4648 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4649 @var{r} is used anywhere within the function. This implies the function
4650 prologue should save register @var{r}, provided it is not one of the
4651 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4652 @code{regs_ever_live}.)
4654 On machines that have ``register windows'', the function entry code does
4655 not save on the stack the registers that are in the windows, even if
4656 they are supposed to be preserved by function calls; instead it takes
4657 appropriate steps to ``push'' the register stack, if any non-call-used
4658 registers are used in the function.
4660 @findex frame_pointer_needed
4661 On machines where functions may or may not have frame-pointers, the
4662 function entry code must vary accordingly; it must set up the frame
4663 pointer if one is wanted, and not otherwise. To determine whether a
4664 frame pointer is in wanted, the macro can refer to the variable
4665 @code{frame_pointer_needed}. The variable's value will be 1 at run
4666 time in a function that needs a frame pointer. @xref{Elimination}.
4668 The function entry code is responsible for allocating any stack space
4669 required for the function. This stack space consists of the regions
4670 listed below. In most cases, these regions are allocated in the
4671 order listed, with the last listed region closest to the top of the
4672 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4673 the highest address if it is not defined). You can use a different order
4674 for a machine if doing so is more convenient or required for
4675 compatibility reasons. Except in cases where required by standard
4676 or by a debugger, there is no reason why the stack layout used by GCC
4677 need agree with that used by other compilers for a machine.
4680 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4681 If defined, a function that outputs assembler code at the end of a
4682 prologue. This should be used when the function prologue is being
4683 emitted as RTL, and you have some extra assembler that needs to be
4684 emitted. @xref{prologue instruction pattern}.
4687 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4688 If defined, a function that outputs assembler code at the start of an
4689 epilogue. This should be used when the function epilogue is being
4690 emitted as RTL, and you have some extra assembler that needs to be
4691 emitted. @xref{epilogue instruction pattern}.
4694 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4695 If defined, a function that outputs the assembler code for exit from a
4696 function. The epilogue is responsible for restoring the saved
4697 registers and stack pointer to their values when the function was
4698 called, and returning control to the caller. This macro takes the
4699 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4700 registers to restore are determined from @code{regs_ever_live} and
4701 @code{CALL_USED_REGISTERS} in the same way.
4703 On some machines, there is a single instruction that does all the work
4704 of returning from the function. On these machines, give that
4705 instruction the name @samp{return} and do not define the macro
4706 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4708 Do not define a pattern named @samp{return} if you want the
4709 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4710 switches to control whether return instructions or epilogues are used,
4711 define a @samp{return} pattern with a validity condition that tests the
4712 target switches appropriately. If the @samp{return} pattern's validity
4713 condition is false, epilogues will be used.
4715 On machines where functions may or may not have frame-pointers, the
4716 function exit code must vary accordingly. Sometimes the code for these
4717 two cases is completely different. To determine whether a frame pointer
4718 is wanted, the macro can refer to the variable
4719 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4720 a function that needs a frame pointer.
4722 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4723 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4724 The C variable @code{current_function_is_leaf} is nonzero for such a
4725 function. @xref{Leaf Functions}.
4727 On some machines, some functions pop their arguments on exit while
4728 others leave that for the caller to do. For example, the 68020 when
4729 given @option{-mrtd} pops arguments in functions that take a fixed
4730 number of arguments.
4732 @findex current_function_pops_args
4733 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4734 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4735 needs to know what was decided. The number of bytes of the current
4736 function's arguments that this function should pop is available in
4737 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4742 @findex current_function_pretend_args_size
4743 A region of @code{current_function_pretend_args_size} bytes of
4744 uninitialized space just underneath the first argument arriving on the
4745 stack. (This may not be at the very start of the allocated stack region
4746 if the calling sequence has pushed anything else since pushing the stack
4747 arguments. But usually, on such machines, nothing else has been pushed
4748 yet, because the function prologue itself does all the pushing.) This
4749 region is used on machines where an argument may be passed partly in
4750 registers and partly in memory, and, in some cases to support the
4751 features in @code{<stdarg.h>}.
4754 An area of memory used to save certain registers used by the function.
4755 The size of this area, which may also include space for such things as
4756 the return address and pointers to previous stack frames, is
4757 machine-specific and usually depends on which registers have been used
4758 in the function. Machines with register windows often do not require
4762 A region of at least @var{size} bytes, possibly rounded up to an allocation
4763 boundary, to contain the local variables of the function. On some machines,
4764 this region and the save area may occur in the opposite order, with the
4765 save area closer to the top of the stack.
4768 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4769 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4770 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4771 argument lists of the function. @xref{Stack Arguments}.
4774 @defmac EXIT_IGNORE_STACK
4775 Define this macro as a C expression that is nonzero if the return
4776 instruction or the function epilogue ignores the value of the stack
4777 pointer; in other words, if it is safe to delete an instruction to
4778 adjust the stack pointer before a return from the function. The
4781 Note that this macro's value is relevant only for functions for which
4782 frame pointers are maintained. It is never safe to delete a final
4783 stack adjustment in a function that has no frame pointer, and the
4784 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4787 @defmac EPILOGUE_USES (@var{regno})
4788 Define this macro as a C expression that is nonzero for registers that are
4789 used by the epilogue or the @samp{return} pattern. The stack and frame
4790 pointer registers are already assumed to be used as needed.
4793 @defmac EH_USES (@var{regno})
4794 Define this macro as a C expression that is nonzero for registers that are
4795 used by the exception handling mechanism, and so should be considered live
4796 on entry to an exception edge.
4799 @defmac DELAY_SLOTS_FOR_EPILOGUE
4800 Define this macro if the function epilogue contains delay slots to which
4801 instructions from the rest of the function can be ``moved''. The
4802 definition should be a C expression whose value is an integer
4803 representing the number of delay slots there.
4806 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4807 A C expression that returns 1 if @var{insn} can be placed in delay
4808 slot number @var{n} of the epilogue.
4810 The argument @var{n} is an integer which identifies the delay slot now
4811 being considered (since different slots may have different rules of
4812 eligibility). It is never negative and is always less than the number
4813 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4814 If you reject a particular insn for a given delay slot, in principle, it
4815 may be reconsidered for a subsequent delay slot. Also, other insns may
4816 (at least in principle) be considered for the so far unfilled delay
4819 @findex current_function_epilogue_delay_list
4820 @findex final_scan_insn
4821 The insns accepted to fill the epilogue delay slots are put in an RTL
4822 list made with @code{insn_list} objects, stored in the variable
4823 @code{current_function_epilogue_delay_list}. The insn for the first
4824 delay slot comes first in the list. Your definition of the macro
4825 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4826 outputting the insns in this list, usually by calling
4827 @code{final_scan_insn}.
4829 You need not define this macro if you did not define
4830 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4833 @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})
4834 A function that outputs the assembler code for a thunk
4835 function, used to implement C++ virtual function calls with multiple
4836 inheritance. The thunk acts as a wrapper around a virtual function,
4837 adjusting the implicit object parameter before handing control off to
4840 First, emit code to add the integer @var{delta} to the location that
4841 contains the incoming first argument. Assume that this argument
4842 contains a pointer, and is the one used to pass the @code{this} pointer
4843 in C++. This is the incoming argument @emph{before} the function prologue,
4844 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4845 all other incoming arguments.
4847 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4848 made after adding @code{delta}. In particular, if @var{p} is the
4849 adjusted pointer, the following adjustment should be made:
4852 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4855 After the additions, emit code to jump to @var{function}, which is a
4856 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4857 not touch the return address. Hence returning from @var{FUNCTION} will
4858 return to whoever called the current @samp{thunk}.
4860 The effect must be as if @var{function} had been called directly with
4861 the adjusted first argument. This macro is responsible for emitting all
4862 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4863 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4865 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4866 have already been extracted from it.) It might possibly be useful on
4867 some targets, but probably not.
4869 If you do not define this macro, the target-independent code in the C++
4870 front end will generate a less efficient heavyweight thunk that calls
4871 @var{function} instead of jumping to it. The generic approach does
4872 not support varargs.
4875 @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})
4876 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4877 to output the assembler code for the thunk function specified by the
4878 arguments it is passed, and false otherwise. In the latter case, the
4879 generic approach will be used by the C++ front end, with the limitations
4884 @subsection Generating Code for Profiling
4885 @cindex profiling, code generation
4887 These macros will help you generate code for profiling.
4889 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4890 A C statement or compound statement to output to @var{file} some
4891 assembler code to call the profiling subroutine @code{mcount}.
4894 The details of how @code{mcount} expects to be called are determined by
4895 your operating system environment, not by GCC@. To figure them out,
4896 compile a small program for profiling using the system's installed C
4897 compiler and look at the assembler code that results.
4899 Older implementations of @code{mcount} expect the address of a counter
4900 variable to be loaded into some register. The name of this variable is
4901 @samp{LP} followed by the number @var{labelno}, so you would generate
4902 the name using @samp{LP%d} in a @code{fprintf}.
4905 @defmac PROFILE_HOOK
4906 A C statement or compound statement to output to @var{file} some assembly
4907 code to call the profiling subroutine @code{mcount} even the target does
4908 not support profiling.
4911 @defmac NO_PROFILE_COUNTERS
4912 Define this macro to be an expression with a nonzero value if the
4913 @code{mcount} subroutine on your system does not need a counter variable
4914 allocated for each function. This is true for almost all modern
4915 implementations. If you define this macro, you must not use the
4916 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4919 @defmac PROFILE_BEFORE_PROLOGUE
4920 Define this macro if the code for function profiling should come before
4921 the function prologue. Normally, the profiling code comes after.
4925 @subsection Permitting tail calls
4928 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4929 True if it is ok to do sibling call optimization for the specified
4930 call expression @var{exp}. @var{decl} will be the called function,
4931 or @code{NULL} if this is an indirect call.
4933 It is not uncommon for limitations of calling conventions to prevent
4934 tail calls to functions outside the current unit of translation, or
4935 during PIC compilation. The hook is used to enforce these restrictions,
4936 as the @code{sibcall} md pattern can not fail, or fall over to a
4937 ``normal'' call. The criteria for successful sibling call optimization
4938 may vary greatly between different architectures.
4941 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4942 Add any hard registers to @var{regs} that are live on entry to the
4943 function. This hook only needs to be defined to provide registers that
4944 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4945 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4946 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4947 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4950 @node Stack Smashing Protection
4951 @subsection Stack smashing protection
4952 @cindex stack smashing protection
4954 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4955 This hook returns a @code{DECL} node for the external variable to use
4956 for the stack protection guard. This variable is initialized by the
4957 runtime to some random value and is used to initialize the guard value
4958 that is placed at the top of the local stack frame. The type of this
4959 variable must be @code{ptr_type_node}.
4961 The default version of this hook creates a variable called
4962 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4965 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4966 This hook returns a tree expression that alerts the runtime that the
4967 stack protect guard variable has been modified. This expression should
4968 involve a call to a @code{noreturn} function.
4970 The default version of this hook invokes a function called
4971 @samp{__stack_chk_fail}, taking no arguments. This function is
4972 normally defined in @file{libgcc2.c}.
4975 @deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
4976 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
4980 @section Implementing the Varargs Macros
4981 @cindex varargs implementation
4983 GCC comes with an implementation of @code{<varargs.h>} and
4984 @code{<stdarg.h>} that work without change on machines that pass arguments
4985 on the stack. Other machines require their own implementations of
4986 varargs, and the two machine independent header files must have
4987 conditionals to include it.
4989 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4990 the calling convention for @code{va_start}. The traditional
4991 implementation takes just one argument, which is the variable in which
4992 to store the argument pointer. The ISO implementation of
4993 @code{va_start} takes an additional second argument. The user is
4994 supposed to write the last named argument of the function here.
4996 However, @code{va_start} should not use this argument. The way to find
4997 the end of the named arguments is with the built-in functions described
5000 @defmac __builtin_saveregs ()
5001 Use this built-in function to save the argument registers in memory so
5002 that the varargs mechanism can access them. Both ISO and traditional
5003 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
5004 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
5006 On some machines, @code{__builtin_saveregs} is open-coded under the
5007 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
5008 other machines, it calls a routine written in assembler language,
5009 found in @file{libgcc2.c}.
5011 Code generated for the call to @code{__builtin_saveregs} appears at the
5012 beginning of the function, as opposed to where the call to
5013 @code{__builtin_saveregs} is written, regardless of what the code is.
5014 This is because the registers must be saved before the function starts
5015 to use them for its own purposes.
5016 @c i rewrote the first sentence above to fix an overfull hbox. --mew
5020 @defmac __builtin_next_arg (@var{lastarg})
5021 This builtin returns the address of the first anonymous stack
5022 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
5023 returns the address of the location above the first anonymous stack
5024 argument. Use it in @code{va_start} to initialize the pointer for
5025 fetching arguments from the stack. Also use it in @code{va_start} to
5026 verify that the second parameter @var{lastarg} is the last named argument
5027 of the current function.
5030 @defmac __builtin_classify_type (@var{object})
5031 Since each machine has its own conventions for which data types are
5032 passed in which kind of register, your implementation of @code{va_arg}
5033 has to embody these conventions. The easiest way to categorize the
5034 specified data type is to use @code{__builtin_classify_type} together
5035 with @code{sizeof} and @code{__alignof__}.
5037 @code{__builtin_classify_type} ignores the value of @var{object},
5038 considering only its data type. It returns an integer describing what
5039 kind of type that is---integer, floating, pointer, structure, and so on.
5041 The file @file{typeclass.h} defines an enumeration that you can use to
5042 interpret the values of @code{__builtin_classify_type}.
5045 These machine description macros help implement varargs:
5047 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5048 If defined, this hook produces the machine-specific code for a call to
5049 @code{__builtin_saveregs}. This code will be moved to the very
5050 beginning of the function, before any parameter access are made. The
5051 return value of this function should be an RTX that contains the value
5052 to use as the return of @code{__builtin_saveregs}.
5055 @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})
5056 This target hook offers an alternative to using
5057 @code{__builtin_saveregs} and defining the hook
5058 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5059 register arguments into the stack so that all the arguments appear to
5060 have been passed consecutively on the stack. Once this is done, you can
5061 use the standard implementation of varargs that works for machines that
5062 pass all their arguments on the stack.
5064 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5065 structure, containing the values that are obtained after processing the
5066 named arguments. The arguments @var{mode} and @var{type} describe the
5067 last named argument---its machine mode and its data type as a tree node.
5069 The target hook should do two things: first, push onto the stack all the
5070 argument registers @emph{not} used for the named arguments, and second,
5071 store the size of the data thus pushed into the @code{int}-valued
5072 variable pointed to by @var{pretend_args_size}. The value that you
5073 store here will serve as additional offset for setting up the stack
5076 Because you must generate code to push the anonymous arguments at
5077 compile time without knowing their data types,
5078 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5079 have just a single category of argument register and use it uniformly
5082 If the argument @var{second_time} is nonzero, it means that the
5083 arguments of the function are being analyzed for the second time. This
5084 happens for an inline function, which is not actually compiled until the
5085 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5086 not generate any instructions in this case.
5089 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca})
5090 Define this hook to return @code{true} if the location where a function
5091 argument is passed depends on whether or not it is a named argument.
5093 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5094 is set for varargs and stdarg functions. If this hook returns
5095 @code{true}, the @var{named} argument is always true for named
5096 arguments, and false for unnamed arguments. If it returns @code{false},
5097 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5098 then all arguments are treated as named. Otherwise, all named arguments
5099 except the last are treated as named.
5101 You need not define this hook if it always returns @code{false}.
5104 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca})
5105 If you need to conditionally change ABIs so that one works with
5106 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5107 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5108 defined, then define this hook to return @code{true} if
5109 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5110 Otherwise, you should not define this hook.
5114 @section Trampolines for Nested Functions
5115 @cindex trampolines for nested functions
5116 @cindex nested functions, trampolines for
5118 A @dfn{trampoline} is a small piece of code that is created at run time
5119 when the address of a nested function is taken. It normally resides on
5120 the stack, in the stack frame of the containing function. These macros
5121 tell GCC how to generate code to allocate and initialize a
5124 The instructions in the trampoline must do two things: load a constant
5125 address into the static chain register, and jump to the real address of
5126 the nested function. On CISC machines such as the m68k, this requires
5127 two instructions, a move immediate and a jump. Then the two addresses
5128 exist in the trampoline as word-long immediate operands. On RISC
5129 machines, it is often necessary to load each address into a register in
5130 two parts. Then pieces of each address form separate immediate
5133 The code generated to initialize the trampoline must store the variable
5134 parts---the static chain value and the function address---into the
5135 immediate operands of the instructions. On a CISC machine, this is
5136 simply a matter of copying each address to a memory reference at the
5137 proper offset from the start of the trampoline. On a RISC machine, it
5138 may be necessary to take out pieces of the address and store them
5141 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5142 This hook is called by @code{assemble_trampoline_template} to output,
5143 on the stream @var{f}, assembler code for a block of data that contains
5144 the constant parts of a trampoline. This code should not include a
5145 label---the label is taken care of automatically.
5147 If you do not define this hook, it means no template is needed
5148 for the target. Do not define this hook on systems where the block move
5149 code to copy the trampoline into place would be larger than the code
5150 to generate it on the spot.
5153 @defmac TRAMPOLINE_SECTION
5154 Return the section into which the trampoline template is to be placed
5155 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5158 @defmac TRAMPOLINE_SIZE
5159 A C expression for the size in bytes of the trampoline, as an integer.
5162 @defmac TRAMPOLINE_ALIGNMENT
5163 Alignment required for trampolines, in bits.
5165 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5166 is used for aligning trampolines.
5169 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5170 This hook is called to initialize a trampoline.
5171 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5172 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5173 RTX for the static chain value that should be passed to the function
5176 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5177 first thing this hook should do is emit a block move into @var{m_tramp}
5178 from the memory block returned by @code{assemble_trampoline_template}.
5179 Note that the block move need only cover the constant parts of the
5180 trampoline. If the target isolates the variable parts of the trampoline
5181 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5183 If the target requires any other actions, such as flushing caches or
5184 enabling stack execution, these actions should be performed after
5185 initializing the trampoline proper.
5188 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5189 This hook should perform any machine-specific adjustment in
5190 the address of the trampoline. Its argument contains the address of the
5191 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5192 the address to be used for a function call should be different from the
5193 address at which the template was stored, the different address should
5194 be returned; otherwise @var{addr} should be returned unchanged.
5195 If this hook is not defined, @var{addr} will be used for function calls.
5198 Implementing trampolines is difficult on many machines because they have
5199 separate instruction and data caches. Writing into a stack location
5200 fails to clear the memory in the instruction cache, so when the program
5201 jumps to that location, it executes the old contents.
5203 Here are two possible solutions. One is to clear the relevant parts of
5204 the instruction cache whenever a trampoline is set up. The other is to
5205 make all trampolines identical, by having them jump to a standard
5206 subroutine. The former technique makes trampoline execution faster; the
5207 latter makes initialization faster.
5209 To clear the instruction cache when a trampoline is initialized, define
5210 the following macro.
5212 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5213 If defined, expands to a C expression clearing the @emph{instruction
5214 cache} in the specified interval. The definition of this macro would
5215 typically be a series of @code{asm} statements. Both @var{beg} and
5216 @var{end} are both pointer expressions.
5219 To use a standard subroutine, define the following macro. In addition,
5220 you must make sure that the instructions in a trampoline fill an entire
5221 cache line with identical instructions, or else ensure that the
5222 beginning of the trampoline code is always aligned at the same point in
5223 its cache line. Look in @file{m68k.h} as a guide.
5225 @defmac TRANSFER_FROM_TRAMPOLINE
5226 Define this macro if trampolines need a special subroutine to do their
5227 work. The macro should expand to a series of @code{asm} statements
5228 which will be compiled with GCC@. They go in a library function named
5229 @code{__transfer_from_trampoline}.
5231 If you need to avoid executing the ordinary prologue code of a compiled
5232 C function when you jump to the subroutine, you can do so by placing a
5233 special label of your own in the assembler code. Use one @code{asm}
5234 statement to generate an assembler label, and another to make the label
5235 global. Then trampolines can use that label to jump directly to your
5236 special assembler code.
5240 @section Implicit Calls to Library Routines
5241 @cindex library subroutine names
5242 @cindex @file{libgcc.a}
5244 @c prevent bad page break with this line
5245 Here is an explanation of implicit calls to library routines.
5247 @defmac DECLARE_LIBRARY_RENAMES
5248 This macro, if defined, should expand to a piece of C code that will get
5249 expanded when compiling functions for libgcc.a. It can be used to
5250 provide alternate names for GCC's internal library functions if there
5251 are ABI-mandated names that the compiler should provide.
5254 @findex set_optab_libfunc
5255 @findex init_one_libfunc
5256 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5257 This hook should declare additional library routines or rename
5258 existing ones, using the functions @code{set_optab_libfunc} and
5259 @code{init_one_libfunc} defined in @file{optabs.c}.
5260 @code{init_optabs} calls this macro after initializing all the normal
5263 The default is to do nothing. Most ports don't need to define this hook.
5266 @deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX
5267 If false (the default), internal library routines start with two
5268 underscores. If set to true, these routines start with @code{__gnu_}
5269 instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This
5270 currently only affects functions defined in @file{libgcc2.c}. If this
5271 is set to true, the @file{tm.h} file must also
5272 @code{#define LIBGCC2_GNU_PREFIX}.
5275 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5276 This macro should return @code{true} if the library routine that
5277 implements the floating point comparison operator @var{comparison} in
5278 mode @var{mode} will return a boolean, and @var{false} if it will
5281 GCC's own floating point libraries return tristates from the
5282 comparison operators, so the default returns false always. Most ports
5283 don't need to define this macro.
5286 @defmac TARGET_LIB_INT_CMP_BIASED
5287 This macro should evaluate to @code{true} if the integer comparison
5288 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5289 operand is smaller than the second, 1 to indicate that they are equal,
5290 and 2 to indicate that the first operand is greater than the second.
5291 If this macro evaluates to @code{false} the comparison functions return
5292 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5293 in @file{libgcc.a}, you do not need to define this macro.
5296 @cindex @code{EDOM}, implicit usage
5299 The value of @code{EDOM} on the target machine, as a C integer constant
5300 expression. If you don't define this macro, GCC does not attempt to
5301 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5302 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5305 If you do not define @code{TARGET_EDOM}, then compiled code reports
5306 domain errors by calling the library function and letting it report the
5307 error. If mathematical functions on your system use @code{matherr} when
5308 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5309 that @code{matherr} is used normally.
5312 @cindex @code{errno}, implicit usage
5313 @defmac GEN_ERRNO_RTX
5314 Define this macro as a C expression to create an rtl expression that
5315 refers to the global ``variable'' @code{errno}. (On certain systems,
5316 @code{errno} may not actually be a variable.) If you don't define this
5317 macro, a reasonable default is used.
5320 @cindex C99 math functions, implicit usage
5321 @defmac TARGET_C99_FUNCTIONS
5322 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5323 @code{sinf} and similarly for other functions defined by C99 standard. The
5324 default is zero because a number of existing systems lack support for these
5325 functions in their runtime so this macro needs to be redefined to one on
5326 systems that do support the C99 runtime.
5329 @cindex sincos math function, implicit usage
5330 @defmac TARGET_HAS_SINCOS
5331 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5332 and @code{cos} with the same argument to a call to @code{sincos}. The
5333 default is zero. The target has to provide the following functions:
5335 void sincos(double x, double *sin, double *cos);
5336 void sincosf(float x, float *sin, float *cos);
5337 void sincosl(long double x, long double *sin, long double *cos);
5341 @defmac NEXT_OBJC_RUNTIME
5342 Define this macro to generate code for Objective-C message sending using
5343 the calling convention of the NeXT system. This calling convention
5344 involves passing the object, the selector and the method arguments all
5345 at once to the method-lookup library function.
5347 The default calling convention passes just the object and the selector
5348 to the lookup function, which returns a pointer to the method.
5351 @node Addressing Modes
5352 @section Addressing Modes
5353 @cindex addressing modes
5355 @c prevent bad page break with this line
5356 This is about addressing modes.
5358 @defmac HAVE_PRE_INCREMENT
5359 @defmacx HAVE_PRE_DECREMENT
5360 @defmacx HAVE_POST_INCREMENT
5361 @defmacx HAVE_POST_DECREMENT
5362 A C expression that is nonzero if the machine supports pre-increment,
5363 pre-decrement, post-increment, or post-decrement addressing respectively.
5366 @defmac HAVE_PRE_MODIFY_DISP
5367 @defmacx HAVE_POST_MODIFY_DISP
5368 A C expression that is nonzero if the machine supports pre- or
5369 post-address side-effect generation involving constants other than
5370 the size of the memory operand.
5373 @defmac HAVE_PRE_MODIFY_REG
5374 @defmacx HAVE_POST_MODIFY_REG
5375 A C expression that is nonzero if the machine supports pre- or
5376 post-address side-effect generation involving a register displacement.
5379 @defmac CONSTANT_ADDRESS_P (@var{x})
5380 A C expression that is 1 if the RTX @var{x} is a constant which
5381 is a valid address. On most machines the default definition of
5382 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5383 is acceptable, but a few machines are more restrictive as to which
5384 constant addresses are supported.
5387 @defmac CONSTANT_P (@var{x})
5388 @code{CONSTANT_P}, which is defined by target-independent code,
5389 accepts integer-values expressions whose values are not explicitly
5390 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5391 expressions and @code{const} arithmetic expressions, in addition to
5392 @code{const_int} and @code{const_double} expressions.
5395 @defmac MAX_REGS_PER_ADDRESS
5396 A number, the maximum number of registers that can appear in a valid
5397 memory address. Note that it is up to you to specify a value equal to
5398 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5402 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5403 A function that returns whether @var{x} (an RTX) is a legitimate memory
5404 address on the target machine for a memory operand of mode @var{mode}.
5406 Legitimate addresses are defined in two variants: a strict variant and a
5407 non-strict one. The @var{strict} parameter chooses which variant is
5408 desired by the caller.
5410 The strict variant is used in the reload pass. It must be defined so
5411 that any pseudo-register that has not been allocated a hard register is
5412 considered a memory reference. This is because in contexts where some
5413 kind of register is required, a pseudo-register with no hard register
5414 must be rejected. For non-hard registers, the strict variant should look
5415 up the @code{reg_renumber} array; it should then proceed using the hard
5416 register number in the array, or treat the pseudo as a memory reference
5417 if the array holds @code{-1}.
5419 The non-strict variant is used in other passes. It must be defined to
5420 accept all pseudo-registers in every context where some kind of
5421 register is required.
5423 Normally, constant addresses which are the sum of a @code{symbol_ref}
5424 and an integer are stored inside a @code{const} RTX to mark them as
5425 constant. Therefore, there is no need to recognize such sums
5426 specifically as legitimate addresses. Normally you would simply
5427 recognize any @code{const} as legitimate.
5429 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5430 sums that are not marked with @code{const}. It assumes that a naked
5431 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5432 naked constant sums as illegitimate addresses, so that none of them will
5433 be given to @code{PRINT_OPERAND_ADDRESS}.
5435 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5436 On some machines, whether a symbolic address is legitimate depends on
5437 the section that the address refers to. On these machines, define the
5438 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5439 into the @code{symbol_ref}, and then check for it here. When you see a
5440 @code{const}, you will have to look inside it to find the
5441 @code{symbol_ref} in order to determine the section. @xref{Assembler
5444 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5445 Some ports are still using a deprecated legacy substitute for
5446 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5450 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5454 and should @code{goto @var{label}} if the address @var{x} is a valid
5455 address on the target machine for a memory operand of mode @var{mode}.
5457 @findex REG_OK_STRICT
5458 Compiler source files that want to use the strict variant of this
5459 macro define the macro @code{REG_OK_STRICT}. You should use an
5460 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5461 that case and the non-strict variant otherwise.
5463 Using the hook is usually simpler because it limits the number of
5464 files that are recompiled when changes are made.
5467 @defmac TARGET_MEM_CONSTRAINT
5468 A single character to be used instead of the default @code{'m'}
5469 character for general memory addresses. This defines the constraint
5470 letter which matches the memory addresses accepted by
5471 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5472 support new address formats in your back end without changing the
5473 semantics of the @code{'m'} constraint. This is necessary in order to
5474 preserve functionality of inline assembly constructs using the
5475 @code{'m'} constraint.
5478 @defmac FIND_BASE_TERM (@var{x})
5479 A C expression to determine the base term of address @var{x},
5480 or to provide a simplified version of @var{x} from which @file{alias.c}
5481 can easily find the base term. This macro is used in only two places:
5482 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5484 It is always safe for this macro to not be defined. It exists so
5485 that alias analysis can understand machine-dependent addresses.
5487 The typical use of this macro is to handle addresses containing
5488 a label_ref or symbol_ref within an UNSPEC@.
5491 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode})
5492 This hook is given an invalid memory address @var{x} for an
5493 operand of mode @var{mode} and should try to return a valid memory
5496 @findex break_out_memory_refs
5497 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5498 and @var{oldx} will be the operand that was given to that function to produce
5501 The code of the hook should not alter the substructure of
5502 @var{x}. If it transforms @var{x} into a more legitimate form, it
5503 should return the new @var{x}.
5505 It is not necessary for this hook to come up with a legitimate address.
5506 The compiler has standard ways of doing so in all cases. In fact, it
5507 is safe to omit this hook or make it return @var{x} if it cannot find
5508 a valid way to legitimize the address. But often a machine-dependent
5509 strategy can generate better code.
5512 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5513 A C compound statement that attempts to replace @var{x}, which is an address
5514 that needs reloading, with a valid memory address for an operand of mode
5515 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5516 It is not necessary to define this macro, but it might be useful for
5517 performance reasons.
5519 For example, on the i386, it is sometimes possible to use a single
5520 reload register instead of two by reloading a sum of two pseudo
5521 registers into a register. On the other hand, for number of RISC
5522 processors offsets are limited so that often an intermediate address
5523 needs to be generated in order to address a stack slot. By defining
5524 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5525 generated for adjacent some stack slots can be made identical, and thus
5528 @emph{Note}: This macro should be used with caution. It is necessary
5529 to know something of how reload works in order to effectively use this,
5530 and it is quite easy to produce macros that build in too much knowledge
5531 of reload internals.
5533 @emph{Note}: This macro must be able to reload an address created by a
5534 previous invocation of this macro. If it fails to handle such addresses
5535 then the compiler may generate incorrect code or abort.
5538 The macro definition should use @code{push_reload} to indicate parts that
5539 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5540 suitable to be passed unaltered to @code{push_reload}.
5542 The code generated by this macro must not alter the substructure of
5543 @var{x}. If it transforms @var{x} into a more legitimate form, it
5544 should assign @var{x} (which will always be a C variable) a new value.
5545 This also applies to parts that you change indirectly by calling
5548 @findex strict_memory_address_p
5549 The macro definition may use @code{strict_memory_address_p} to test if
5550 the address has become legitimate.
5553 If you want to change only a part of @var{x}, one standard way of doing
5554 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5555 single level of rtl. Thus, if the part to be changed is not at the
5556 top level, you'll need to replace first the top level.
5557 It is not necessary for this macro to come up with a legitimate
5558 address; but often a machine-dependent strategy can generate better code.
5561 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr})
5562 This hook returns @code{true} if memory address @var{addr} can have
5563 different meanings depending on the machine mode of the memory
5564 reference it is used for or if the address is valid for some modes
5567 Autoincrement and autodecrement addresses typically have mode-dependent
5568 effects because the amount of the increment or decrement is the size
5569 of the operand being addressed. Some machines have other mode-dependent
5570 addresses. Many RISC machines have no mode-dependent addresses.
5572 You may assume that @var{addr} is a valid address for the machine.
5574 The default version of this hook returns @code{false}.
5577 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5578 A C statement or compound statement with a conditional @code{goto
5579 @var{label};} executed if memory address @var{x} (an RTX) can have
5580 different meanings depending on the machine mode of the memory
5581 reference it is used for or if the address is valid for some modes
5584 Autoincrement and autodecrement addresses typically have mode-dependent
5585 effects because the amount of the increment or decrement is the size
5586 of the operand being addressed. Some machines have other mode-dependent
5587 addresses. Many RISC machines have no mode-dependent addresses.
5589 You may assume that @var{addr} is a valid address for the machine.
5591 These are obsolete macros, replaced by the
5592 @code{TARGET_MODE_DEPENDENT_ADDRESS_P} target hook.
5595 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5596 This hook returns true if @var{x} is a legitimate constant for a
5597 @var{mode}-mode immediate operand on the target machine. You can assume that
5598 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5600 The default definition returns true.
5603 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5604 This hook is used to undo the possibly obfuscating effects of the
5605 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5606 macros. Some backend implementations of these macros wrap symbol
5607 references inside an @code{UNSPEC} rtx to represent PIC or similar
5608 addressing modes. This target hook allows GCC's optimizers to understand
5609 the semantics of these opaque @code{UNSPEC}s by converting them back
5610 into their original form.
5613 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (enum machine_mode @var{mode}, rtx @var{x})
5614 This hook should return true if @var{x} is of a form that cannot (or
5615 should not) be spilled to the constant pool. @var{mode} is the mode
5618 The default version of this hook returns false.
5620 The primary reason to define this hook is to prevent reload from
5621 deciding that a non-legitimate constant would be better reloaded
5622 from the constant pool instead of spilling and reloading a register
5623 holding the constant. This restriction is often true of addresses
5624 of TLS symbols for various targets.
5627 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, const_rtx @var{x})
5628 This hook should return true if pool entries for constant @var{x} can
5629 be placed in an @code{object_block} structure. @var{mode} is the mode
5632 The default version returns false for all constants.
5635 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (unsigned @var{fn}, bool @var{md_fn}, bool @var{sqrt})
5636 This hook should return the DECL of a function that implements reciprocal of
5637 the builtin function with builtin function code @var{fn}, or
5638 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5639 when @var{fn} is a code of a machine-dependent builtin function. When
5640 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5641 of a square root function are performed, and only reciprocals of @code{sqrt}
5645 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5646 This hook should return the DECL of a function @var{f} that given an
5647 address @var{addr} as an argument returns a mask @var{m} that can be
5648 used to extract from two vectors the relevant data that resides in
5649 @var{addr} in case @var{addr} is not properly aligned.
5651 The autovectorizer, when vectorizing a load operation from an address
5652 @var{addr} that may be unaligned, will generate two vector loads from
5653 the two aligned addresses around @var{addr}. It then generates a
5654 @code{REALIGN_LOAD} operation to extract the relevant data from the
5655 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5656 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5657 the third argument, @var{OFF}, defines how the data will be extracted
5658 from these two vectors: if @var{OFF} is 0, then the returned vector is
5659 @var{v2}; otherwise, the returned vector is composed from the last
5660 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5661 @var{OFF} elements of @var{v2}.
5663 If this hook is defined, the autovectorizer will generate a call
5664 to @var{f} (using the DECL tree that this hook returns) and will
5665 use the return value of @var{f} as the argument @var{OFF} to
5666 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5667 should comply with the semantics expected by @code{REALIGN_LOAD}
5669 If this hook is not defined, then @var{addr} will be used as
5670 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5671 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5674 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5675 This hook should return the DECL of a function @var{f} that implements
5676 widening multiplication of the even elements of two input vectors of type @var{x}.
5678 If this hook is defined, the autovectorizer will use it along with the
5679 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5680 widening multiplication in cases that the order of the results does not have to be
5681 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5682 @code{widen_mult_hi/lo} idioms will be used.
5685 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5686 This hook should return the DECL of a function @var{f} that implements
5687 widening multiplication of the odd elements of two input vectors of type @var{x}.
5689 If this hook is defined, the autovectorizer will use it along with the
5690 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5691 widening multiplication in cases that the order of the results does not have to be
5692 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5693 @code{widen_mult_hi/lo} idioms will be used.
5696 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5697 Returns cost of different scalar or vector statements for vectorization cost model.
5698 For vector memory operations the cost may depend on type (@var{vectype}) and
5699 misalignment value (@var{misalign}).
5702 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5703 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5706 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VEC_PERM (tree @var{type}, tree *@var{mask_element_type})
5707 Target builtin that implements vector permute.
5710 @deftypefn {Target Hook} bool TARGET_VECTORIZE_BUILTIN_VEC_PERM_OK (tree @var{vec_type}, tree @var{mask})
5711 Return true if a vector created for @code{builtin_vec_perm} is valid.
5714 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5715 This hook should return the DECL of a function that implements conversion of the
5716 input vector of type @var{src_type} to type @var{dest_type}.
5717 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5718 specifies how the conversion is to be applied
5719 (truncation, rounding, etc.).
5721 If this hook is defined, the autovectorizer will use the
5722 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5723 conversion. Otherwise, it will return @code{NULL_TREE}.
5726 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5727 This hook should return the decl of a function that implements the
5728 vectorized variant of the builtin function with builtin function code
5729 @var{code} or @code{NULL_TREE} if such a function is not available.
5730 The value of @var{fndecl} is the builtin function declaration. The
5731 return type of the vectorized function shall be of vector type
5732 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5735 @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})
5736 This hook should return true if the target supports misaligned vector
5737 store/load of a specific factor denoted in the @var{misalignment}
5738 parameter. The vector store/load should be of machine mode @var{mode} and
5739 the elements in the vectors should be of type @var{type}. @var{is_packed}
5740 parameter is true if the memory access is defined in a packed struct.
5743 @deftypefn {Target Hook} {enum machine_mode} TARGET_VECTORIZE_PREFERRED_SIMD_MODE (enum machine_mode @var{mode})
5744 This hook should return the preferred mode for vectorizing scalar
5745 mode @var{mode}. The default is
5746 equal to @code{word_mode}, because the vectorizer can do some
5747 transformations even in absence of specialized @acronym{SIMD} hardware.
5750 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void)
5751 This hook should return a mask of sizes that should be iterated over
5752 after trying to autovectorize using the vector size derived from the
5753 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5754 The default is zero which means to not iterate over other vector sizes.
5757 @node Anchored Addresses
5758 @section Anchored Addresses
5759 @cindex anchored addresses
5760 @cindex @option{-fsection-anchors}
5762 GCC usually addresses every static object as a separate entity.
5763 For example, if we have:
5767 int foo (void) @{ return a + b + c; @}
5770 the code for @code{foo} will usually calculate three separate symbolic
5771 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5772 it would be better to calculate just one symbolic address and access
5773 the three variables relative to it. The equivalent pseudocode would
5779 register int *xr = &x;
5780 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5784 (which isn't valid C). We refer to shared addresses like @code{x} as
5785 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5787 The hooks below describe the target properties that GCC needs to know
5788 in order to make effective use of section anchors. It won't use
5789 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5790 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5792 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5793 The minimum offset that should be applied to a section anchor.
5794 On most targets, it should be the smallest offset that can be
5795 applied to a base register while still giving a legitimate address
5796 for every mode. The default value is 0.
5799 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5800 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5801 offset that should be applied to section anchors. The default
5805 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5806 Write the assembly code to define section anchor @var{x}, which is a
5807 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5808 The hook is called with the assembly output position set to the beginning
5809 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5811 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5812 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5813 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5814 is @code{NULL}, which disables the use of section anchors altogether.
5817 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
5818 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5819 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5820 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5822 The default version is correct for most targets, but you might need to
5823 intercept this hook to handle things like target-specific attributes
5824 or target-specific sections.
5827 @node Condition Code
5828 @section Condition Code Status
5829 @cindex condition code status
5831 The macros in this section can be split in two families, according to the
5832 two ways of representing condition codes in GCC.
5834 The first representation is the so called @code{(cc0)} representation
5835 (@pxref{Jump Patterns}), where all instructions can have an implicit
5836 clobber of the condition codes. The second is the condition code
5837 register representation, which provides better schedulability for
5838 architectures that do have a condition code register, but on which
5839 most instructions do not affect it. The latter category includes
5842 The implicit clobbering poses a strong restriction on the placement of
5843 the definition and use of the condition code, which need to be in adjacent
5844 insns for machines using @code{(cc0)}. This can prevent important
5845 optimizations on some machines. For example, on the IBM RS/6000, there
5846 is a delay for taken branches unless the condition code register is set
5847 three instructions earlier than the conditional branch. The instruction
5848 scheduler cannot perform this optimization if it is not permitted to
5849 separate the definition and use of the condition code register.
5851 For this reason, it is possible and suggested to use a register to
5852 represent the condition code for new ports. If there is a specific
5853 condition code register in the machine, use a hard register. If the
5854 condition code or comparison result can be placed in any general register,
5855 or if there are multiple condition registers, use a pseudo register.
5856 Registers used to store the condition code value will usually have a mode
5857 that is in class @code{MODE_CC}.
5859 Alternatively, you can use @code{BImode} if the comparison operator is
5860 specified already in the compare instruction. In this case, you are not
5861 interested in most macros in this section.
5864 * CC0 Condition Codes:: Old style representation of condition codes.
5865 * MODE_CC Condition Codes:: Modern representation of condition codes.
5866 * Cond Exec Macros:: Macros to control conditional execution.
5869 @node CC0 Condition Codes
5870 @subsection Representation of condition codes using @code{(cc0)}
5874 The file @file{conditions.h} defines a variable @code{cc_status} to
5875 describe how the condition code was computed (in case the interpretation of
5876 the condition code depends on the instruction that it was set by). This
5877 variable contains the RTL expressions on which the condition code is
5878 currently based, and several standard flags.
5880 Sometimes additional machine-specific flags must be defined in the machine
5881 description header file. It can also add additional machine-specific
5882 information by defining @code{CC_STATUS_MDEP}.
5884 @defmac CC_STATUS_MDEP
5885 C code for a data type which is used for declaring the @code{mdep}
5886 component of @code{cc_status}. It defaults to @code{int}.
5888 This macro is not used on machines that do not use @code{cc0}.
5891 @defmac CC_STATUS_MDEP_INIT
5892 A C expression to initialize the @code{mdep} field to ``empty''.
5893 The default definition does nothing, since most machines don't use
5894 the field anyway. If you want to use the field, you should probably
5895 define this macro to initialize it.
5897 This macro is not used on machines that do not use @code{cc0}.
5900 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5901 A C compound statement to set the components of @code{cc_status}
5902 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5903 this macro's responsibility to recognize insns that set the condition
5904 code as a byproduct of other activity as well as those that explicitly
5907 This macro is not used on machines that do not use @code{cc0}.
5909 If there are insns that do not set the condition code but do alter
5910 other machine registers, this macro must check to see whether they
5911 invalidate the expressions that the condition code is recorded as
5912 reflecting. For example, on the 68000, insns that store in address
5913 registers do not set the condition code, which means that usually
5914 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5915 insns. But suppose that the previous insn set the condition code
5916 based on location @samp{a4@@(102)} and the current insn stores a new
5917 value in @samp{a4}. Although the condition code is not changed by
5918 this, it will no longer be true that it reflects the contents of
5919 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5920 @code{cc_status} in this case to say that nothing is known about the
5921 condition code value.
5923 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5924 with the results of peephole optimization: insns whose patterns are
5925 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5926 constants which are just the operands. The RTL structure of these
5927 insns is not sufficient to indicate what the insns actually do. What
5928 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5929 @code{CC_STATUS_INIT}.
5931 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5932 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5933 @samp{cc}. This avoids having detailed information about patterns in
5934 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5937 @node MODE_CC Condition Codes
5938 @subsection Representation of condition codes using registers
5942 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5943 On many machines, the condition code may be produced by other instructions
5944 than compares, for example the branch can use directly the condition
5945 code set by a subtract instruction. However, on some machines
5946 when the condition code is set this way some bits (such as the overflow
5947 bit) are not set in the same way as a test instruction, so that a different
5948 branch instruction must be used for some conditional branches. When
5949 this happens, use the machine mode of the condition code register to
5950 record different formats of the condition code register. Modes can
5951 also be used to record which compare instruction (e.g. a signed or an
5952 unsigned comparison) produced the condition codes.
5954 If other modes than @code{CCmode} are required, add them to
5955 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
5956 a mode given an operand of a compare. This is needed because the modes
5957 have to be chosen not only during RTL generation but also, for example,
5958 by instruction combination. The result of @code{SELECT_CC_MODE} should
5959 be consistent with the mode used in the patterns; for example to support
5960 the case of the add on the SPARC discussed above, we have the pattern
5964 [(set (reg:CC_NOOV 0)
5966 (plus:SI (match_operand:SI 0 "register_operand" "%r")
5967 (match_operand:SI 1 "arith_operand" "rI"))
5974 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
5975 for comparisons whose argument is a @code{plus}:
5978 #define SELECT_CC_MODE(OP,X,Y) \
5979 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5980 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5981 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5982 || GET_CODE (X) == NEG) \
5983 ? CC_NOOVmode : CCmode))
5986 Another reason to use modes is to retain information on which operands
5987 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
5990 You should define this macro if and only if you define extra CC modes
5991 in @file{@var{machine}-modes.def}.
5994 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5995 On some machines not all possible comparisons are defined, but you can
5996 convert an invalid comparison into a valid one. For example, the Alpha
5997 does not have a @code{GT} comparison, but you can use an @code{LT}
5998 comparison instead and swap the order of the operands.
6000 On such machines, define this macro to be a C statement to do any
6001 required conversions. @var{code} is the initial comparison code
6002 and @var{op0} and @var{op1} are the left and right operands of the
6003 comparison, respectively. You should modify @var{code}, @var{op0}, and
6004 @var{op1} as required.
6006 GCC will not assume that the comparison resulting from this macro is
6007 valid but will see if the resulting insn matches a pattern in the
6010 You need not define this macro if it would never change the comparison
6014 @defmac REVERSIBLE_CC_MODE (@var{mode})
6015 A C expression whose value is one if it is always safe to reverse a
6016 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6017 can ever return @var{mode} for a floating-point inequality comparison,
6018 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6020 You need not define this macro if it would always returns zero or if the
6021 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6022 For example, here is the definition used on the SPARC, where floating-point
6023 inequality comparisons are always given @code{CCFPEmode}:
6026 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
6030 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6031 A C expression whose value is reversed condition code of the @var{code} for
6032 comparison done in CC_MODE @var{mode}. The macro is used only in case
6033 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6034 machine has some non-standard way how to reverse certain conditionals. For
6035 instance in case all floating point conditions are non-trapping, compiler may
6036 freely convert unordered compares to ordered one. Then definition may look
6040 #define REVERSE_CONDITION(CODE, MODE) \
6041 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6042 : reverse_condition_maybe_unordered (CODE))
6046 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6047 On targets which do not use @code{(cc0)}, and which use a hard
6048 register rather than a pseudo-register to hold condition codes, the
6049 regular CSE passes are often not able to identify cases in which the
6050 hard register is set to a common value. Use this hook to enable a
6051 small pass which optimizes such cases. This hook should return true
6052 to enable this pass, and it should set the integers to which its
6053 arguments point to the hard register numbers used for condition codes.
6054 When there is only one such register, as is true on most systems, the
6055 integer pointed to by @var{p2} should be set to
6056 @code{INVALID_REGNUM}.
6058 The default version of this hook returns false.
6061 @deftypefn {Target Hook} {enum machine_mode} TARGET_CC_MODES_COMPATIBLE (enum machine_mode @var{m1}, enum machine_mode @var{m2})
6062 On targets which use multiple condition code modes in class
6063 @code{MODE_CC}, it is sometimes the case that a comparison can be
6064 validly done in more than one mode. On such a system, define this
6065 target hook to take two mode arguments and to return a mode in which
6066 both comparisons may be validly done. If there is no such mode,
6067 return @code{VOIDmode}.
6069 The default version of this hook checks whether the modes are the
6070 same. If they are, it returns that mode. If they are different, it
6071 returns @code{VOIDmode}.
6074 @node Cond Exec Macros
6075 @subsection Macros to control conditional execution
6076 @findex conditional execution
6079 There is one macro that may need to be defined for targets
6080 supporting conditional execution, independent of how they
6081 represent conditional branches.
6083 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6084 A C expression that returns true if the conditional execution predicate
6085 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6086 versa. Define this to return 0 if the target has conditional execution
6087 predicates that cannot be reversed safely. There is no need to validate
6088 that the arguments of op1 and op2 are the same, this is done separately.
6089 If no expansion is specified, this macro is defined as follows:
6092 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6093 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6098 @section Describing Relative Costs of Operations
6099 @cindex costs of instructions
6100 @cindex relative costs
6101 @cindex speed of instructions
6103 These macros let you describe the relative speed of various operations
6104 on the target machine.
6106 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6107 A C expression for the cost of moving data of mode @var{mode} from a
6108 register in class @var{from} to one in class @var{to}. The classes are
6109 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6110 value of 2 is the default; other values are interpreted relative to
6113 It is not required that the cost always equal 2 when @var{from} is the
6114 same as @var{to}; on some machines it is expensive to move between
6115 registers if they are not general registers.
6117 If reload sees an insn consisting of a single @code{set} between two
6118 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6119 classes returns a value of 2, reload does not check to ensure that the
6120 constraints of the insn are met. Setting a cost of other than 2 will
6121 allow reload to verify that the constraints are met. You should do this
6122 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6124 These macros are obsolete, new ports should use the target hook
6125 @code{TARGET_REGISTER_MOVE_COST} instead.
6128 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6129 This target hook should return the cost of moving data of mode @var{mode}
6130 from a register in class @var{from} to one in class @var{to}. The classes
6131 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6132 A value of 2 is the default; other values are interpreted relative to
6135 It is not required that the cost always equal 2 when @var{from} is the
6136 same as @var{to}; on some machines it is expensive to move between
6137 registers if they are not general registers.
6139 If reload sees an insn consisting of a single @code{set} between two
6140 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6141 classes returns a value of 2, reload does not check to ensure that the
6142 constraints of the insn are met. Setting a cost of other than 2 will
6143 allow reload to verify that the constraints are met. You should do this
6144 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6146 The default version of this function returns 2.
6149 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6150 A C expression for the cost of moving data of mode @var{mode} between a
6151 register of class @var{class} and memory; @var{in} is zero if the value
6152 is to be written to memory, nonzero if it is to be read in. This cost
6153 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6154 registers and memory is more expensive than between two registers, you
6155 should define this macro to express the relative cost.
6157 If you do not define this macro, GCC uses a default cost of 4 plus
6158 the cost of copying via a secondary reload register, if one is
6159 needed. If your machine requires a secondary reload register to copy
6160 between memory and a register of @var{class} but the reload mechanism is
6161 more complex than copying via an intermediate, define this macro to
6162 reflect the actual cost of the move.
6164 GCC defines the function @code{memory_move_secondary_cost} if
6165 secondary reloads are needed. It computes the costs due to copying via
6166 a secondary register. If your machine copies from memory using a
6167 secondary register in the conventional way but the default base value of
6168 4 is not correct for your machine, define this macro to add some other
6169 value to the result of that function. The arguments to that function
6170 are the same as to this macro.
6172 These macros are obsolete, new ports should use the target hook
6173 @code{TARGET_MEMORY_MOVE_COST} instead.
6176 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6177 This target hook should return the cost of moving data of mode @var{mode}
6178 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6179 if the value is to be written to memory, @code{true} if it is to be read in.
6180 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6181 If moving between registers and memory is more expensive than between two
6182 registers, you should add this target hook to express the relative cost.
6184 If you do not add this target hook, GCC uses a default cost of 4 plus
6185 the cost of copying via a secondary reload register, if one is
6186 needed. If your machine requires a secondary reload register to copy
6187 between memory and a register of @var{rclass} but the reload mechanism is
6188 more complex than copying via an intermediate, use this target hook to
6189 reflect the actual cost of the move.
6191 GCC defines the function @code{memory_move_secondary_cost} if
6192 secondary reloads are needed. It computes the costs due to copying via
6193 a secondary register. If your machine copies from memory using a
6194 secondary register in the conventional way but the default base value of
6195 4 is not correct for your machine, use this target hook to add some other
6196 value to the result of that function. The arguments to that function
6197 are the same as to this target hook.
6200 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6201 A C expression for the cost of a branch instruction. A value of 1 is
6202 the default; other values are interpreted relative to that. Parameter
6203 @var{speed_p} is true when the branch in question should be optimized
6204 for speed. When it is false, @code{BRANCH_COST} should return a value
6205 optimal for code size rather than performance. @var{predictable_p} is
6206 true for well-predicted branches. On many architectures the
6207 @code{BRANCH_COST} can be reduced then.
6210 Here are additional macros which do not specify precise relative costs,
6211 but only that certain actions are more expensive than GCC would
6214 @defmac SLOW_BYTE_ACCESS
6215 Define this macro as a C expression which is nonzero if accessing less
6216 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6217 faster than accessing a word of memory, i.e., if such access
6218 require more than one instruction or if there is no difference in cost
6219 between byte and (aligned) word loads.
6221 When this macro is not defined, the compiler will access a field by
6222 finding the smallest containing object; when it is defined, a fullword
6223 load will be used if alignment permits. Unless bytes accesses are
6224 faster than word accesses, using word accesses is preferable since it
6225 may eliminate subsequent memory access if subsequent accesses occur to
6226 other fields in the same word of the structure, but to different bytes.
6229 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6230 Define this macro to be the value 1 if memory accesses described by the
6231 @var{mode} and @var{alignment} parameters have a cost many times greater
6232 than aligned accesses, for example if they are emulated in a trap
6235 When this macro is nonzero, the compiler will act as if
6236 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6237 moves. This can cause significantly more instructions to be produced.
6238 Therefore, do not set this macro nonzero if unaligned accesses only add a
6239 cycle or two to the time for a memory access.
6241 If the value of this macro is always zero, it need not be defined. If
6242 this macro is defined, it should produce a nonzero value when
6243 @code{STRICT_ALIGNMENT} is nonzero.
6246 @defmac MOVE_RATIO (@var{speed})
6247 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6248 which a sequence of insns should be generated instead of a
6249 string move insn or a library call. Increasing the value will always
6250 make code faster, but eventually incurs high cost in increased code size.
6252 Note that on machines where the corresponding move insn is a
6253 @code{define_expand} that emits a sequence of insns, this macro counts
6254 the number of such sequences.
6256 The parameter @var{speed} is true if the code is currently being
6257 optimized for speed rather than size.
6259 If you don't define this, a reasonable default is used.
6262 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6263 A C expression used to determine whether @code{move_by_pieces} will be used to
6264 copy a chunk of memory, or whether some other block move mechanism
6265 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6266 than @code{MOVE_RATIO}.
6269 @defmac MOVE_MAX_PIECES
6270 A C expression used by @code{move_by_pieces} to determine the largest unit
6271 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6274 @defmac CLEAR_RATIO (@var{speed})
6275 The threshold of number of scalar move insns, @emph{below} which a sequence
6276 of insns should be generated to clear memory instead of a string clear insn
6277 or a library call. Increasing the value will always make code faster, but
6278 eventually incurs high cost in increased code size.
6280 The parameter @var{speed} is true if the code is currently being
6281 optimized for speed rather than size.
6283 If you don't define this, a reasonable default is used.
6286 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6287 A C expression used to determine whether @code{clear_by_pieces} will be used
6288 to clear a chunk of memory, or whether some other block clear mechanism
6289 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6290 than @code{CLEAR_RATIO}.
6293 @defmac SET_RATIO (@var{speed})
6294 The threshold of number of scalar move insns, @emph{below} which a sequence
6295 of insns should be generated to set memory to a constant value, instead of
6296 a block set insn or a library call.
6297 Increasing the value will always make code faster, but
6298 eventually incurs high cost in increased code size.
6300 The parameter @var{speed} is true if the code is currently being
6301 optimized for speed rather than size.
6303 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6306 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6307 A C expression used to determine whether @code{store_by_pieces} will be
6308 used to set a chunk of memory to a constant value, or whether some
6309 other mechanism will be used. Used by @code{__builtin_memset} when
6310 storing values other than constant zero.
6311 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6312 than @code{SET_RATIO}.
6315 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6316 A C expression used to determine whether @code{store_by_pieces} will be
6317 used to set a chunk of memory to a constant string value, or whether some
6318 other mechanism will be used. Used by @code{__builtin_strcpy} when
6319 called with a constant source string.
6320 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6321 than @code{MOVE_RATIO}.
6324 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6325 A C expression used to determine whether a load postincrement is a good
6326 thing to use for a given mode. Defaults to the value of
6327 @code{HAVE_POST_INCREMENT}.
6330 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6331 A C expression used to determine whether a load postdecrement is a good
6332 thing to use for a given mode. Defaults to the value of
6333 @code{HAVE_POST_DECREMENT}.
6336 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6337 A C expression used to determine whether a load preincrement is a good
6338 thing to use for a given mode. Defaults to the value of
6339 @code{HAVE_PRE_INCREMENT}.
6342 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6343 A C expression used to determine whether a load predecrement is a good
6344 thing to use for a given mode. Defaults to the value of
6345 @code{HAVE_PRE_DECREMENT}.
6348 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6349 A C expression used to determine whether a store postincrement is a good
6350 thing to use for a given mode. Defaults to the value of
6351 @code{HAVE_POST_INCREMENT}.
6354 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6355 A C expression used to determine whether a store postdecrement is a good
6356 thing to use for a given mode. Defaults to the value of
6357 @code{HAVE_POST_DECREMENT}.
6360 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6361 This macro is used to determine whether a store preincrement is a good
6362 thing to use for a given mode. Defaults to the value of
6363 @code{HAVE_PRE_INCREMENT}.
6366 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6367 This macro is used to determine whether a store predecrement is a good
6368 thing to use for a given mode. Defaults to the value of
6369 @code{HAVE_PRE_DECREMENT}.
6372 @defmac NO_FUNCTION_CSE
6373 Define this macro if it is as good or better to call a constant
6374 function address than to call an address kept in a register.
6377 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6378 Define this macro if a non-short-circuit operation produced by
6379 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6380 @code{BRANCH_COST} is greater than or equal to the value 2.
6383 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total}, bool @var{speed})
6384 This target hook describes the relative costs of RTL expressions.
6386 The cost may depend on the precise form of the expression, which is
6387 available for examination in @var{x}, and the rtx code of the expression
6388 in which it is contained, found in @var{outer_code}. @var{code} is the
6389 expression code---redundant, since it can be obtained with
6390 @code{GET_CODE (@var{x})}.
6392 In implementing this hook, you can use the construct
6393 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6396 On entry to the hook, @code{*@var{total}} contains a default estimate
6397 for the cost of the expression. The hook should modify this value as
6398 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6399 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6400 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6402 When optimizing for code size, i.e.@: when @code{speed} is
6403 false, this target hook should be used to estimate the relative
6404 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6406 The hook returns true when all subexpressions of @var{x} have been
6407 processed, and false when @code{rtx_cost} should recurse.
6410 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, bool @var{speed})
6411 This hook computes the cost of an addressing mode that contains
6412 @var{address}. If not defined, the cost is computed from
6413 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6415 For most CISC machines, the default cost is a good approximation of the
6416 true cost of the addressing mode. However, on RISC machines, all
6417 instructions normally have the same length and execution time. Hence
6418 all addresses will have equal costs.
6420 In cases where more than one form of an address is known, the form with
6421 the lowest cost will be used. If multiple forms have the same, lowest,
6422 cost, the one that is the most complex will be used.
6424 For example, suppose an address that is equal to the sum of a register
6425 and a constant is used twice in the same basic block. When this macro
6426 is not defined, the address will be computed in a register and memory
6427 references will be indirect through that register. On machines where
6428 the cost of the addressing mode containing the sum is no higher than
6429 that of a simple indirect reference, this will produce an additional
6430 instruction and possibly require an additional register. Proper
6431 specification of this macro eliminates this overhead for such machines.
6433 This hook is never called with an invalid address.
6435 On machines where an address involving more than one register is as
6436 cheap as an address computation involving only one register, defining
6437 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6438 be live over a region of code where only one would have been if
6439 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6440 should be considered in the definition of this macro. Equivalent costs
6441 should probably only be given to addresses with different numbers of
6442 registers on machines with lots of registers.
6446 @section Adjusting the Instruction Scheduler
6448 The instruction scheduler may need a fair amount of machine-specific
6449 adjustment in order to produce good code. GCC provides several target
6450 hooks for this purpose. It is usually enough to define just a few of
6451 them: try the first ones in this list first.
6453 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6454 This hook returns the maximum number of instructions that can ever
6455 issue at the same time on the target machine. The default is one.
6456 Although the insn scheduler can define itself the possibility of issue
6457 an insn on the same cycle, the value can serve as an additional
6458 constraint to issue insns on the same simulated processor cycle (see
6459 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6460 This value must be constant over the entire compilation. If you need
6461 it to vary depending on what the instructions are, you must use
6462 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6465 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6466 This hook is executed by the scheduler after it has scheduled an insn
6467 from the ready list. It should return the number of insns which can
6468 still be issued in the current cycle. The default is
6469 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6470 @code{USE}, which normally are not counted against the issue rate.
6471 You should define this hook if some insns take more machine resources
6472 than others, so that fewer insns can follow them in the same cycle.
6473 @var{file} is either a null pointer, or a stdio stream to write any
6474 debug output to. @var{verbose} is the verbose level provided by
6475 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6479 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6480 This function corrects the value of @var{cost} based on the
6481 relationship between @var{insn} and @var{dep_insn} through the
6482 dependence @var{link}. It should return the new value. The default
6483 is to make no adjustment to @var{cost}. This can be used for example
6484 to specify to the scheduler using the traditional pipeline description
6485 that an output- or anti-dependence does not incur the same cost as a
6486 data-dependence. If the scheduler using the automaton based pipeline
6487 description, the cost of anti-dependence is zero and the cost of
6488 output-dependence is maximum of one and the difference of latency
6489 times of the first and the second insns. If these values are not
6490 acceptable, you could use the hook to modify them too. See also
6491 @pxref{Processor pipeline description}.
6494 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6495 This hook adjusts the integer scheduling priority @var{priority} of
6496 @var{insn}. It should return the new priority. Increase the priority to
6497 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6498 later. Do not define this hook if you do not need to adjust the
6499 scheduling priorities of insns.
6502 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6503 This hook is executed by the scheduler after it has scheduled the ready
6504 list, to allow the machine description to reorder it (for example to
6505 combine two small instructions together on @samp{VLIW} machines).
6506 @var{file} is either a null pointer, or a stdio stream to write any
6507 debug output to. @var{verbose} is the verbose level provided by
6508 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6509 list of instructions that are ready to be scheduled. @var{n_readyp} is
6510 a pointer to the number of elements in the ready list. The scheduler
6511 reads the ready list in reverse order, starting with
6512 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6513 is the timer tick of the scheduler. You may modify the ready list and
6514 the number of ready insns. The return value is the number of insns that
6515 can issue this cycle; normally this is just @code{issue_rate}. See also
6516 @samp{TARGET_SCHED_REORDER2}.
6519 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6520 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6521 function is called whenever the scheduler starts a new cycle. This one
6522 is called once per iteration over a cycle, immediately after
6523 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6524 return the number of insns to be scheduled in the same cycle. Defining
6525 this hook can be useful if there are frequent situations where
6526 scheduling one insn causes other insns to become ready in the same
6527 cycle. These other insns can then be taken into account properly.
6530 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6531 This hook is called after evaluation forward dependencies of insns in
6532 chain given by two parameter values (@var{head} and @var{tail}
6533 correspondingly) but before insns scheduling of the insn chain. For
6534 example, it can be used for better insn classification if it requires
6535 analysis of dependencies. This hook can use backward and forward
6536 dependencies of the insn scheduler because they are already
6540 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6541 This hook is executed by the scheduler at the beginning of each block of
6542 instructions that are to be scheduled. @var{file} is either a null
6543 pointer, or a stdio stream to write any debug output to. @var{verbose}
6544 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6545 @var{max_ready} is the maximum number of insns in the current scheduling
6546 region that can be live at the same time. This can be used to allocate
6547 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6550 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6551 This hook is executed by the scheduler at the end of each block of
6552 instructions that are to be scheduled. It can be used to perform
6553 cleanup of any actions done by the other scheduling hooks. @var{file}
6554 is either a null pointer, or a stdio stream to write any debug output
6555 to. @var{verbose} is the verbose level provided by
6556 @option{-fsched-verbose-@var{n}}.
6559 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6560 This hook is executed by the scheduler after function level initializations.
6561 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6562 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6563 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6566 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6567 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6568 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6569 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6572 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6573 The hook returns an RTL insn. The automaton state used in the
6574 pipeline hazard recognizer is changed as if the insn were scheduled
6575 when the new simulated processor cycle starts. Usage of the hook may
6576 simplify the automaton pipeline description for some @acronym{VLIW}
6577 processors. If the hook is defined, it is used only for the automaton
6578 based pipeline description. The default is not to change the state
6579 when the new simulated processor cycle starts.
6582 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6583 The hook can be used to initialize data used by the previous hook.
6586 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6587 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6588 to changed the state as if the insn were scheduled when the new
6589 simulated processor cycle finishes.
6592 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6593 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6594 used to initialize data used by the previous hook.
6597 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
6598 The hook to notify target that the current simulated cycle is about to finish.
6599 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6600 to change the state in more complicated situations - e.g., when advancing
6601 state on a single insn is not enough.
6604 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
6605 The hook to notify target that new simulated cycle has just started.
6606 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6607 to change the state in more complicated situations - e.g., when advancing
6608 state on a single insn is not enough.
6611 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6612 This hook controls better choosing an insn from the ready insn queue
6613 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6614 chooses the first insn from the queue. If the hook returns a positive
6615 value, an additional scheduler code tries all permutations of
6616 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6617 subsequent ready insns to choose an insn whose issue will result in
6618 maximal number of issued insns on the same cycle. For the
6619 @acronym{VLIW} processor, the code could actually solve the problem of
6620 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6621 rules of @acronym{VLIW} packing are described in the automaton.
6623 This code also could be used for superscalar @acronym{RISC}
6624 processors. Let us consider a superscalar @acronym{RISC} processor
6625 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6626 @var{B}, some insns can be executed only in pipelines @var{B} or
6627 @var{C}, and one insn can be executed in pipeline @var{B}. The
6628 processor may issue the 1st insn into @var{A} and the 2nd one into
6629 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6630 until the next cycle. If the scheduler issues the 3rd insn the first,
6631 the processor could issue all 3 insns per cycle.
6633 Actually this code demonstrates advantages of the automaton based
6634 pipeline hazard recognizer. We try quickly and easy many insn
6635 schedules to choose the best one.
6637 The default is no multipass scheduling.
6640 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx @var{insn})
6642 This hook controls what insns from the ready insn queue will be
6643 considered for the multipass insn scheduling. If the hook returns
6644 zero for @var{insn}, the insn will be not chosen to
6647 The default is that any ready insns can be chosen to be issued.
6650 @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})
6651 This hook prepares the target backend for a new round of multipass
6655 @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})
6656 This hook is called when multipass scheduling evaluates instruction INSN.
6659 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, char *@var{ready_try}, int @var{n_ready})
6660 This is called when multipass scheduling backtracks from evaluation of
6664 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
6665 This hook notifies the target about the result of the concluded current
6666 round of multipass scheduling.
6669 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
6670 This hook initializes target-specific data used in multipass scheduling.
6673 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
6674 This hook finalizes target-specific data used in multipass scheduling.
6677 @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})
6678 This hook is called by the insn scheduler before issuing @var{insn}
6679 on cycle @var{clock}. If the hook returns nonzero,
6680 @var{insn} is not issued on this processor cycle. Instead,
6681 the processor cycle is advanced. If *@var{sort_p}
6682 is zero, the insn ready queue is not sorted on the new cycle
6683 start as usually. @var{dump} and @var{verbose} specify the file and
6684 verbosity level to use for debugging output.
6685 @var{last_clock} and @var{clock} are, respectively, the
6686 processor cycle on which the previous insn has been issued,
6687 and the current processor cycle.
6690 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
6691 This hook is used to define which dependences are considered costly by
6692 the target, so costly that it is not advisable to schedule the insns that
6693 are involved in the dependence too close to one another. The parameters
6694 to this hook are as follows: The first parameter @var{_dep} is the dependence
6695 being evaluated. The second parameter @var{cost} is the cost of the
6696 dependence as estimated by the scheduler, and the third
6697 parameter @var{distance} is the distance in cycles between the two insns.
6698 The hook returns @code{true} if considering the distance between the two
6699 insns the dependence between them is considered costly by the target,
6700 and @code{false} otherwise.
6702 Defining this hook can be useful in multiple-issue out-of-order machines,
6703 where (a) it's practically hopeless to predict the actual data/resource
6704 delays, however: (b) there's a better chance to predict the actual grouping
6705 that will be formed, and (c) correctly emulating the grouping can be very
6706 important. In such targets one may want to allow issuing dependent insns
6707 closer to one another---i.e., closer than the dependence distance; however,
6708 not in cases of ``costly dependences'', which this hooks allows to define.
6711 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6712 This hook is called by the insn scheduler after emitting a new instruction to
6713 the instruction stream. The hook notifies a target backend to extend its
6714 per instruction data structures.
6717 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6718 Return a pointer to a store large enough to hold target scheduling context.
6721 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6722 Initialize store pointed to by @var{tc} to hold target scheduling context.
6723 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6724 beginning of the block. Otherwise, copy the current context into @var{tc}.
6727 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6728 Copy target scheduling context pointed to by @var{tc} to the current context.
6731 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6732 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6735 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6736 Deallocate a store for target scheduling context pointed to by @var{tc}.
6739 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6740 This hook is called by the insn scheduler when @var{insn} has only
6741 speculative dependencies and therefore can be scheduled speculatively.
6742 The hook is used to check if the pattern of @var{insn} has a speculative
6743 version and, in case of successful check, to generate that speculative
6744 pattern. The hook should return 1, if the instruction has a speculative form,
6745 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6746 speculation. If the return value equals 1 then @var{new_pat} is assigned
6747 the generated speculative pattern.
6750 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (int @var{dep_status})
6751 This hook is called by the insn scheduler during generation of recovery code
6752 for @var{insn}. It should return @code{true}, if the corresponding check
6753 instruction should branch to recovery code, or @code{false} otherwise.
6756 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6757 This hook is called by the insn scheduler to generate a pattern for recovery
6758 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6759 speculative instruction for which the check should be generated.
6760 @var{label} is either a label of a basic block, where recovery code should
6761 be emitted, or a null pointer, when requested check doesn't branch to
6762 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6763 a pattern for a branchy check corresponding to a simple check denoted by
6764 @var{insn} should be generated. In this case @var{label} can't be null.
6767 @deftypefn {Target Hook} bool TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (const_rtx @var{insn})
6768 This hook is used as a workaround for
6769 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6770 called on the first instruction of the ready list. The hook is used to
6771 discard speculative instructions that stand first in the ready list from
6772 being scheduled on the current cycle. If the hook returns @code{false},
6773 @var{insn} will not be chosen to be issued.
6774 For non-speculative instructions,
6775 the hook should always return @code{true}. For example, in the ia64 backend
6776 the hook is used to cancel data speculative insns when the ALAT table
6780 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
6781 This hook is used by the insn scheduler to find out what features should be
6783 The structure *@var{spec_info} should be filled in by the target.
6784 The structure describes speculation types that can be used in the scheduler.
6787 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6788 This hook is called by the swing modulo scheduler to calculate a
6789 resource-based lower bound which is based on the resources available in
6790 the machine and the resources required by each instruction. The target
6791 backend can use @var{g} to calculate such bound. A very simple lower
6792 bound will be used in case this hook is not implemented: the total number
6793 of instructions divided by the issue rate.
6796 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx @var{insn}, int @var{x})
6797 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6798 is supported in hardware and the condition specified in the parameter is true.
6801 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx @var{insn}, int @var{x})
6802 This hook is called by Haifa Scheduler. It performs the operation specified
6803 in its second parameter.
6806 @deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE
6807 True if the processor has an exposed pipeline, which means that not just
6808 the order of instructions is important for correctness when scheduling, but
6809 also the latencies of operations.
6813 @section Dividing the Output into Sections (Texts, Data, @dots{})
6814 @c the above section title is WAY too long. maybe cut the part between
6815 @c the (...)? --mew 10feb93
6817 An object file is divided into sections containing different types of
6818 data. In the most common case, there are three sections: the @dfn{text
6819 section}, which holds instructions and read-only data; the @dfn{data
6820 section}, which holds initialized writable data; and the @dfn{bss
6821 section}, which holds uninitialized data. Some systems have other kinds
6824 @file{varasm.c} provides several well-known sections, such as
6825 @code{text_section}, @code{data_section} and @code{bss_section}.
6826 The normal way of controlling a @code{@var{foo}_section} variable
6827 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6828 as described below. The macros are only read once, when @file{varasm.c}
6829 initializes itself, so their values must be run-time constants.
6830 They may however depend on command-line flags.
6832 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6833 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6834 to be string literals.
6836 Some assemblers require a different string to be written every time a
6837 section is selected. If your assembler falls into this category, you
6838 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6839 @code{get_unnamed_section} to set up the sections.
6841 You must always create a @code{text_section}, either by defining
6842 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6843 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6844 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6845 create a distinct @code{readonly_data_section}, the default is to
6846 reuse @code{text_section}.
6848 All the other @file{varasm.c} sections are optional, and are null
6849 if the target does not provide them.
6851 @defmac TEXT_SECTION_ASM_OP
6852 A C expression whose value is a string, including spacing, containing the
6853 assembler operation that should precede instructions and read-only data.
6854 Normally @code{"\t.text"} is right.
6857 @defmac HOT_TEXT_SECTION_NAME
6858 If defined, a C string constant for the name of the section containing most
6859 frequently executed functions of the program. If not defined, GCC will provide
6860 a default definition if the target supports named sections.
6863 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6864 If defined, a C string constant for the name of the section containing unlikely
6865 executed functions in the program.
6868 @defmac DATA_SECTION_ASM_OP
6869 A C expression whose value is a string, including spacing, containing the
6870 assembler operation to identify the following data as writable initialized
6871 data. Normally @code{"\t.data"} is right.
6874 @defmac SDATA_SECTION_ASM_OP
6875 If defined, a C expression whose value is a string, including spacing,
6876 containing the assembler operation to identify the following data as
6877 initialized, writable small data.
6880 @defmac READONLY_DATA_SECTION_ASM_OP
6881 A C expression whose value is a string, including spacing, containing the
6882 assembler operation to identify the following data as read-only initialized
6886 @defmac BSS_SECTION_ASM_OP
6887 If defined, a C expression whose value is a string, including spacing,
6888 containing the assembler operation to identify the following data as
6889 uninitialized global data. If not defined, and
6890 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
6891 uninitialized global data will be output in the data section if
6892 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6896 @defmac SBSS_SECTION_ASM_OP
6897 If defined, a C expression whose value is a string, including spacing,
6898 containing the assembler operation to identify the following data as
6899 uninitialized, writable small data.
6902 @defmac TLS_COMMON_ASM_OP
6903 If defined, a C expression whose value is a string containing the
6904 assembler operation to identify the following data as thread-local
6905 common data. The default is @code{".tls_common"}.
6908 @defmac TLS_SECTION_ASM_FLAG
6909 If defined, a C expression whose value is a character constant
6910 containing the flag used to mark a section as a TLS section. The
6911 default is @code{'T'}.
6914 @defmac INIT_SECTION_ASM_OP
6915 If defined, a C expression whose value is a string, including spacing,
6916 containing the assembler operation to identify the following data as
6917 initialization code. If not defined, GCC will assume such a section does
6918 not exist. This section has no corresponding @code{init_section}
6919 variable; it is used entirely in runtime code.
6922 @defmac FINI_SECTION_ASM_OP
6923 If defined, a C expression whose value is a string, including spacing,
6924 containing the assembler operation to identify the following data as
6925 finalization code. If not defined, GCC will assume such a section does
6926 not exist. This section has no corresponding @code{fini_section}
6927 variable; it is used entirely in runtime code.
6930 @defmac INIT_ARRAY_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 part of the @code{.init_array} (or equivalent) section. If not
6934 defined, GCC will assume such a section does not exist. Do not define
6935 both this macro and @code{INIT_SECTION_ASM_OP}.
6938 @defmac FINI_ARRAY_SECTION_ASM_OP
6939 If defined, a C expression whose value is a string, including spacing,
6940 containing the assembler operation to identify the following data as
6941 part of the @code{.fini_array} (or equivalent) section. If not
6942 defined, GCC will assume such a section does not exist. Do not define
6943 both this macro and @code{FINI_SECTION_ASM_OP}.
6946 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6947 If defined, an ASM statement that switches to a different section
6948 via @var{section_op}, calls @var{function}, and switches back to
6949 the text section. This is used in @file{crtstuff.c} if
6950 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6951 to initialization and finalization functions from the init and fini
6952 sections. By default, this macro uses a simple function call. Some
6953 ports need hand-crafted assembly code to avoid dependencies on
6954 registers initialized in the function prologue or to ensure that
6955 constant pools don't end up too far way in the text section.
6958 @defmac TARGET_LIBGCC_SDATA_SECTION
6959 If defined, a string which names the section into which small
6960 variables defined in crtstuff and libgcc should go. This is useful
6961 when the target has options for optimizing access to small data, and
6962 you want the crtstuff and libgcc routines to be conservative in what
6963 they expect of your application yet liberal in what your application
6964 expects. For example, for targets with a @code{.sdata} section (like
6965 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6966 require small data support from your application, but use this macro
6967 to put small data into @code{.sdata} so that your application can
6968 access these variables whether it uses small data or not.
6971 @defmac FORCE_CODE_SECTION_ALIGN
6972 If defined, an ASM statement that aligns a code section to some
6973 arbitrary boundary. This is used to force all fragments of the
6974 @code{.init} and @code{.fini} sections to have to same alignment
6975 and thus prevent the linker from having to add any padding.
6978 @defmac JUMP_TABLES_IN_TEXT_SECTION
6979 Define this macro to be an expression with a nonzero value if jump
6980 tables (for @code{tablejump} insns) should be output in the text
6981 section, along with the assembler instructions. Otherwise, the
6982 readonly data section is used.
6984 This macro is irrelevant if there is no separate readonly data section.
6987 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6988 Define this hook if you need to do something special to set up the
6989 @file{varasm.c} sections, or if your target has some special sections
6990 of its own that you need to create.
6992 GCC calls this hook after processing the command line, but before writing
6993 any assembly code, and before calling any of the section-returning hooks
6997 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
6998 Return a mask describing how relocations should be treated when
6999 selecting sections. Bit 1 should be set if global relocations
7000 should be placed in a read-write section; bit 0 should be set if
7001 local relocations should be placed in a read-write section.
7003 The default version of this function returns 3 when @option{-fpic}
7004 is in effect, and 0 otherwise. The hook is typically redefined
7005 when the target cannot support (some kinds of) dynamic relocations
7006 in read-only sections even in executables.
7009 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
7010 Return the section into which @var{exp} should be placed. You can
7011 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
7012 some sort. @var{reloc} indicates whether the initial value of @var{exp}
7013 requires link-time relocations. Bit 0 is set when variable contains
7014 local relocations only, while bit 1 is set for global relocations.
7015 @var{align} is the constant alignment in bits.
7017 The default version of this function takes care of putting read-only
7018 variables in @code{readonly_data_section}.
7020 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7023 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7024 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7025 for @code{FUNCTION_DECL}s as well as for variables and constants.
7027 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7028 function has been determined to be likely to be called, and nonzero if
7029 it is unlikely to be called.
7032 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7033 Build up a unique section name, expressed as a @code{STRING_CST} node,
7034 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7035 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7036 the initial value of @var{exp} requires link-time relocations.
7038 The default version of this function appends the symbol name to the
7039 ELF section name that would normally be used for the symbol. For
7040 example, the function @code{foo} would be placed in @code{.text.foo}.
7041 Whatever the actual target object format, this is often good enough.
7044 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7045 Return the readonly data section associated with
7046 @samp{DECL_SECTION_NAME (@var{decl})}.
7047 The default version of this function selects @code{.gnu.linkonce.r.name} if
7048 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7049 if function is in @code{.text.name}, and the normal readonly-data section
7053 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7054 Return the section into which a constant @var{x}, of mode @var{mode},
7055 should be placed. You can assume that @var{x} is some kind of
7056 constant in RTL@. The argument @var{mode} is redundant except in the
7057 case of a @code{const_int} rtx. @var{align} is the constant alignment
7060 The default version of this function takes care of putting symbolic
7061 constants in @code{flag_pic} mode in @code{data_section} and everything
7062 else in @code{readonly_data_section}.
7065 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7066 Define this hook if you need to postprocess the assembler name generated
7067 by target-independent code. The @var{id} provided to this hook will be
7068 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7069 or the mangled name of the @var{decl} in C++). The return value of the
7070 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7071 your target system. The default implementation of this hook just
7072 returns the @var{id} provided.
7075 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7076 Define this hook if references to a symbol or a constant must be
7077 treated differently depending on something about the variable or
7078 function named by the symbol (such as what section it is in).
7080 The hook is executed immediately after rtl has been created for
7081 @var{decl}, which may be a variable or function declaration or
7082 an entry in the constant pool. In either case, @var{rtl} is the
7083 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7084 in this hook; that field may not have been initialized yet.
7086 In the case of a constant, it is safe to assume that the rtl is
7087 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7088 will also have this form, but that is not guaranteed. Global
7089 register variables, for instance, will have a @code{reg} for their
7090 rtl. (Normally the right thing to do with such unusual rtl is
7093 The @var{new_decl_p} argument will be true if this is the first time
7094 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7095 be false for subsequent invocations, which will happen for duplicate
7096 declarations. Whether or not anything must be done for the duplicate
7097 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7098 @var{new_decl_p} is always true when the hook is called for a constant.
7100 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7101 The usual thing for this hook to do is to record flags in the
7102 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7103 Historically, the name string was modified if it was necessary to
7104 encode more than one bit of information, but this practice is now
7105 discouraged; use @code{SYMBOL_REF_FLAGS}.
7107 The default definition of this hook, @code{default_encode_section_info}
7108 in @file{varasm.c}, sets a number of commonly-useful bits in
7109 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7110 before overriding it.
7113 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7114 Decode @var{name} and return the real name part, sans
7115 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7119 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7120 Returns true if @var{exp} should be placed into a ``small data'' section.
7121 The default version of this hook always returns false.
7124 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7125 Contains the value true if the target places read-only
7126 ``small data'' into a separate section. The default value is false.
7129 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7130 It returns true if target wants profile code emitted before prologue.
7132 The default version of this hook use the target macro
7133 @code{PROFILE_BEFORE_PROLOGUE}.
7136 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7137 Returns true if @var{exp} names an object for which name resolution
7138 rules must resolve to the current ``module'' (dynamic shared library
7139 or executable image).
7141 The default version of this hook implements the name resolution rules
7142 for ELF, which has a looser model of global name binding than other
7143 currently supported object file formats.
7146 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7147 Contains the value true if the target supports thread-local storage.
7148 The default value is false.
7153 @section Position Independent Code
7154 @cindex position independent code
7157 This section describes macros that help implement generation of position
7158 independent code. Simply defining these macros is not enough to
7159 generate valid PIC; you must also add support to the hook
7160 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7161 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7162 must modify the definition of @samp{movsi} to do something appropriate
7163 when the source operand contains a symbolic address. You may also
7164 need to alter the handling of switch statements so that they use
7166 @c i rearranged the order of the macros above to try to force one of
7167 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7169 @defmac PIC_OFFSET_TABLE_REGNUM
7170 The register number of the register used to address a table of static
7171 data addresses in memory. In some cases this register is defined by a
7172 processor's ``application binary interface'' (ABI)@. When this macro
7173 is defined, RTL is generated for this register once, as with the stack
7174 pointer and frame pointer registers. If this macro is not defined, it
7175 is up to the machine-dependent files to allocate such a register (if
7176 necessary). Note that this register must be fixed when in use (e.g.@:
7177 when @code{flag_pic} is true).
7180 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7181 A C expression that is nonzero if the register defined by
7182 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7183 the default is zero. Do not define
7184 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7187 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7188 A C expression that is nonzero if @var{x} is a legitimate immediate
7189 operand on the target machine when generating position independent code.
7190 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7191 check this. You can also assume @var{flag_pic} is true, so you need not
7192 check it either. You need not define this macro if all constants
7193 (including @code{SYMBOL_REF}) can be immediate operands when generating
7194 position independent code.
7197 @node Assembler Format
7198 @section Defining the Output Assembler Language
7200 This section describes macros whose principal purpose is to describe how
7201 to write instructions in assembler language---rather than what the
7205 * File Framework:: Structural information for the assembler file.
7206 * Data Output:: Output of constants (numbers, strings, addresses).
7207 * Uninitialized Data:: Output of uninitialized variables.
7208 * Label Output:: Output and generation of labels.
7209 * Initialization:: General principles of initialization
7210 and termination routines.
7211 * Macros for Initialization::
7212 Specific macros that control the handling of
7213 initialization and termination routines.
7214 * Instruction Output:: Output of actual instructions.
7215 * Dispatch Tables:: Output of jump tables.
7216 * Exception Region Output:: Output of exception region code.
7217 * Alignment Output:: Pseudo ops for alignment and skipping data.
7220 @node File Framework
7221 @subsection The Overall Framework of an Assembler File
7222 @cindex assembler format
7223 @cindex output of assembler code
7225 @c prevent bad page break with this line
7226 This describes the overall framework of an assembly file.
7228 @findex default_file_start
7229 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7230 Output to @code{asm_out_file} any text which the assembler expects to
7231 find at the beginning of a file. The default behavior is controlled
7232 by two flags, documented below. Unless your target's assembler is
7233 quite unusual, if you override the default, you should call
7234 @code{default_file_start} at some point in your target hook. This
7235 lets other target files rely on these variables.
7238 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7239 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7240 printed as the very first line in the assembly file, unless
7241 @option{-fverbose-asm} is in effect. (If that macro has been defined
7242 to the empty string, this variable has no effect.) With the normal
7243 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7244 assembler that it need not bother stripping comments or extra
7245 whitespace from its input. This allows it to work a bit faster.
7247 The default is false. You should not set it to true unless you have
7248 verified that your port does not generate any extra whitespace or
7249 comments that will cause GAS to issue errors in NO_APP mode.
7252 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7253 If this flag is true, @code{output_file_directive} will be called
7254 for the primary source file, immediately after printing
7255 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7256 this to be done. The default is false.
7259 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7260 Output to @code{asm_out_file} any text which the assembler expects
7261 to find at the end of a file. The default is to output nothing.
7264 @deftypefun void file_end_indicate_exec_stack ()
7265 Some systems use a common convention, the @samp{.note.GNU-stack}
7266 special section, to indicate whether or not an object file relies on
7267 the stack being executable. If your system uses this convention, you
7268 should define @code{TARGET_ASM_FILE_END} to this function. If you
7269 need to do other things in that hook, have your hook function call
7273 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7274 Output to @code{asm_out_file} any text which the assembler expects
7275 to find at the start of an LTO section. The default is to output
7279 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7280 Output to @code{asm_out_file} any text which the assembler expects
7281 to find at the end of an LTO section. The default is to output
7285 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7286 Output to @code{asm_out_file} any text which is needed before emitting
7287 unwind info and debug info at the end of a file. Some targets emit
7288 here PIC setup thunks that cannot be emitted at the end of file,
7289 because they couldn't have unwind info then. The default is to output
7293 @defmac ASM_COMMENT_START
7294 A C string constant describing how to begin a comment in the target
7295 assembler language. The compiler assumes that the comment will end at
7296 the end of the line.
7300 A C string constant for text to be output before each @code{asm}
7301 statement or group of consecutive ones. Normally this is
7302 @code{"#APP"}, which is a comment that has no effect on most
7303 assemblers but tells the GNU assembler that it must check the lines
7304 that follow for all valid assembler constructs.
7308 A C string constant for text to be output after each @code{asm}
7309 statement or group of consecutive ones. Normally this is
7310 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7311 time-saving assumptions that are valid for ordinary compiler output.
7314 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7315 A C statement to output COFF information or DWARF debugging information
7316 which indicates that filename @var{name} is the current source file to
7317 the stdio stream @var{stream}.
7319 This macro need not be defined if the standard form of output
7320 for the file format in use is appropriate.
7323 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7324 Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7326 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7329 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7330 A C statement to output the string @var{string} to the stdio stream
7331 @var{stream}. If you do not call the function @code{output_quoted_string}
7332 in your config files, GCC will only call it to output filenames to
7333 the assembler source. So you can use it to canonicalize the format
7334 of the filename using this macro.
7337 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7338 A C statement to output something to the assembler file to handle a
7339 @samp{#ident} directive containing the text @var{string}. If this
7340 macro is not defined, nothing is output for a @samp{#ident} directive.
7343 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7344 Output assembly directives to switch to section @var{name}. The section
7345 should have attributes as specified by @var{flags}, which is a bit mask
7346 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7347 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7348 this section is associated.
7351 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
7352 Return preferred text (sub)section for function @var{decl}.
7353 Main purpose of this function is to separate cold, normal and hot
7354 functions. @var{startup} is true when function is known to be used only
7355 at startup (from static constructors or it is @code{main()}).
7356 @var{exit} is true when function is known to be used only at exit
7357 (from static destructors).
7358 Return NULL if function should go to default text section.
7361 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
7362 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}.
7365 @deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7366 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7367 It must not be modified by command-line option processing.
7370 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7371 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7372 This flag is true if we can create zeroed data by switching to a BSS
7373 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7374 This is true on most ELF targets.
7377 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7378 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7379 based on a variable or function decl, a section name, and whether or not the
7380 declaration's initializer may contain runtime relocations. @var{decl} may be
7381 null, in which case read-write data should be assumed.
7383 The default version of this function handles choosing code vs data,
7384 read-only vs read-write data, and @code{flag_pic}. You should only
7385 need to override this if your target has special flags that might be
7386 set via @code{__attribute__}.
7389 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
7390 Provides the target with the ability to record the gcc command line
7391 switches that have been passed to the compiler, and options that are
7392 enabled. The @var{type} argument specifies what is being recorded.
7393 It can take the following values:
7396 @item SWITCH_TYPE_PASSED
7397 @var{text} is a command line switch that has been set by the user.
7399 @item SWITCH_TYPE_ENABLED
7400 @var{text} is an option which has been enabled. This might be as a
7401 direct result of a command line switch, or because it is enabled by
7402 default or because it has been enabled as a side effect of a different
7403 command line switch. For example, the @option{-O2} switch enables
7404 various different individual optimization passes.
7406 @item SWITCH_TYPE_DESCRIPTIVE
7407 @var{text} is either NULL or some descriptive text which should be
7408 ignored. If @var{text} is NULL then it is being used to warn the
7409 target hook that either recording is starting or ending. The first
7410 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7411 warning is for start up and the second time the warning is for
7412 wind down. This feature is to allow the target hook to make any
7413 necessary preparations before it starts to record switches and to
7414 perform any necessary tidying up after it has finished recording
7417 @item SWITCH_TYPE_LINE_START
7418 This option can be ignored by this target hook.
7420 @item SWITCH_TYPE_LINE_END
7421 This option can be ignored by this target hook.
7424 The hook's return value must be zero. Other return values may be
7425 supported in the future.
7427 By default this hook is set to NULL, but an example implementation is
7428 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7429 it records the switches as ASCII text inside a new, string mergeable
7430 section in the assembler output file. The name of the new section is
7431 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7435 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7436 This is the name of the section that will be created by the example
7437 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7443 @subsection Output of Data
7446 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7447 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7448 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7449 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7450 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7451 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7452 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7453 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7454 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7455 These hooks specify assembly directives for creating certain kinds
7456 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7457 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7458 aligned two-byte object, and so on. Any of the hooks may be
7459 @code{NULL}, indicating that no suitable directive is available.
7461 The compiler will print these strings at the start of a new line,
7462 followed immediately by the object's initial value. In most cases,
7463 the string should contain a tab, a pseudo-op, and then another tab.
7466 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7467 The @code{assemble_integer} function uses this hook to output an
7468 integer object. @var{x} is the object's value, @var{size} is its size
7469 in bytes and @var{aligned_p} indicates whether it is aligned. The
7470 function should return @code{true} if it was able to output the
7471 object. If it returns false, @code{assemble_integer} will try to
7472 split the object into smaller parts.
7474 The default implementation of this hook will use the
7475 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7476 when the relevant string is @code{NULL}.
7479 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
7480 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7481 can't deal with, and output assembly code to @var{file} corresponding to
7482 the pattern @var{x}. This may be used to allow machine-dependent
7483 @code{UNSPEC}s to appear within constants.
7485 If target hook fails to recognize a pattern, it must return @code{false},
7486 so that a standard error message is printed. If it prints an error message
7487 itself, by calling, for example, @code{output_operand_lossage}, it may just
7491 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
7492 A C statement to recognize @var{rtx} patterns that
7493 @code{output_addr_const} can't deal with, and output assembly code to
7494 @var{stream} corresponding to the pattern @var{x}. This may be used to
7495 allow machine-dependent @code{UNSPEC}s to appear within constants.
7497 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
7498 @code{goto fail}, so that a standard error message is printed. If it
7499 prints an error message itself, by calling, for example,
7500 @code{output_operand_lossage}, it may just complete normally.
7503 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7504 A C statement to output to the stdio stream @var{stream} an assembler
7505 instruction to assemble a string constant containing the @var{len}
7506 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7507 @code{char *} and @var{len} a C expression of type @code{int}.
7509 If the assembler has a @code{.ascii} pseudo-op as found in the
7510 Berkeley Unix assembler, do not define the macro
7511 @code{ASM_OUTPUT_ASCII}.
7514 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7515 A C statement to output word @var{n} of a function descriptor for
7516 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7517 is defined, and is otherwise unused.
7520 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7521 You may define this macro as a C expression. You should define the
7522 expression to have a nonzero value if GCC should output the constant
7523 pool for a function before the code for the function, or a zero value if
7524 GCC should output the constant pool after the function. If you do
7525 not define this macro, the usual case, GCC will output the constant
7526 pool before the function.
7529 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7530 A C statement to output assembler commands to define the start of the
7531 constant pool for a function. @var{funname} is a string giving
7532 the name of the function. Should the return type of the function
7533 be required, it can be obtained via @var{fundecl}. @var{size}
7534 is the size, in bytes, of the constant pool that will be written
7535 immediately after this call.
7537 If no constant-pool prefix is required, the usual case, this macro need
7541 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7542 A C statement (with or without semicolon) to output a constant in the
7543 constant pool, if it needs special treatment. (This macro need not do
7544 anything for RTL expressions that can be output normally.)
7546 The argument @var{file} is the standard I/O stream to output the
7547 assembler code on. @var{x} is the RTL expression for the constant to
7548 output, and @var{mode} is the machine mode (in case @var{x} is a
7549 @samp{const_int}). @var{align} is the required alignment for the value
7550 @var{x}; you should output an assembler directive to force this much
7553 The argument @var{labelno} is a number to use in an internal label for
7554 the address of this pool entry. The definition of this macro is
7555 responsible for outputting the label definition at the proper place.
7556 Here is how to do this:
7559 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7562 When you output a pool entry specially, you should end with a
7563 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7564 entry from being output a second time in the usual manner.
7566 You need not define this macro if it would do nothing.
7569 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7570 A C statement to output assembler commands to at the end of the constant
7571 pool for a function. @var{funname} is a string giving the name of the
7572 function. Should the return type of the function be required, you can
7573 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7574 constant pool that GCC wrote immediately before this call.
7576 If no constant-pool epilogue is required, the usual case, you need not
7580 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7581 Define this macro as a C expression which is nonzero if @var{C} is
7582 used as a logical line separator by the assembler. @var{STR} points
7583 to the position in the string where @var{C} was found; this can be used if
7584 a line separator uses multiple characters.
7586 If you do not define this macro, the default is that only
7587 the character @samp{;} is treated as a logical line separator.
7590 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7591 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7592 These target hooks are C string constants, describing the syntax in the
7593 assembler for grouping arithmetic expressions. If not overridden, they
7594 default to normal parentheses, which is correct for most assemblers.
7597 These macros are provided by @file{real.h} for writing the definitions
7598 of @code{ASM_OUTPUT_DOUBLE} and the like:
7600 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7601 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7602 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7603 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7604 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7605 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7606 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7607 target's floating point representation, and store its bit pattern in
7608 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7609 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7610 simple @code{long int}. For the others, it should be an array of
7611 @code{long int}. The number of elements in this array is determined
7612 by the size of the desired target floating point data type: 32 bits of
7613 it go in each @code{long int} array element. Each array element holds
7614 32 bits of the result, even if @code{long int} is wider than 32 bits
7615 on the host machine.
7617 The array element values are designed so that you can print them out
7618 using @code{fprintf} in the order they should appear in the target
7622 @node Uninitialized Data
7623 @subsection Output of Uninitialized Variables
7625 Each of the macros in this section is used to do the whole job of
7626 outputting a single uninitialized variable.
7628 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7629 A C statement (sans semicolon) to output to the stdio stream
7630 @var{stream} the assembler definition of a common-label named
7631 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7632 is the size rounded up to whatever alignment the caller wants. It is
7633 possible that @var{size} may be zero, for instance if a struct with no
7634 other member than a zero-length array is defined. In this case, the
7635 backend must output a symbol definition that allocates at least one
7636 byte, both so that the address of the resulting object does not compare
7637 equal to any other, and because some object formats cannot even express
7638 the concept of a zero-sized common symbol, as that is how they represent
7639 an ordinary undefined external.
7641 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7642 output the name itself; before and after that, output the additional
7643 assembler syntax for defining the name, and a newline.
7645 This macro controls how the assembler definitions of uninitialized
7646 common global variables are output.
7649 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7650 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7651 separate, explicit argument. If you define this macro, it is used in
7652 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7653 handling the required alignment of the variable. The alignment is specified
7654 as the number of bits.
7657 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7658 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7659 variable to be output, if there is one, or @code{NULL_TREE} if there
7660 is no corresponding variable. If you define this macro, GCC will use it
7661 in place of both @code{ASM_OUTPUT_COMMON} and
7662 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7663 the variable's decl in order to chose what to output.
7666 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7667 A C statement (sans semicolon) to output to the stdio stream
7668 @var{stream} the assembler definition of uninitialized global @var{decl} named
7669 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
7670 is the alignment specified as the number of bits.
7672 Try to use function @code{asm_output_aligned_bss} defined in file
7673 @file{varasm.c} when defining this macro. If unable, use the expression
7674 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7675 before and after that, output the additional assembler syntax for defining
7676 the name, and a newline.
7678 There are two ways of handling global BSS@. One is to define this macro.
7679 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7680 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7681 You do not need to do both.
7683 Some languages do not have @code{common} data, and require a
7684 non-common form of global BSS in order to handle uninitialized globals
7685 efficiently. C++ is one example of this. However, if the target does
7686 not support global BSS, the front end may choose to make globals
7687 common in order to save space in the object file.
7690 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7691 A C statement (sans semicolon) to output to the stdio stream
7692 @var{stream} the assembler definition of a local-common-label named
7693 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7694 is the size rounded up to whatever alignment the caller wants.
7696 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7697 output the name itself; before and after that, output the additional
7698 assembler syntax for defining the name, and a newline.
7700 This macro controls how the assembler definitions of uninitialized
7701 static variables are output.
7704 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7705 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7706 separate, explicit argument. If you define this macro, it is used in
7707 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7708 handling the required alignment of the variable. The alignment is specified
7709 as the number of bits.
7712 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7713 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7714 variable to be output, if there is one, or @code{NULL_TREE} if there
7715 is no corresponding variable. If you define this macro, GCC will use it
7716 in place of both @code{ASM_OUTPUT_DECL} and
7717 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7718 the variable's decl in order to chose what to output.
7722 @subsection Output and Generation of Labels
7724 @c prevent bad page break with this line
7725 This is about outputting labels.
7727 @findex assemble_name
7728 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7729 A C statement (sans semicolon) to output to the stdio stream
7730 @var{stream} the assembler definition of a label named @var{name}.
7731 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7732 output the name itself; before and after that, output the additional
7733 assembler syntax for defining the name, and a newline. A default
7734 definition of this macro is provided which is correct for most systems.
7737 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7738 A C statement (sans semicolon) to output to the stdio stream
7739 @var{stream} the assembler definition of a label named @var{name} of
7741 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7742 output the name itself; before and after that, output the additional
7743 assembler syntax for defining the name, and a newline. A default
7744 definition of this macro is provided which is correct for most systems.
7746 If this macro is not defined, then the function name is defined in the
7747 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7750 @findex assemble_name_raw
7751 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7752 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7753 to refer to a compiler-generated label. The default definition uses
7754 @code{assemble_name_raw}, which is like @code{assemble_name} except
7755 that it is more efficient.
7759 A C string containing the appropriate assembler directive to specify the
7760 size of a symbol, without any arguments. On systems that use ELF, the
7761 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7762 systems, the default is not to define this macro.
7764 Define this macro only if it is correct to use the default definitions
7765 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7766 for your system. If you need your own custom definitions of those
7767 macros, or if you do not need explicit symbol sizes at all, do not
7771 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7772 A C statement (sans semicolon) to output to the stdio stream
7773 @var{stream} a directive telling the assembler that the size of the
7774 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7775 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7779 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7780 A C statement (sans semicolon) to output to the stdio stream
7781 @var{stream} a directive telling the assembler to calculate the size of
7782 the symbol @var{name} by subtracting its address from the current
7785 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7786 provided. The default assumes that the assembler recognizes a special
7787 @samp{.} symbol as referring to the current address, and can calculate
7788 the difference between this and another symbol. If your assembler does
7789 not recognize @samp{.} or cannot do calculations with it, you will need
7790 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7794 A C string containing the appropriate assembler directive to specify the
7795 type of a symbol, without any arguments. On systems that use ELF, the
7796 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7797 systems, the default is not to define this macro.
7799 Define this macro only if it is correct to use the default definition of
7800 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7801 custom definition of this macro, or if you do not need explicit symbol
7802 types at all, do not define this macro.
7805 @defmac TYPE_OPERAND_FMT
7806 A C string which specifies (using @code{printf} syntax) the format of
7807 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7808 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7809 the default is not to define this macro.
7811 Define this macro only if it is correct to use the default definition of
7812 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7813 custom definition of this macro, or if you do not need explicit symbol
7814 types at all, do not define this macro.
7817 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7818 A C statement (sans semicolon) to output to the stdio stream
7819 @var{stream} a directive telling the assembler that the type of the
7820 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7821 that string is always either @samp{"function"} or @samp{"object"}, but
7822 you should not count on this.
7824 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7825 definition of this macro is provided.
7828 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7829 A C statement (sans semicolon) to output to the stdio stream
7830 @var{stream} any text necessary for declaring the name @var{name} of a
7831 function which is being defined. This macro is responsible for
7832 outputting the label definition (perhaps using
7833 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7834 @code{FUNCTION_DECL} tree node representing the function.
7836 If this macro is not defined, then the function name is defined in the
7837 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7839 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7843 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7844 A C statement (sans semicolon) to output to the stdio stream
7845 @var{stream} any text necessary for declaring the size of a function
7846 which is being defined. The argument @var{name} is the name of the
7847 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7848 representing the function.
7850 If this macro is not defined, then the function size is not defined.
7852 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7856 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7857 A C statement (sans semicolon) to output to the stdio stream
7858 @var{stream} any text necessary for declaring the name @var{name} of an
7859 initialized variable which is being defined. This macro must output the
7860 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7861 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7863 If this macro is not defined, then the variable name is defined in the
7864 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7866 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7867 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7870 @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})
7871 A target hook to output to the stdio stream @var{file} any text necessary
7872 for declaring the name @var{name} of a constant which is being defined. This
7873 target hook is responsible for outputting the label definition (perhaps using
7874 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7875 and @var{size} is the size of the constant in bytes. The @var{name}
7876 will be an internal label.
7878 The default version of this target hook, define the @var{name} in the
7879 usual manner as a label (by means of @code{assemble_label}).
7881 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7884 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7885 A C statement (sans semicolon) to output to the stdio stream
7886 @var{stream} any text necessary for claiming a register @var{regno}
7887 for a global variable @var{decl} with name @var{name}.
7889 If you don't define this macro, that is equivalent to defining it to do
7893 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7894 A C statement (sans semicolon) to finish up declaring a variable name
7895 once the compiler has processed its initializer fully and thus has had a
7896 chance to determine the size of an array when controlled by an
7897 initializer. This is used on systems where it's necessary to declare
7898 something about the size of the object.
7900 If you don't define this macro, that is equivalent to defining it to do
7903 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7904 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7907 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7908 This target hook is a function to output to the stdio stream
7909 @var{stream} some commands that will make the label @var{name} global;
7910 that is, available for reference from other files.
7912 The default implementation relies on a proper definition of
7913 @code{GLOBAL_ASM_OP}.
7916 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7917 This target hook is a function to output to the stdio stream
7918 @var{stream} some commands that will make the name associated with @var{decl}
7919 global; that is, available for reference from other files.
7921 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7924 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7925 A C statement (sans semicolon) to output to the stdio stream
7926 @var{stream} some commands that will make the label @var{name} weak;
7927 that is, available for reference from other files but only used if
7928 no other definition is available. Use the expression
7929 @code{assemble_name (@var{stream}, @var{name})} to output the name
7930 itself; before and after that, output the additional assembler syntax
7931 for making that name weak, and a newline.
7933 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7934 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7938 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7939 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7940 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7941 or variable decl. If @var{value} is not @code{NULL}, this C statement
7942 should output to the stdio stream @var{stream} assembler code which
7943 defines (equates) the weak symbol @var{name} to have the value
7944 @var{value}. If @var{value} is @code{NULL}, it should output commands
7945 to make @var{name} weak.
7948 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7949 Outputs a directive that enables @var{name} to be used to refer to
7950 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7951 declaration of @code{name}.
7954 @defmac SUPPORTS_WEAK
7955 A preprocessor constant expression which evaluates to true if the target
7956 supports weak symbols.
7958 If you don't define this macro, @file{defaults.h} provides a default
7959 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7960 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
7963 @defmac TARGET_SUPPORTS_WEAK
7964 A C expression which evaluates to true if the target supports weak symbols.
7966 If you don't define this macro, @file{defaults.h} provides a default
7967 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
7968 this macro if you want to control weak symbol support with a compiler
7969 flag such as @option{-melf}.
7972 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7973 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7974 public symbol such that extra copies in multiple translation units will
7975 be discarded by the linker. Define this macro if your object file
7976 format provides support for this concept, such as the @samp{COMDAT}
7977 section flags in the Microsoft Windows PE/COFF format, and this support
7978 requires changes to @var{decl}, such as putting it in a separate section.
7981 @defmac SUPPORTS_ONE_ONLY
7982 A C expression which evaluates to true if the target supports one-only
7985 If you don't define this macro, @file{varasm.c} provides a default
7986 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7987 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7988 you want to control one-only symbol support with a compiler flag, or if
7989 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7990 be emitted as one-only.
7993 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
7994 This target hook is a function to output to @var{asm_out_file} some
7995 commands that will make the symbol(s) associated with @var{decl} have
7996 hidden, protected or internal visibility as specified by @var{visibility}.
7999 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
8000 A C expression that evaluates to true if the target's linker expects
8001 that weak symbols do not appear in a static archive's table of contents.
8002 The default is @code{0}.
8004 Leaving weak symbols out of an archive's table of contents means that,
8005 if a symbol will only have a definition in one translation unit and
8006 will have undefined references from other translation units, that
8007 symbol should not be weak. Defining this macro to be nonzero will
8008 thus have the effect that certain symbols that would normally be weak
8009 (explicit template instantiations, and vtables for polymorphic classes
8010 with noninline key methods) will instead be nonweak.
8012 The C++ ABI requires this macro to be zero. Define this macro for
8013 targets where full C++ ABI compliance is impossible and where linker
8014 restrictions require weak symbols to be left out of a static archive's
8018 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8019 A C statement (sans semicolon) to output to the stdio stream
8020 @var{stream} any text necessary for declaring the name of an external
8021 symbol named @var{name} which is referenced in this compilation but
8022 not defined. The value of @var{decl} is the tree node for the
8025 This macro need not be defined if it does not need to output anything.
8026 The GNU assembler and most Unix assemblers don't require anything.
8029 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8030 This target hook is a function to output to @var{asm_out_file} an assembler
8031 pseudo-op to declare a library function name external. The name of the
8032 library function is given by @var{symref}, which is a @code{symbol_ref}.
8035 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8036 This target hook is a function to output to @var{asm_out_file} an assembler
8037 directive to annotate @var{symbol} as used. The Darwin target uses the
8038 .no_dead_code_strip directive.
8041 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8042 A C statement (sans semicolon) to output to the stdio stream
8043 @var{stream} a reference in assembler syntax to a label named
8044 @var{name}. This should add @samp{_} to the front of the name, if that
8045 is customary on your operating system, as it is in most Berkeley Unix
8046 systems. This macro is used in @code{assemble_name}.
8049 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8050 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.
8053 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8054 A C statement (sans semicolon) to output a reference to
8055 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8056 will be used to output the name of the symbol. This macro may be used
8057 to modify the way a symbol is referenced depending on information
8058 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8061 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8062 A C statement (sans semicolon) to output a reference to @var{buf}, the
8063 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8064 @code{assemble_name} will be used to output the name of the symbol.
8065 This macro is not used by @code{output_asm_label}, or the @code{%l}
8066 specifier that calls it; the intention is that this macro should be set
8067 when it is necessary to output a label differently when its address is
8071 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8072 A function to output to the stdio stream @var{stream} a label whose
8073 name is made from the string @var{prefix} and the number @var{labelno}.
8075 It is absolutely essential that these labels be distinct from the labels
8076 used for user-level functions and variables. Otherwise, certain programs
8077 will have name conflicts with internal labels.
8079 It is desirable to exclude internal labels from the symbol table of the
8080 object file. Most assemblers have a naming convention for labels that
8081 should be excluded; on many systems, the letter @samp{L} at the
8082 beginning of a label has this effect. You should find out what
8083 convention your system uses, and follow it.
8085 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8088 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8089 A C statement to output to the stdio stream @var{stream} a debug info
8090 label whose name is made from the string @var{prefix} and the number
8091 @var{num}. This is useful for VLIW targets, where debug info labels
8092 may need to be treated differently than branch target labels. On some
8093 systems, branch target labels must be at the beginning of instruction
8094 bundles, but debug info labels can occur in the middle of instruction
8097 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8101 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8102 A C statement to store into the string @var{string} a label whose name
8103 is made from the string @var{prefix} and the number @var{num}.
8105 This string, when output subsequently by @code{assemble_name}, should
8106 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8107 with the same @var{prefix} and @var{num}.
8109 If the string begins with @samp{*}, then @code{assemble_name} will
8110 output the rest of the string unchanged. It is often convenient for
8111 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8112 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8113 to output the string, and may change it. (Of course,
8114 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8115 you should know what it does on your machine.)
8118 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8119 A C expression to assign to @var{outvar} (which is a variable of type
8120 @code{char *}) a newly allocated string made from the string
8121 @var{name} and the number @var{number}, with some suitable punctuation
8122 added. Use @code{alloca} to get space for the string.
8124 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8125 produce an assembler label for an internal static variable whose name is
8126 @var{name}. Therefore, the string must be such as to result in valid
8127 assembler code. The argument @var{number} is different each time this
8128 macro is executed; it prevents conflicts between similarly-named
8129 internal static variables in different scopes.
8131 Ideally this string should not be a valid C identifier, to prevent any
8132 conflict with the user's own symbols. Most assemblers allow periods
8133 or percent signs in assembler symbols; putting at least one of these
8134 between the name and the number will suffice.
8136 If this macro is not defined, a default definition will be provided
8137 which is correct for most systems.
8140 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8141 A C statement to output to the stdio stream @var{stream} assembler code
8142 which defines (equates) the symbol @var{name} to have the value @var{value}.
8145 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8146 correct for most systems.
8149 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8150 A C statement to output to the stdio stream @var{stream} assembler code
8151 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8152 to have the value of the tree node @var{decl_of_value}. This macro will
8153 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8154 the tree nodes are available.
8157 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8158 correct for most systems.
8161 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8162 A C statement that evaluates to true if the assembler code which defines
8163 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8164 of the tree node @var{decl_of_value} should be emitted near the end of the
8165 current compilation unit. The default is to not defer output of defines.
8166 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8167 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8170 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8171 A C statement to output to the stdio stream @var{stream} assembler code
8172 which defines (equates) the weak symbol @var{name} to have the value
8173 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8174 an undefined weak symbol.
8176 Define this macro if the target only supports weak aliases; define
8177 @code{ASM_OUTPUT_DEF} instead if possible.
8180 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8181 Define this macro to override the default assembler names used for
8182 Objective-C methods.
8184 The default name is a unique method number followed by the name of the
8185 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8186 the category is also included in the assembler name (e.g.@:
8189 These names are safe on most systems, but make debugging difficult since
8190 the method's selector is not present in the name. Therefore, particular
8191 systems define other ways of computing names.
8193 @var{buf} is an expression of type @code{char *} which gives you a
8194 buffer in which to store the name; its length is as long as
8195 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8196 50 characters extra.
8198 The argument @var{is_inst} specifies whether the method is an instance
8199 method or a class method; @var{class_name} is the name of the class;
8200 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8201 in a category); and @var{sel_name} is the name of the selector.
8203 On systems where the assembler can handle quoted names, you can use this
8204 macro to provide more human-readable names.
8207 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
8208 A C statement (sans semicolon) to output to the stdio stream
8209 @var{stream} commands to declare that the label @var{name} is an
8210 Objective-C class reference. This is only needed for targets whose
8211 linkers have special support for NeXT-style runtimes.
8214 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
8215 A C statement (sans semicolon) to output to the stdio stream
8216 @var{stream} commands to declare that the label @var{name} is an
8217 unresolved Objective-C class reference. This is only needed for targets
8218 whose linkers have special support for NeXT-style runtimes.
8221 @node Initialization
8222 @subsection How Initialization Functions Are Handled
8223 @cindex initialization routines
8224 @cindex termination routines
8225 @cindex constructors, output of
8226 @cindex destructors, output of
8228 The compiled code for certain languages includes @dfn{constructors}
8229 (also called @dfn{initialization routines})---functions to initialize
8230 data in the program when the program is started. These functions need
8231 to be called before the program is ``started''---that is to say, before
8232 @code{main} is called.
8234 Compiling some languages generates @dfn{destructors} (also called
8235 @dfn{termination routines}) that should be called when the program
8238 To make the initialization and termination functions work, the compiler
8239 must output something in the assembler code to cause those functions to
8240 be called at the appropriate time. When you port the compiler to a new
8241 system, you need to specify how to do this.
8243 There are two major ways that GCC currently supports the execution of
8244 initialization and termination functions. Each way has two variants.
8245 Much of the structure is common to all four variations.
8247 @findex __CTOR_LIST__
8248 @findex __DTOR_LIST__
8249 The linker must build two lists of these functions---a list of
8250 initialization functions, called @code{__CTOR_LIST__}, and a list of
8251 termination functions, called @code{__DTOR_LIST__}.
8253 Each list always begins with an ignored function pointer (which may hold
8254 0, @minus{}1, or a count of the function pointers after it, depending on
8255 the environment). This is followed by a series of zero or more function
8256 pointers to constructors (or destructors), followed by a function
8257 pointer containing zero.
8259 Depending on the operating system and its executable file format, either
8260 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8261 time and exit time. Constructors are called in reverse order of the
8262 list; destructors in forward order.
8264 The best way to handle static constructors works only for object file
8265 formats which provide arbitrarily-named sections. A section is set
8266 aside for a list of constructors, and another for a list of destructors.
8267 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8268 object file that defines an initialization function also puts a word in
8269 the constructor section to point to that function. The linker
8270 accumulates all these words into one contiguous @samp{.ctors} section.
8271 Termination functions are handled similarly.
8273 This method will be chosen as the default by @file{target-def.h} if
8274 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8275 support arbitrary sections, but does support special designated
8276 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8277 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8279 When arbitrary sections are available, there are two variants, depending
8280 upon how the code in @file{crtstuff.c} is called. On systems that
8281 support a @dfn{.init} section which is executed at program startup,
8282 parts of @file{crtstuff.c} are compiled into that section. The
8283 program is linked by the @command{gcc} driver like this:
8286 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8289 The prologue of a function (@code{__init}) appears in the @code{.init}
8290 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8291 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8292 files are provided by the operating system or by the GNU C library, but
8293 are provided by GCC for a few targets.
8295 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8296 compiled from @file{crtstuff.c}. They contain, among other things, code
8297 fragments within the @code{.init} and @code{.fini} sections that branch
8298 to routines in the @code{.text} section. The linker will pull all parts
8299 of a section together, which results in a complete @code{__init} function
8300 that invokes the routines we need at startup.
8302 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8305 If no init section is available, when GCC compiles any function called
8306 @code{main} (or more accurately, any function designated as a program
8307 entry point by the language front end calling @code{expand_main_function}),
8308 it inserts a procedure call to @code{__main} as the first executable code
8309 after the function prologue. The @code{__main} function is defined
8310 in @file{libgcc2.c} and runs the global constructors.
8312 In file formats that don't support arbitrary sections, there are again
8313 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8314 and an `a.out' format must be used. In this case,
8315 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8316 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8317 and with the address of the void function containing the initialization
8318 code as its value. The GNU linker recognizes this as a request to add
8319 the value to a @dfn{set}; the values are accumulated, and are eventually
8320 placed in the executable as a vector in the format described above, with
8321 a leading (ignored) count and a trailing zero element.
8322 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8323 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8324 the compilation of @code{main} to call @code{__main} as above, starting
8325 the initialization process.
8327 The last variant uses neither arbitrary sections nor the GNU linker.
8328 This is preferable when you want to do dynamic linking and when using
8329 file formats which the GNU linker does not support, such as `ECOFF'@. In
8330 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8331 termination functions are recognized simply by their names. This requires
8332 an extra program in the linkage step, called @command{collect2}. This program
8333 pretends to be the linker, for use with GCC; it does its job by running
8334 the ordinary linker, but also arranges to include the vectors of
8335 initialization and termination functions. These functions are called
8336 via @code{__main} as described above. In order to use this method,
8337 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8340 The following section describes the specific macros that control and
8341 customize the handling of initialization and termination functions.
8344 @node Macros for Initialization
8345 @subsection Macros Controlling Initialization Routines
8347 Here are the macros that control how the compiler handles initialization
8348 and termination functions:
8350 @defmac INIT_SECTION_ASM_OP
8351 If defined, a C string constant, including spacing, for the assembler
8352 operation to identify the following data as initialization code. If not
8353 defined, GCC will assume such a section does not exist. When you are
8354 using special sections for initialization and termination functions, this
8355 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8356 run the initialization functions.
8359 @defmac HAS_INIT_SECTION
8360 If defined, @code{main} will not call @code{__main} as described above.
8361 This macro should be defined for systems that control start-up code
8362 on a symbol-by-symbol basis, such as OSF/1, and should not
8363 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8366 @defmac LD_INIT_SWITCH
8367 If defined, a C string constant for a switch that tells the linker that
8368 the following symbol is an initialization routine.
8371 @defmac LD_FINI_SWITCH
8372 If defined, a C string constant for a switch that tells the linker that
8373 the following symbol is a finalization routine.
8376 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8377 If defined, a C statement that will write a function that can be
8378 automatically called when a shared library is loaded. The function
8379 should call @var{func}, which takes no arguments. If not defined, and
8380 the object format requires an explicit initialization function, then a
8381 function called @code{_GLOBAL__DI} will be generated.
8383 This function and the following one are used by collect2 when linking a
8384 shared library that needs constructors or destructors, or has DWARF2
8385 exception tables embedded in the code.
8388 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8389 If defined, a C statement that will write a function that can be
8390 automatically called when a shared library is unloaded. The function
8391 should call @var{func}, which takes no arguments. If not defined, and
8392 the object format requires an explicit finalization function, then a
8393 function called @code{_GLOBAL__DD} will be generated.
8396 @defmac INVOKE__main
8397 If defined, @code{main} will call @code{__main} despite the presence of
8398 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8399 where the init section is not actually run automatically, but is still
8400 useful for collecting the lists of constructors and destructors.
8403 @defmac SUPPORTS_INIT_PRIORITY
8404 If nonzero, the C++ @code{init_priority} attribute is supported and the
8405 compiler should emit instructions to control the order of initialization
8406 of objects. If zero, the compiler will issue an error message upon
8407 encountering an @code{init_priority} attribute.
8410 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8411 This value is true if the target supports some ``native'' method of
8412 collecting constructors and destructors to be run at startup and exit.
8413 It is false if we must use @command{collect2}.
8416 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8417 If defined, a function that outputs assembler code to arrange to call
8418 the function referenced by @var{symbol} at initialization time.
8420 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8421 no arguments and with no return value. If the target supports initialization
8422 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8423 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8425 If this macro is not defined by the target, a suitable default will
8426 be chosen if (1) the target supports arbitrary section names, (2) the
8427 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8431 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8432 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8433 functions rather than initialization functions.
8436 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8437 generated for the generated object file will have static linkage.
8439 If your system uses @command{collect2} as the means of processing
8440 constructors, then that program normally uses @command{nm} to scan
8441 an object file for constructor functions to be called.
8443 On certain kinds of systems, you can define this macro to make
8444 @command{collect2} work faster (and, in some cases, make it work at all):
8446 @defmac OBJECT_FORMAT_COFF
8447 Define this macro if the system uses COFF (Common Object File Format)
8448 object files, so that @command{collect2} can assume this format and scan
8449 object files directly for dynamic constructor/destructor functions.
8451 This macro is effective only in a native compiler; @command{collect2} as
8452 part of a cross compiler always uses @command{nm} for the target machine.
8455 @defmac REAL_NM_FILE_NAME
8456 Define this macro as a C string constant containing the file name to use
8457 to execute @command{nm}. The default is to search the path normally for
8462 @command{collect2} calls @command{nm} to scan object files for static
8463 constructors and destructors and LTO info. By default, @option{-n} is
8464 passed. Define @code{NM_FLAGS} to a C string constant if other options
8465 are needed to get the same output format as GNU @command{nm -n}
8469 If your system supports shared libraries and has a program to list the
8470 dynamic dependencies of a given library or executable, you can define
8471 these macros to enable support for running initialization and
8472 termination functions in shared libraries:
8475 Define this macro to a C string constant containing the name of the program
8476 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8479 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8480 Define this macro to be C code that extracts filenames from the output
8481 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8482 of type @code{char *} that points to the beginning of a line of output
8483 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8484 code must advance @var{ptr} to the beginning of the filename on that
8485 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8488 @defmac SHLIB_SUFFIX
8489 Define this macro to a C string constant containing the default shared
8490 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8491 strips version information after this suffix when generating global
8492 constructor and destructor names. This define is only needed on targets
8493 that use @command{collect2} to process constructors and destructors.
8496 @node Instruction Output
8497 @subsection Output of Assembler Instructions
8499 @c prevent bad page break with this line
8500 This describes assembler instruction output.
8502 @defmac REGISTER_NAMES
8503 A C initializer containing the assembler's names for the machine
8504 registers, each one as a C string constant. This is what translates
8505 register numbers in the compiler into assembler language.
8508 @defmac ADDITIONAL_REGISTER_NAMES
8509 If defined, a C initializer for an array of structures containing a name
8510 and a register number. This macro defines additional names for hard
8511 registers, thus allowing the @code{asm} option in declarations to refer
8512 to registers using alternate names.
8515 @defmac OVERLAPPING_REGISTER_NAMES
8516 If defined, a C initializer for an array of structures containing a
8517 name, a register number and a count of the number of consecutive
8518 machine registers the name overlaps. This macro defines additional
8519 names for hard registers, thus allowing the @code{asm} option in
8520 declarations to refer to registers using alternate names. Unlike
8521 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8522 register name implies multiple underlying registers.
8524 This macro should be used when it is important that a clobber in an
8525 @code{asm} statement clobbers all the underlying values implied by the
8526 register name. For example, on ARM, clobbering the double-precision
8527 VFP register ``d0'' implies clobbering both single-precision registers
8531 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8532 Define this macro if you are using an unusual assembler that
8533 requires different names for the machine instructions.
8535 The definition is a C statement or statements which output an
8536 assembler instruction opcode to the stdio stream @var{stream}. The
8537 macro-operand @var{ptr} is a variable of type @code{char *} which
8538 points to the opcode name in its ``internal'' form---the form that is
8539 written in the machine description. The definition should output the
8540 opcode name to @var{stream}, performing any translation you desire, and
8541 increment the variable @var{ptr} to point at the end of the opcode
8542 so that it will not be output twice.
8544 In fact, your macro definition may process less than the entire opcode
8545 name, or more than the opcode name; but if you want to process text
8546 that includes @samp{%}-sequences to substitute operands, you must take
8547 care of the substitution yourself. Just be sure to increment
8548 @var{ptr} over whatever text should not be output normally.
8550 @findex recog_data.operand
8551 If you need to look at the operand values, they can be found as the
8552 elements of @code{recog_data.operand}.
8554 If the macro definition does nothing, the instruction is output
8558 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8559 If defined, a C statement to be executed just prior to the output of
8560 assembler code for @var{insn}, to modify the extracted operands so
8561 they will be output differently.
8563 Here the argument @var{opvec} is the vector containing the operands
8564 extracted from @var{insn}, and @var{noperands} is the number of
8565 elements of the vector which contain meaningful data for this insn.
8566 The contents of this vector are what will be used to convert the insn
8567 template into assembler code, so you can change the assembler output
8568 by changing the contents of the vector.
8570 This macro is useful when various assembler syntaxes share a single
8571 file of instruction patterns; by defining this macro differently, you
8572 can cause a large class of instructions to be output differently (such
8573 as with rearranged operands). Naturally, variations in assembler
8574 syntax affecting individual insn patterns ought to be handled by
8575 writing conditional output routines in those patterns.
8577 If this macro is not defined, it is equivalent to a null statement.
8580 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx @var{insn}, rtx *@var{opvec}, int @var{noperands})
8581 If defined, this target hook is a function which is executed just after the
8582 output of assembler code for @var{insn}, to change the mode of the assembler
8585 Here the argument @var{opvec} is the vector containing the operands
8586 extracted from @var{insn}, and @var{noperands} is the number of
8587 elements of the vector which contain meaningful data for this insn.
8588 The contents of this vector are what was used to convert the insn
8589 template into assembler code, so you can change the assembler mode
8590 by checking the contents of the vector.
8593 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8594 A C compound statement to output to stdio stream @var{stream} the
8595 assembler syntax for an instruction operand @var{x}. @var{x} is an
8598 @var{code} is a value that can be used to specify one of several ways
8599 of printing the operand. It is used when identical operands must be
8600 printed differently depending on the context. @var{code} comes from
8601 the @samp{%} specification that was used to request printing of the
8602 operand. If the specification was just @samp{%@var{digit}} then
8603 @var{code} is 0; if the specification was @samp{%@var{ltr}
8604 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8607 If @var{x} is a register, this macro should print the register's name.
8608 The names can be found in an array @code{reg_names} whose type is
8609 @code{char *[]}. @code{reg_names} is initialized from
8610 @code{REGISTER_NAMES}.
8612 When the machine description has a specification @samp{%@var{punct}}
8613 (a @samp{%} followed by a punctuation character), this macro is called
8614 with a null pointer for @var{x} and the punctuation character for
8618 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8619 A C expression which evaluates to true if @var{code} is a valid
8620 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8621 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8622 punctuation characters (except for the standard one, @samp{%}) are used
8626 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8627 A C compound statement to output to stdio stream @var{stream} the
8628 assembler syntax for an instruction operand that is a memory reference
8629 whose address is @var{x}. @var{x} is an RTL expression.
8631 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8632 On some machines, the syntax for a symbolic address depends on the
8633 section that the address refers to. On these machines, define the hook
8634 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8635 @code{symbol_ref}, and then check for it here. @xref{Assembler
8639 @findex dbr_sequence_length
8640 @defmac DBR_OUTPUT_SEQEND (@var{file})
8641 A C statement, to be executed after all slot-filler instructions have
8642 been output. If necessary, call @code{dbr_sequence_length} to
8643 determine the number of slots filled in a sequence (zero if not
8644 currently outputting a sequence), to decide how many no-ops to output,
8647 Don't define this macro if it has nothing to do, but it is helpful in
8648 reading assembly output if the extent of the delay sequence is made
8649 explicit (e.g.@: with white space).
8652 @findex final_sequence
8653 Note that output routines for instructions with delay slots must be
8654 prepared to deal with not being output as part of a sequence
8655 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8656 found.) The variable @code{final_sequence} is null when not
8657 processing a sequence, otherwise it contains the @code{sequence} rtx
8661 @defmac REGISTER_PREFIX
8662 @defmacx LOCAL_LABEL_PREFIX
8663 @defmacx USER_LABEL_PREFIX
8664 @defmacx IMMEDIATE_PREFIX
8665 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8666 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8667 @file{final.c}). These are useful when a single @file{md} file must
8668 support multiple assembler formats. In that case, the various @file{tm.h}
8669 files can define these macros differently.
8672 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8673 If defined this macro should expand to a series of @code{case}
8674 statements which will be parsed inside the @code{switch} statement of
8675 the @code{asm_fprintf} function. This allows targets to define extra
8676 printf formats which may useful when generating their assembler
8677 statements. Note that uppercase letters are reserved for future
8678 generic extensions to asm_fprintf, and so are not available to target
8679 specific code. The output file is given by the parameter @var{file}.
8680 The varargs input pointer is @var{argptr} and the rest of the format
8681 string, starting the character after the one that is being switched
8682 upon, is pointed to by @var{format}.
8685 @defmac ASSEMBLER_DIALECT
8686 If your target supports multiple dialects of assembler language (such as
8687 different opcodes), define this macro as a C expression that gives the
8688 numeric index of the assembler language dialect to use, with zero as the
8691 If this macro is defined, you may use constructs of the form
8693 @samp{@{option0|option1|option2@dots{}@}}
8696 in the output templates of patterns (@pxref{Output Template}) or in the
8697 first argument of @code{asm_fprintf}. This construct outputs
8698 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8699 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8700 within these strings retain their usual meaning. If there are fewer
8701 alternatives within the braces than the value of
8702 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8704 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8705 @samp{@}} do not have any special meaning when used in templates or
8706 operands to @code{asm_fprintf}.
8708 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8709 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8710 the variations in assembler language syntax with that mechanism. Define
8711 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8712 if the syntax variant are larger and involve such things as different
8713 opcodes or operand order.
8716 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8717 A C expression to output to @var{stream} some assembler code
8718 which will push hard register number @var{regno} onto the stack.
8719 The code need not be optimal, since this macro is used only when
8723 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8724 A C expression to output to @var{stream} some assembler code
8725 which will pop hard register number @var{regno} off of the stack.
8726 The code need not be optimal, since this macro is used only when
8730 @node Dispatch Tables
8731 @subsection Output of Dispatch Tables
8733 @c prevent bad page break with this line
8734 This concerns dispatch tables.
8736 @cindex dispatch table
8737 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8738 A C statement to output to the stdio stream @var{stream} an assembler
8739 pseudo-instruction to generate a difference between two labels.
8740 @var{value} and @var{rel} are the numbers of two internal labels. The
8741 definitions of these labels are output using
8742 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8743 way here. For example,
8746 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8747 @var{value}, @var{rel})
8750 You must provide this macro on machines where the addresses in a
8751 dispatch table are relative to the table's own address. If defined, GCC
8752 will also use this macro on all machines when producing PIC@.
8753 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8754 mode and flags can be read.
8757 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8758 This macro should be provided on machines where the addresses
8759 in a dispatch table are absolute.
8761 The definition should be a C statement to output to the stdio stream
8762 @var{stream} an assembler pseudo-instruction to generate a reference to
8763 a label. @var{value} is the number of an internal label whose
8764 definition is output using @code{(*targetm.asm_out.internal_label)}.
8768 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8772 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8773 Define this if the label before a jump-table needs to be output
8774 specially. The first three arguments are the same as for
8775 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8776 jump-table which follows (a @code{jump_insn} containing an
8777 @code{addr_vec} or @code{addr_diff_vec}).
8779 This feature is used on system V to output a @code{swbeg} statement
8782 If this macro is not defined, these labels are output with
8783 @code{(*targetm.asm_out.internal_label)}.
8786 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8787 Define this if something special must be output at the end of a
8788 jump-table. The definition should be a C statement to be executed
8789 after the assembler code for the table is written. It should write
8790 the appropriate code to stdio stream @var{stream}. The argument
8791 @var{table} is the jump-table insn, and @var{num} is the label-number
8792 of the preceding label.
8794 If this macro is not defined, nothing special is output at the end of
8798 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
8799 This target hook emits a label at the beginning of each FDE@. It
8800 should be defined on targets where FDEs need special labels, and it
8801 should write the appropriate label, for the FDE associated with the
8802 function declaration @var{decl}, to the stdio stream @var{stream}.
8803 The third argument, @var{for_eh}, is a boolean: true if this is for an
8804 exception table. The fourth argument, @var{empty}, is a boolean:
8805 true if this is a placeholder label for an omitted FDE@.
8807 The default is that FDEs are not given nonlocal labels.
8810 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
8811 This target hook emits a label at the beginning of the exception table.
8812 It should be defined on targets where it is desirable for the table
8813 to be broken up according to function.
8815 The default is that no label is emitted.
8818 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
8819 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.
8822 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx @var{insn})
8823 This target hook emits assembly directives required to unwind the
8824 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8825 returns @code{UI_TARGET}.
8828 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8829 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.
8832 @node Exception Region Output
8833 @subsection Assembler Commands for Exception Regions
8835 @c prevent bad page break with this line
8837 This describes commands marking the start and the end of an exception
8840 @defmac EH_FRAME_SECTION_NAME
8841 If defined, a C string constant for the name of the section containing
8842 exception handling frame unwind information. If not defined, GCC will
8843 provide a default definition if the target supports named sections.
8844 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8846 You should define this symbol if your target supports DWARF 2 frame
8847 unwind information and the default definition does not work.
8850 @defmac EH_FRAME_IN_DATA_SECTION
8851 If defined, DWARF 2 frame unwind information will be placed in the
8852 data section even though the target supports named sections. This
8853 might be necessary, for instance, if the system linker does garbage
8854 collection and sections cannot be marked as not to be collected.
8856 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8860 @defmac EH_TABLES_CAN_BE_READ_ONLY
8861 Define this macro to 1 if your target is such that no frame unwind
8862 information encoding used with non-PIC code will ever require a
8863 runtime relocation, but the linker may not support merging read-only
8864 and read-write sections into a single read-write section.
8867 @defmac MASK_RETURN_ADDR
8868 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8869 that it does not contain any extraneous set bits in it.
8872 @defmac DWARF2_UNWIND_INFO
8873 Define this macro to 0 if your target supports DWARF 2 frame unwind
8874 information, but it does not yet work with exception handling.
8875 Otherwise, if your target supports this information (if it defines
8876 @code{INCOMING_RETURN_ADDR_RTX} and either @code{UNALIGNED_INT_ASM_OP}
8877 or @code{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8880 @deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
8881 This hook defines the mechanism that will be used for exception handling
8882 by the target. If the target has ABI specified unwind tables, the hook
8883 should return @code{UI_TARGET}. If the target is to use the
8884 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8885 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8886 information, the hook should return @code{UI_DWARF2}.
8888 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8889 This may end up simplifying other parts of target-specific code. The
8890 default implementation of this hook never returns @code{UI_NONE}.
8892 Note that the value returned by this hook should be constant. It should
8893 not depend on anything except the command-line switches described by
8894 @var{opts}. In particular, the
8895 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8896 macros and builtin functions related to exception handling are set up
8897 depending on this setting.
8899 The default implementation of the hook first honors the
8900 @option{--enable-sjlj-exceptions} configure option, then
8901 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
8902 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
8903 must define this hook so that @var{opts} is used correctly.
8906 @deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8907 This variable should be set to @code{true} if the target ABI requires unwinding
8908 tables even when exceptions are not used. It must not be modified by
8909 command-line option processing.
8912 @defmac DONT_USE_BUILTIN_SETJMP
8913 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8914 should use the @code{setjmp}/@code{longjmp} functions from the C library
8915 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8918 @defmac DWARF_CIE_DATA_ALIGNMENT
8919 This macro need only be defined if the target might save registers in the
8920 function prologue at an offset to the stack pointer that is not aligned to
8921 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8922 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8923 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8924 the target supports DWARF 2 frame unwind information.
8927 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8928 Contains the value true if the target should add a zero word onto the
8929 end of a Dwarf-2 frame info section when used for exception handling.
8930 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8934 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8935 Given a register, this hook should return a parallel of registers to
8936 represent where to find the register pieces. Define this hook if the
8937 register and its mode are represented in Dwarf in non-contiguous
8938 locations, or if the register should be represented in more than one
8939 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8940 If not defined, the default is to return @code{NULL_RTX}.
8943 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8944 If some registers are represented in Dwarf-2 unwind information in
8945 multiple pieces, define this hook to fill in information about the
8946 sizes of those pieces in the table used by the unwinder at runtime.
8947 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8948 filling in a single size corresponding to each hard register;
8949 @var{address} is the address of the table.
8952 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8953 This hook is used to output a reference from a frame unwinding table to
8954 the type_info object identified by @var{sym}. It should return @code{true}
8955 if the reference was output. Returning @code{false} will cause the
8956 reference to be output using the normal Dwarf2 routines.
8959 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8960 This flag should be set to @code{true} on targets that use an ARM EABI
8961 based unwinding library, and @code{false} on other targets. This effects
8962 the format of unwinding tables, and how the unwinder in entered after
8963 running a cleanup. The default is @code{false}.
8966 @node Alignment Output
8967 @subsection Assembler Commands for Alignment
8969 @c prevent bad page break with this line
8970 This describes commands for alignment.
8972 @defmac JUMP_ALIGN (@var{label})
8973 The alignment (log base 2) to put in front of @var{label}, which is
8974 a common destination of jumps and has no fallthru incoming edge.
8976 This macro need not be defined if you don't want any special alignment
8977 to be done at such a time. Most machine descriptions do not currently
8980 Unless it's necessary to inspect the @var{label} parameter, it is better
8981 to set the variable @var{align_jumps} in the target's
8982 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8983 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8986 @deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx @var{label})
8987 The maximum number of bytes to skip before @var{label} when applying
8988 @code{JUMP_ALIGN}. This works only if
8989 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8992 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8993 The alignment (log base 2) to put in front of @var{label}, which follows
8996 This macro need not be defined if you don't want any special alignment
8997 to be done at such a time. Most machine descriptions do not currently
9001 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx @var{label})
9002 The maximum number of bytes to skip before @var{label} when applying
9003 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
9004 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9007 @defmac LOOP_ALIGN (@var{label})
9008 The alignment (log base 2) to put in front of @var{label}, which follows
9009 a @code{NOTE_INSN_LOOP_BEG} note.
9011 This macro need not be defined if you don't want any special alignment
9012 to be done at such a time. Most machine descriptions do not currently
9015 Unless it's necessary to inspect the @var{label} parameter, it is better
9016 to set the variable @code{align_loops} in the target's
9017 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9018 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
9021 @deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx @var{label})
9022 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
9023 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
9027 @defmac LABEL_ALIGN (@var{label})
9028 The alignment (log base 2) to put in front of @var{label}.
9029 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
9030 the maximum of the specified values is used.
9032 Unless it's necessary to inspect the @var{label} parameter, it is better
9033 to set the variable @code{align_labels} in the target's
9034 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9035 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9038 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx @var{label})
9039 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
9040 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
9044 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9045 A C statement to output to the stdio stream @var{stream} an assembler
9046 instruction to advance the location counter by @var{nbytes} bytes.
9047 Those bytes should be zero when loaded. @var{nbytes} will be a C
9048 expression of type @code{unsigned HOST_WIDE_INT}.
9051 @defmac ASM_NO_SKIP_IN_TEXT
9052 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9053 text section because it fails to put zeros in the bytes that are skipped.
9054 This is true on many Unix systems, where the pseudo--op to skip bytes
9055 produces no-op instructions rather than zeros when used in the text
9059 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9060 A C statement to output to the stdio stream @var{stream} an assembler
9061 command to advance the location counter to a multiple of 2 to the
9062 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9065 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9066 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9067 for padding, if necessary.
9070 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9071 A C statement to output to the stdio stream @var{stream} an assembler
9072 command to advance the location counter to a multiple of 2 to the
9073 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9074 satisfy the alignment request. @var{power} and @var{max_skip} will be
9075 a C expression of type @code{int}.
9079 @node Debugging Info
9080 @section Controlling Debugging Information Format
9082 @c prevent bad page break with this line
9083 This describes how to specify debugging information.
9086 * All Debuggers:: Macros that affect all debugging formats uniformly.
9087 * DBX Options:: Macros enabling specific options in DBX format.
9088 * DBX Hooks:: Hook macros for varying DBX format.
9089 * File Names and DBX:: Macros controlling output of file names in DBX format.
9090 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9091 * VMS Debug:: Macros for VMS debug format.
9095 @subsection Macros Affecting All Debugging Formats
9097 @c prevent bad page break with this line
9098 These macros affect all debugging formats.
9100 @defmac DBX_REGISTER_NUMBER (@var{regno})
9101 A C expression that returns the DBX register number for the compiler
9102 register number @var{regno}. In the default macro provided, the value
9103 of this expression will be @var{regno} itself. But sometimes there are
9104 some registers that the compiler knows about and DBX does not, or vice
9105 versa. In such cases, some register may need to have one number in the
9106 compiler and another for DBX@.
9108 If two registers have consecutive numbers inside GCC, and they can be
9109 used as a pair to hold a multiword value, then they @emph{must} have
9110 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9111 Otherwise, debuggers will be unable to access such a pair, because they
9112 expect register pairs to be consecutive in their own numbering scheme.
9114 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9115 does not preserve register pairs, then what you must do instead is
9116 redefine the actual register numbering scheme.
9119 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9120 A C expression that returns the integer offset value for an automatic
9121 variable having address @var{x} (an RTL expression). The default
9122 computation assumes that @var{x} is based on the frame-pointer and
9123 gives the offset from the frame-pointer. This is required for targets
9124 that produce debugging output for DBX or COFF-style debugging output
9125 for SDB and allow the frame-pointer to be eliminated when the
9126 @option{-g} options is used.
9129 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9130 A C expression that returns the integer offset value for an argument
9131 having address @var{x} (an RTL expression). The nominal offset is
9135 @defmac PREFERRED_DEBUGGING_TYPE
9136 A C expression that returns the type of debugging output GCC should
9137 produce when the user specifies just @option{-g}. Define
9138 this if you have arranged for GCC to support more than one format of
9139 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9140 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9141 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9143 When the user specifies @option{-ggdb}, GCC normally also uses the
9144 value of this macro to select the debugging output format, but with two
9145 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9146 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9147 defined, GCC uses @code{DBX_DEBUG}.
9149 The value of this macro only affects the default debugging output; the
9150 user can always get a specific type of output by using @option{-gstabs},
9151 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9155 @subsection Specific Options for DBX Output
9157 @c prevent bad page break with this line
9158 These are specific options for DBX output.
9160 @defmac DBX_DEBUGGING_INFO
9161 Define this macro if GCC should produce debugging output for DBX
9162 in response to the @option{-g} option.
9165 @defmac XCOFF_DEBUGGING_INFO
9166 Define this macro if GCC should produce XCOFF format debugging output
9167 in response to the @option{-g} option. This is a variant of DBX format.
9170 @defmac DEFAULT_GDB_EXTENSIONS
9171 Define this macro to control whether GCC should by default generate
9172 GDB's extended version of DBX debugging information (assuming DBX-format
9173 debugging information is enabled at all). If you don't define the
9174 macro, the default is 1: always generate the extended information
9175 if there is any occasion to.
9178 @defmac DEBUG_SYMS_TEXT
9179 Define this macro if all @code{.stabs} commands should be output while
9180 in the text section.
9183 @defmac ASM_STABS_OP
9184 A C string constant, including spacing, naming the assembler pseudo op to
9185 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9186 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9187 applies only to DBX debugging information format.
9190 @defmac ASM_STABD_OP
9191 A C string constant, including spacing, naming the assembler pseudo op to
9192 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9193 value is the current location. If you don't define this macro,
9194 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9198 @defmac ASM_STABN_OP
9199 A C string constant, including spacing, naming the assembler pseudo op to
9200 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9201 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9202 macro applies only to DBX debugging information format.
9205 @defmac DBX_NO_XREFS
9206 Define this macro if DBX on your system does not support the construct
9207 @samp{xs@var{tagname}}. On some systems, this construct is used to
9208 describe a forward reference to a structure named @var{tagname}.
9209 On other systems, this construct is not supported at all.
9212 @defmac DBX_CONTIN_LENGTH
9213 A symbol name in DBX-format debugging information is normally
9214 continued (split into two separate @code{.stabs} directives) when it
9215 exceeds a certain length (by default, 80 characters). On some
9216 operating systems, DBX requires this splitting; on others, splitting
9217 must not be done. You can inhibit splitting by defining this macro
9218 with the value zero. You can override the default splitting-length by
9219 defining this macro as an expression for the length you desire.
9222 @defmac DBX_CONTIN_CHAR
9223 Normally continuation is indicated by adding a @samp{\} character to
9224 the end of a @code{.stabs} string when a continuation follows. To use
9225 a different character instead, define this macro as a character
9226 constant for the character you want to use. Do not define this macro
9227 if backslash is correct for your system.
9230 @defmac DBX_STATIC_STAB_DATA_SECTION
9231 Define this macro if it is necessary to go to the data section before
9232 outputting the @samp{.stabs} pseudo-op for a non-global static
9236 @defmac DBX_TYPE_DECL_STABS_CODE
9237 The value to use in the ``code'' field of the @code{.stabs} directive
9238 for a typedef. The default is @code{N_LSYM}.
9241 @defmac DBX_STATIC_CONST_VAR_CODE
9242 The value to use in the ``code'' field of the @code{.stabs} directive
9243 for a static variable located in the text section. DBX format does not
9244 provide any ``right'' way to do this. The default is @code{N_FUN}.
9247 @defmac DBX_REGPARM_STABS_CODE
9248 The value to use in the ``code'' field of the @code{.stabs} directive
9249 for a parameter passed in registers. DBX format does not provide any
9250 ``right'' way to do this. The default is @code{N_RSYM}.
9253 @defmac DBX_REGPARM_STABS_LETTER
9254 The letter to use in DBX symbol data to identify a symbol as a parameter
9255 passed in registers. DBX format does not customarily provide any way to
9256 do this. The default is @code{'P'}.
9259 @defmac DBX_FUNCTION_FIRST
9260 Define this macro if the DBX information for a function and its
9261 arguments should precede the assembler code for the function. Normally,
9262 in DBX format, the debugging information entirely follows the assembler
9266 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9267 Define this macro, with value 1, if the value of a symbol describing
9268 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9269 relative to the start of the enclosing function. Normally, GCC uses
9270 an absolute address.
9273 @defmac DBX_LINES_FUNCTION_RELATIVE
9274 Define this macro, with value 1, if the value of a symbol indicating
9275 the current line number (@code{N_SLINE}) should be relative to the
9276 start of the enclosing function. Normally, GCC uses an absolute address.
9279 @defmac DBX_USE_BINCL
9280 Define this macro if GCC should generate @code{N_BINCL} and
9281 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9282 macro also directs GCC to output a type number as a pair of a file
9283 number and a type number within the file. Normally, GCC does not
9284 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9285 number for a type number.
9289 @subsection Open-Ended Hooks for DBX Format
9291 @c prevent bad page break with this line
9292 These are hooks for DBX format.
9294 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9295 Define this macro to say how to output to @var{stream} the debugging
9296 information for the start of a scope level for variable names. The
9297 argument @var{name} is the name of an assembler symbol (for use with
9298 @code{assemble_name}) whose value is the address where the scope begins.
9301 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9302 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9305 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9306 Define this macro if the target machine requires special handling to
9307 output an @code{N_FUN} entry for the function @var{decl}.
9310 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9311 A C statement to output DBX debugging information before code for line
9312 number @var{line} of the current source file to the stdio stream
9313 @var{stream}. @var{counter} is the number of time the macro was
9314 invoked, including the current invocation; it is intended to generate
9315 unique labels in the assembly output.
9317 This macro should not be defined if the default output is correct, or
9318 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9321 @defmac NO_DBX_FUNCTION_END
9322 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9323 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9324 On those machines, define this macro to turn this feature off without
9325 disturbing the rest of the gdb extensions.
9328 @defmac NO_DBX_BNSYM_ENSYM
9329 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9330 extension construct. On those machines, define this macro to turn this
9331 feature off without disturbing the rest of the gdb extensions.
9334 @node File Names and DBX
9335 @subsection File Names in DBX Format
9337 @c prevent bad page break with this line
9338 This describes file names in DBX format.
9340 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9341 A C statement to output DBX debugging information to the stdio stream
9342 @var{stream}, which indicates that file @var{name} is the main source
9343 file---the file specified as the input file for compilation.
9344 This macro is called only once, at the beginning of compilation.
9346 This macro need not be defined if the standard form of output
9347 for DBX debugging information is appropriate.
9349 It may be necessary to refer to a label equal to the beginning of the
9350 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9351 to do so. If you do this, you must also set the variable
9352 @var{used_ltext_label_name} to @code{true}.
9355 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9356 Define this macro, with value 1, if GCC should not emit an indication
9357 of the current directory for compilation and current source language at
9358 the beginning of the file.
9361 @defmac NO_DBX_GCC_MARKER
9362 Define this macro, with value 1, if GCC should not emit an indication
9363 that this object file was compiled by GCC@. The default is to emit
9364 an @code{N_OPT} stab at the beginning of every source file, with
9365 @samp{gcc2_compiled.} for the string and value 0.
9368 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9369 A C statement to output DBX debugging information at the end of
9370 compilation of the main source file @var{name}. Output should be
9371 written to the stdio stream @var{stream}.
9373 If you don't define this macro, nothing special is output at the end
9374 of compilation, which is correct for most machines.
9377 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9378 Define this macro @emph{instead of} defining
9379 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9380 the end of compilation is an @code{N_SO} stab with an empty string,
9381 whose value is the highest absolute text address in the file.
9386 @subsection Macros for SDB and DWARF Output
9388 @c prevent bad page break with this line
9389 Here are macros for SDB and DWARF output.
9391 @defmac SDB_DEBUGGING_INFO
9392 Define this macro if GCC should produce COFF-style debugging output
9393 for SDB in response to the @option{-g} option.
9396 @defmac DWARF2_DEBUGGING_INFO
9397 Define this macro if GCC should produce dwarf version 2 format
9398 debugging output in response to the @option{-g} option.
9400 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
9401 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9402 be emitted for each function. Instead of an integer return the enum
9403 value for the @code{DW_CC_} tag.
9406 To support optional call frame debugging information, you must also
9407 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9408 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9409 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9410 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9413 @defmac DWARF2_FRAME_INFO
9414 Define this macro to a nonzero value if GCC should always output
9415 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9416 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9417 exceptions are enabled, GCC will output this information not matter
9418 how you define @code{DWARF2_FRAME_INFO}.
9421 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
9422 This hook defines the mechanism that will be used for describing frame
9423 unwind information to the debugger. Normally the hook will return
9424 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9425 return @code{UI_NONE} otherwise.
9427 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9428 is disabled in order to always output DWARF 2 frame information.
9430 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9431 This will suppress generation of the normal debug frame unwind information.
9434 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9435 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9436 line debug info sections. This will result in much more compact line number
9437 tables, and hence is desirable if it works.
9440 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
9441 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.
9444 @deftypevr {Target Hook} bool TARGET_DELAY_SCHED2
9445 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.
9448 @deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK
9449 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.
9452 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9453 A C statement to issue assembly directives that create a difference
9454 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9457 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9458 A C statement to issue assembly directives that create a difference
9459 between the two given labels in system defined units, e.g. instruction
9460 slots on IA64 VMS, using an integer of the given size.
9463 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9464 A C statement to issue assembly directives that create a
9465 section-relative reference to the given @var{label}, using an integer of the
9466 given @var{size}. The label is known to be defined in the given @var{section}.
9469 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9470 A C statement to issue assembly directives that create a self-relative
9471 reference to the given @var{label}, using an integer of the given @var{size}.
9474 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9475 A C statement to issue assembly directives that create a reference to
9476 the DWARF table identifier @var{label} from the current section. This
9477 is used on some systems to avoid garbage collecting a DWARF table which
9478 is referenced by a function.
9481 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
9482 If defined, this target hook is a function which outputs a DTP-relative
9483 reference to the given TLS symbol of the specified size.
9486 @defmac PUT_SDB_@dots{}
9487 Define these macros to override the assembler syntax for the special
9488 SDB assembler directives. See @file{sdbout.c} for a list of these
9489 macros and their arguments. If the standard syntax is used, you need
9490 not define them yourself.
9494 Some assemblers do not support a semicolon as a delimiter, even between
9495 SDB assembler directives. In that case, define this macro to be the
9496 delimiter to use (usually @samp{\n}). It is not necessary to define
9497 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9501 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9502 Define this macro to allow references to unknown structure,
9503 union, or enumeration tags to be emitted. Standard COFF does not
9504 allow handling of unknown references, MIPS ECOFF has support for
9508 @defmac SDB_ALLOW_FORWARD_REFERENCES
9509 Define this macro to allow references to structure, union, or
9510 enumeration tags that have not yet been seen to be handled. Some
9511 assemblers choke if forward tags are used, while some require it.
9514 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9515 A C statement to output SDB debugging information before code for line
9516 number @var{line} of the current source file to the stdio stream
9517 @var{stream}. The default is to emit an @code{.ln} directive.
9522 @subsection Macros for VMS Debug Format
9524 @c prevent bad page break with this line
9525 Here are macros for VMS debug format.
9527 @defmac VMS_DEBUGGING_INFO
9528 Define this macro if GCC should produce debugging output for VMS
9529 in response to the @option{-g} option. The default behavior for VMS
9530 is to generate minimal debug info for a traceback in the absence of
9531 @option{-g} unless explicitly overridden with @option{-g0}. This
9532 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9533 @code{TARGET_OPTION_OVERRIDE}.
9536 @node Floating Point
9537 @section Cross Compilation and Floating Point
9538 @cindex cross compilation and floating point
9539 @cindex floating point and cross compilation
9541 While all modern machines use twos-complement representation for integers,
9542 there are a variety of representations for floating point numbers. This
9543 means that in a cross-compiler the representation of floating point numbers
9544 in the compiled program may be different from that used in the machine
9545 doing the compilation.
9547 Because different representation systems may offer different amounts of
9548 range and precision, all floating point constants must be represented in
9549 the target machine's format. Therefore, the cross compiler cannot
9550 safely use the host machine's floating point arithmetic; it must emulate
9551 the target's arithmetic. To ensure consistency, GCC always uses
9552 emulation to work with floating point values, even when the host and
9553 target floating point formats are identical.
9555 The following macros are provided by @file{real.h} for the compiler to
9556 use. All parts of the compiler which generate or optimize
9557 floating-point calculations must use these macros. They may evaluate
9558 their operands more than once, so operands must not have side effects.
9560 @defmac REAL_VALUE_TYPE
9561 The C data type to be used to hold a floating point value in the target
9562 machine's format. Typically this is a @code{struct} containing an
9563 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9567 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9568 Compares for equality the two values, @var{x} and @var{y}. If the target
9569 floating point format supports negative zeroes and/or NaNs,
9570 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9571 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9574 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9575 Tests whether @var{x} is less than @var{y}.
9578 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9579 Truncates @var{x} to a signed integer, rounding toward zero.
9582 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9583 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9584 @var{x} is negative, returns zero.
9587 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9588 Converts @var{string} into a floating point number in the target machine's
9589 representation for mode @var{mode}. This routine can handle both
9590 decimal and hexadecimal floating point constants, using the syntax
9591 defined by the C language for both.
9594 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9595 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9598 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9599 Determines whether @var{x} represents infinity (positive or negative).
9602 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9603 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9606 @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})
9607 Calculates an arithmetic operation on the two floating point values
9608 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9611 The operation to be performed is specified by @var{code}. Only the
9612 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9613 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9615 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9616 target's floating point format cannot represent infinity, it will call
9617 @code{abort}. Callers should check for this situation first, using
9618 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9621 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9622 Returns the negative of the floating point value @var{x}.
9625 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9626 Returns the absolute value of @var{x}.
9629 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9630 Truncates the floating point value @var{x} to fit in @var{mode}. The
9631 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9632 appropriate bit pattern to be output as a floating constant whose
9633 precision accords with mode @var{mode}.
9636 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9637 Converts a floating point value @var{x} into a double-precision integer
9638 which is then stored into @var{low} and @var{high}. If the value is not
9639 integral, it is truncated.
9642 @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})
9643 Converts a double-precision integer found in @var{low} and @var{high},
9644 into a floating point value which is then stored into @var{x}. The
9645 value is truncated to fit in mode @var{mode}.
9648 @node Mode Switching
9649 @section Mode Switching Instructions
9650 @cindex mode switching
9651 The following macros control mode switching optimizations:
9653 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9654 Define this macro if the port needs extra instructions inserted for mode
9655 switching in an optimizing compilation.
9657 For an example, the SH4 can perform both single and double precision
9658 floating point operations, but to perform a single precision operation,
9659 the FPSCR PR bit has to be cleared, while for a double precision
9660 operation, this bit has to be set. Changing the PR bit requires a general
9661 purpose register as a scratch register, hence these FPSCR sets have to
9662 be inserted before reload, i.e.@: you can't put this into instruction emitting
9663 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9665 You can have multiple entities that are mode-switched, and select at run time
9666 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9667 return nonzero for any @var{entity} that needs mode-switching.
9668 If you define this macro, you also have to define
9669 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9670 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9671 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9675 @defmac NUM_MODES_FOR_MODE_SWITCHING
9676 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9677 initializer for an array of integers. Each initializer element
9678 N refers to an entity that needs mode switching, and specifies the number
9679 of different modes that might need to be set for this entity.
9680 The position of the initializer in the initializer---starting counting at
9681 zero---determines the integer that is used to refer to the mode-switched
9683 In macros that take mode arguments / yield a mode result, modes are
9684 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9685 switch is needed / supplied.
9688 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9689 @var{entity} is an integer specifying a mode-switched entity. If
9690 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9691 return an integer value not larger than the corresponding element in
9692 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9693 be switched into prior to the execution of @var{insn}.
9696 @defmac MODE_AFTER (@var{mode}, @var{insn})
9697 If this macro is defined, it is evaluated for every @var{insn} during
9698 mode switching. It determines the mode that an insn results in (if
9699 different from the incoming mode).
9702 @defmac MODE_ENTRY (@var{entity})
9703 If this macro is defined, it is evaluated for every @var{entity} that needs
9704 mode switching. It should evaluate to an integer, which is a mode that
9705 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9706 is defined then @code{MODE_EXIT} must be defined.
9709 @defmac MODE_EXIT (@var{entity})
9710 If this macro is defined, it is evaluated for every @var{entity} that needs
9711 mode switching. It should evaluate to an integer, which is a mode that
9712 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9713 is defined then @code{MODE_ENTRY} must be defined.
9716 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9717 This macro specifies the order in which modes for @var{entity} are processed.
9718 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9719 lowest. The value of the macro should be an integer designating a mode
9720 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9721 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9722 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9725 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9726 Generate one or more insns to set @var{entity} to @var{mode}.
9727 @var{hard_reg_live} is the set of hard registers live at the point where
9728 the insn(s) are to be inserted.
9731 @node Target Attributes
9732 @section Defining target-specific uses of @code{__attribute__}
9733 @cindex target attributes
9734 @cindex machine attributes
9735 @cindex attributes, target-specific
9737 Target-specific attributes may be defined for functions, data and types.
9738 These are described using the following target hooks; they also need to
9739 be documented in @file{extend.texi}.
9741 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9742 If defined, this target hook points to an array of @samp{struct
9743 attribute_spec} (defined in @file{tree.h}) specifying the machine
9744 specific attributes for this target and some of the restrictions on the
9745 entities to which these attributes are applied and the arguments they
9749 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
9750 If defined, this target hook is a function which returns true if the
9751 machine-specific attribute named @var{name} expects an identifier
9752 given as its first argument to be passed on as a plain identifier, not
9753 subjected to name lookup. If this is not defined, the default is
9754 false for all machine-specific attributes.
9757 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
9758 If defined, this target hook is a function which returns zero if the attributes on
9759 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9760 and two if they are nearly compatible (which causes a warning to be
9761 generated). If this is not defined, machine-specific attributes are
9762 supposed always to be compatible.
9765 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9766 If defined, this target hook is a function which assigns default attributes to
9767 the newly defined @var{type}.
9770 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9771 Define this target hook if the merging of type attributes needs special
9772 handling. If defined, the result is a list of the combined
9773 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9774 that @code{comptypes} has already been called and returned 1. This
9775 function may call @code{merge_attributes} to handle machine-independent
9779 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9780 Define this target hook if the merging of decl attributes needs special
9781 handling. If defined, the result is a list of the combined
9782 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9783 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9784 when this is needed are when one attribute overrides another, or when an
9785 attribute is nullified by a subsequent definition. This function may
9786 call @code{merge_attributes} to handle machine-independent merging.
9788 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9789 If the only target-specific handling you require is @samp{dllimport}
9790 for Microsoft Windows targets, you should define the macro
9791 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9792 will then define a function called
9793 @code{merge_dllimport_decl_attributes} which can then be defined as
9794 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9795 add @code{handle_dll_attribute} in the attribute table for your port
9796 to perform initial processing of the @samp{dllimport} and
9797 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9798 @file{i386/i386.c}, for example.
9801 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
9802 @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}.
9805 @defmac TARGET_DECLSPEC
9806 Define this macro to a nonzero value if you want to treat
9807 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9808 default, this behavior is enabled only for targets that define
9809 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9810 of @code{__declspec} is via a built-in macro, but you should not rely
9811 on this implementation detail.
9814 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9815 Define this target hook if you want to be able to add attributes to a decl
9816 when it is being created. This is normally useful for back ends which
9817 wish to implement a pragma by using the attributes which correspond to
9818 the pragma's effect. The @var{node} argument is the decl which is being
9819 created. The @var{attr_ptr} argument is a pointer to the attribute list
9820 for this decl. The list itself should not be modified, since it may be
9821 shared with other decls, but attributes may be chained on the head of
9822 the list and @code{*@var{attr_ptr}} modified to point to the new
9823 attributes, or a copy of the list may be made if further changes are
9827 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
9829 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9830 into the current function, despite its having target-specific
9831 attributes, @code{false} otherwise. By default, if a function has a
9832 target specific attribute attached to it, it will not be inlined.
9835 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9836 This hook is called to parse the @code{attribute(option("..."))}, and
9837 it allows the function to set different target machine compile time
9838 options for the current function that might be different than the
9839 options specified on the command line. The hook should return
9840 @code{true} if the options are valid.
9842 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9843 the function declaration to hold a pointer to a target specific
9844 @var{struct cl_target_option} structure.
9847 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr})
9848 This hook is called to save any additional target specific information
9849 in the @var{struct cl_target_option} structure for function specific
9851 @xref{Option file format}.
9854 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct cl_target_option *@var{ptr})
9855 This hook is called to restore any additional target specific
9856 information in the @var{struct cl_target_option} structure for
9857 function specific options.
9860 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
9861 This hook is called to print any additional target specific
9862 information in the @var{struct cl_target_option} structure for
9863 function specific options.
9866 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
9867 This target hook parses the options for @code{#pragma GCC option} to
9868 set the machine specific options for functions that occur later in the
9869 input stream. The options should be the same as handled by the
9870 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9873 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
9874 Sometimes certain combinations of command options do not make sense on
9875 a particular target machine. You can override the hook
9876 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9877 once just after all the command options have been parsed.
9879 Don't use this hook to turn on various extra optimizations for
9880 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9882 If you need to do something whenever the optimization level is
9883 changed via the optimize attribute or pragma, see
9884 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9887 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
9888 This target hook returns @code{false} if the @var{caller} function
9889 cannot inline @var{callee}, based on target specific information. By
9890 default, inlining is not allowed if the callee function has function
9891 specific target options and the caller does not use the same options.
9895 @section Emulating TLS
9896 @cindex Emulated TLS
9898 For targets whose psABI does not provide Thread Local Storage via
9899 specific relocations and instruction sequences, an emulation layer is
9900 used. A set of target hooks allows this emulation layer to be
9901 configured for the requirements of a particular target. For instance
9902 the psABI may in fact specify TLS support in terms of an emulation
9905 The emulation layer works by creating a control object for every TLS
9906 object. To access the TLS object, a lookup function is provided
9907 which, when given the address of the control object, will return the
9908 address of the current thread's instance of the TLS object.
9910 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9911 Contains the name of the helper function that uses a TLS control
9912 object to locate a TLS instance. The default causes libgcc's
9913 emulated TLS helper function to be used.
9916 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9917 Contains the name of the helper function that should be used at
9918 program startup to register TLS objects that are implicitly
9919 initialized to zero. If this is @code{NULL}, all TLS objects will
9920 have explicit initializers. The default causes libgcc's emulated TLS
9921 registration function to be used.
9924 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9925 Contains the name of the section in which TLS control variables should
9926 be placed. The default of @code{NULL} allows these to be placed in
9930 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9931 Contains the name of the section in which TLS initializers should be
9932 placed. The default of @code{NULL} allows these to be placed in any
9936 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9937 Contains the prefix to be prepended to TLS control variable names.
9938 The default of @code{NULL} uses a target-specific prefix.
9941 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9942 Contains the prefix to be prepended to TLS initializer objects. The
9943 default of @code{NULL} uses a target-specific prefix.
9946 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
9947 Specifies a function that generates the FIELD_DECLs for a TLS control
9948 object type. @var{type} is the RECORD_TYPE the fields are for and
9949 @var{name} should be filled with the structure tag, if the default of
9950 @code{__emutls_object} is unsuitable. The default creates a type suitable
9951 for libgcc's emulated TLS function.
9954 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
9955 Specifies a function that generates the CONSTRUCTOR to initialize a
9956 TLS control object. @var{var} is the TLS control object, @var{decl}
9957 is the TLS object and @var{tmpl_addr} is the address of the
9958 initializer. The default initializes libgcc's emulated TLS control object.
9961 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
9962 Specifies whether the alignment of TLS control variable objects is
9963 fixed and should not be increased as some backends may do to optimize
9964 single objects. The default is false.
9967 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9968 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9969 may be used to describe emulated TLS control objects.
9972 @node MIPS Coprocessors
9973 @section Defining coprocessor specifics for MIPS targets.
9974 @cindex MIPS coprocessor-definition macros
9976 The MIPS specification allows MIPS implementations to have as many as 4
9977 coprocessors, each with as many as 32 private registers. GCC supports
9978 accessing these registers and transferring values between the registers
9979 and memory using asm-ized variables. For example:
9982 register unsigned int cp0count asm ("c0r1");
9988 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9989 names may be added as described below, or the default names may be
9990 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9992 Coprocessor registers are assumed to be epilogue-used; sets to them will
9993 be preserved even if it does not appear that the register is used again
9994 later in the function.
9996 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9997 the FPU@. One accesses COP1 registers through standard mips
9998 floating-point support; they are not included in this mechanism.
10000 There is one macro used in defining the MIPS coprocessor interface which
10001 you may want to override in subtargets; it is described below.
10003 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
10004 A comma-separated list (with leading comma) of pairs describing the
10005 alternate names of coprocessor registers. The format of each entry should be
10007 @{ @var{alternatename}, @var{register_number}@}
10013 @section Parameters for Precompiled Header Validity Checking
10014 @cindex parameters, precompiled headers
10016 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
10017 This hook returns a pointer to the data needed by
10018 @code{TARGET_PCH_VALID_P} and sets
10019 @samp{*@var{sz}} to the size of the data in bytes.
10022 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
10023 This hook checks whether the options used to create a PCH file are
10024 compatible with the current settings. It returns @code{NULL}
10025 if so and a suitable error message if not. Error messages will
10026 be presented to the user and must be localized using @samp{_(@var{msg})}.
10028 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
10029 when the PCH file was created and @var{sz} is the size of that data in bytes.
10030 It's safe to assume that the data was created by the same version of the
10031 compiler, so no format checking is needed.
10033 The default definition of @code{default_pch_valid_p} should be
10034 suitable for most targets.
10037 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
10038 If this hook is nonnull, the default implementation of
10039 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
10040 of @code{target_flags}. @var{pch_flags} specifies the value that
10041 @code{target_flags} had when the PCH file was created. The return
10042 value is the same as for @code{TARGET_PCH_VALID_P}.
10046 @section C++ ABI parameters
10047 @cindex parameters, c++ abi
10049 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10050 Define this hook to override the integer type used for guard variables.
10051 These are used to implement one-time construction of static objects. The
10052 default is long_long_integer_type_node.
10055 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10056 This hook determines how guard variables are used. It should return
10057 @code{false} (the default) if the first byte should be used. A return value of
10058 @code{true} indicates that only the least significant bit should be used.
10061 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10062 This hook returns the size of the cookie to use when allocating an array
10063 whose elements have the indicated @var{type}. Assumes that it is already
10064 known that a cookie is needed. The default is
10065 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10066 IA64/Generic C++ ABI@.
10069 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10070 This hook should return @code{true} if the element size should be stored in
10071 array cookies. The default is to return @code{false}.
10074 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10075 If defined by a backend this hook allows the decision made to export
10076 class @var{type} to be overruled. Upon entry @var{import_export}
10077 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10078 to be imported and 0 otherwise. This function should return the
10079 modified value and perform any other actions necessary to support the
10080 backend's targeted operating system.
10083 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10084 This hook should return @code{true} if constructors and destructors return
10085 the address of the object created/destroyed. The default is to return
10089 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10090 This hook returns true if the key method for a class (i.e., the method
10091 which, if defined in the current translation unit, causes the virtual
10092 table to be emitted) may be an inline function. Under the standard
10093 Itanium C++ ABI the key method may be an inline function so long as
10094 the function is not declared inline in the class definition. Under
10095 some variants of the ABI, an inline function can never be the key
10096 method. The default is to return @code{true}.
10099 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10100 @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}.
10103 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10104 This hook returns true (the default) if virtual tables and other
10105 similar implicit class data objects are always COMDAT if they have
10106 external linkage. If this hook returns false, then class data for
10107 classes whose virtual table will be emitted in only one translation
10108 unit will not be COMDAT.
10111 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10112 This hook returns true (the default) if the RTTI information for
10113 the basic types which is defined in the C++ runtime should always
10114 be COMDAT, false if it should not be COMDAT.
10117 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10118 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10119 should be used to register static destructors when @option{-fuse-cxa-atexit}
10120 is in effect. The default is to return false to use @code{__cxa_atexit}.
10123 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10124 This hook returns true if the target @code{atexit} function can be used
10125 in the same manner as @code{__cxa_atexit} to register C++ static
10126 destructors. This requires that @code{atexit}-registered functions in
10127 shared libraries are run in the correct order when the libraries are
10128 unloaded. The default is to return false.
10131 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10132 @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).
10135 @node Named Address Spaces
10136 @section Adding support for named address spaces
10137 @cindex named address spaces
10139 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10140 standards committee, @cite{Programming Languages - C - Extensions to
10141 support embedded processors}, specifies a syntax for embedded
10142 processors to specify alternate address spaces. You can configure a
10143 GCC port to support section 5.1 of the draft report to add support for
10144 address spaces other than the default address space. These address
10145 spaces are new keywords that are similar to the @code{volatile} and
10146 @code{const} type attributes.
10148 Pointers to named address spaces can have a different size than
10149 pointers to the generic address space.
10151 For example, the SPU port uses the @code{__ea} address space to refer
10152 to memory in the host processor, rather than memory local to the SPU
10153 processor. Access to memory in the @code{__ea} address space involves
10154 issuing DMA operations to move data between the host processor and the
10155 local processor memory address space. Pointers in the @code{__ea}
10156 address space are either 32 bits or 64 bits based on the
10157 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10160 Internally, address spaces are represented as a small integer in the
10161 range 0 to 15 with address space 0 being reserved for the generic
10164 To register a named address space qualifier keyword with the C front end,
10165 the target may call the @code{c_register_addr_space} routine. For example,
10166 the SPU port uses the following to declare @code{__ea} as the keyword for
10167 named address space #1:
10169 #define ADDR_SPACE_EA 1
10170 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10173 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10174 Define this to return the machine mode to use for pointers to
10175 @var{address_space} if the target supports named address spaces.
10176 The default version of this hook returns @code{ptr_mode} for the
10177 generic address space only.
10180 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10181 Define this to return the machine mode to use for addresses in
10182 @var{address_space} if the target supports named address spaces.
10183 The default version of this hook returns @code{Pmode} for the
10184 generic address space only.
10187 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (enum machine_mode @var{mode}, addr_space_t @var{as})
10188 Define this to return nonzero if the port can handle pointers
10189 with machine mode @var{mode} to address space @var{as}. This target
10190 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10191 except that it includes explicit named address space support. The default
10192 version of this hook returns true for the modes returned by either the
10193 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10194 target hooks for the given address space.
10197 @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})
10198 Define this to return true if @var{exp} is a valid address for mode
10199 @var{mode} in the named address space @var{as}. The @var{strict}
10200 parameter says whether strict addressing is in effect after reload has
10201 finished. This target hook is the same as the
10202 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10203 explicit named address space support.
10206 @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})
10207 Define this to modify an invalid address @var{x} to be a valid address
10208 with mode @var{mode} in the named address space @var{as}. This target
10209 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10210 except that it includes explicit named address space support.
10213 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{superset}, addr_space_t @var{subset})
10214 Define this to return whether the @var{subset} named address space is
10215 contained within the @var{superset} named address space. Pointers to
10216 a named address space that is a subset of another named address space
10217 will be converted automatically without a cast if used together in
10218 arithmetic operations. Pointers to a superset address space can be
10219 converted to pointers to a subset address space via explicit casts.
10222 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10223 Define this to convert the pointer expression represented by the RTL
10224 @var{op} with type @var{from_type} that points to a named address
10225 space to a new pointer expression with type @var{to_type} that points
10226 to a different named address space. When this hook it called, it is
10227 guaranteed that one of the two address spaces is a subset of the other,
10228 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10232 @section Miscellaneous Parameters
10233 @cindex parameters, miscellaneous
10235 @c prevent bad page break with this line
10236 Here are several miscellaneous parameters.
10238 @defmac HAS_LONG_COND_BRANCH
10239 Define this boolean macro to indicate whether or not your architecture
10240 has conditional branches that can span all of memory. It is used in
10241 conjunction with an optimization that partitions hot and cold basic
10242 blocks into separate sections of the executable. If this macro is
10243 set to false, gcc will convert any conditional branches that attempt
10244 to cross between sections into unconditional branches or indirect jumps.
10247 @defmac HAS_LONG_UNCOND_BRANCH
10248 Define this boolean macro to indicate whether or not your architecture
10249 has unconditional branches that can span all of memory. It is used in
10250 conjunction with an optimization that partitions hot and cold basic
10251 blocks into separate sections of the executable. If this macro is
10252 set to false, gcc will convert any unconditional branches that attempt
10253 to cross between sections into indirect jumps.
10256 @defmac CASE_VECTOR_MODE
10257 An alias for a machine mode name. This is the machine mode that
10258 elements of a jump-table should have.
10261 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10262 Optional: return the preferred mode for an @code{addr_diff_vec}
10263 when the minimum and maximum offset are known. If you define this,
10264 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10265 To make this work, you also have to define @code{INSN_ALIGN} and
10266 make the alignment for @code{addr_diff_vec} explicit.
10267 The @var{body} argument is provided so that the offset_unsigned and scale
10268 flags can be updated.
10271 @defmac CASE_VECTOR_PC_RELATIVE
10272 Define this macro to be a C expression to indicate when jump-tables
10273 should contain relative addresses. You need not define this macro if
10274 jump-tables never contain relative addresses, or jump-tables should
10275 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10279 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10280 This function return the smallest number of different values for which it
10281 is best to use a jump-table instead of a tree of conditional branches.
10282 The default is four for machines with a @code{casesi} instruction and
10283 five otherwise. This is best for most machines.
10286 @defmac CASE_USE_BIT_TESTS
10287 Define this macro to be a C expression to indicate whether C switch
10288 statements may be implemented by a sequence of bit tests. This is
10289 advantageous on processors that can efficiently implement left shift
10290 of 1 by the number of bits held in a register, but inappropriate on
10291 targets that would require a loop. By default, this macro returns
10292 @code{true} if the target defines an @code{ashlsi3} pattern, and
10293 @code{false} otherwise.
10296 @defmac WORD_REGISTER_OPERATIONS
10297 Define this macro if operations between registers with integral mode
10298 smaller than a word are always performed on the entire register.
10299 Most RISC machines have this property and most CISC machines do not.
10302 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10303 Define this macro to be a C expression indicating when insns that read
10304 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10305 bits outside of @var{mem_mode} to be either the sign-extension or the
10306 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10307 of @var{mem_mode} for which the
10308 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10309 @code{UNKNOWN} for other modes.
10311 This macro is not called with @var{mem_mode} non-integral or with a width
10312 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10313 value in this case. Do not define this macro if it would always return
10314 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10315 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10317 You may return a non-@code{UNKNOWN} value even if for some hard registers
10318 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10319 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10320 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10321 integral mode larger than this but not larger than @code{word_mode}.
10323 You must return @code{UNKNOWN} if for some hard registers that allow this
10324 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10325 @code{word_mode}, but that they can change to another integral mode that
10326 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10329 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10330 Define this macro if loading short immediate values into registers sign
10334 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
10335 Define this macro if the same instructions that convert a floating
10336 point number to a signed fixed point number also convert validly to an
10340 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
10341 When @option{-ffast-math} is in effect, GCC tries to optimize
10342 divisions by the same divisor, by turning them into multiplications by
10343 the reciprocal. This target hook specifies the minimum number of divisions
10344 that should be there for GCC to perform the optimization for a variable
10345 of mode @var{mode}. The default implementation returns 3 if the machine
10346 has an instruction for the division, and 2 if it does not.
10350 The maximum number of bytes that a single instruction can move quickly
10351 between memory and registers or between two memory locations.
10354 @defmac MAX_MOVE_MAX
10355 The maximum number of bytes that a single instruction can move quickly
10356 between memory and registers or between two memory locations. If this
10357 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10358 constant value that is the largest value that @code{MOVE_MAX} can have
10362 @defmac SHIFT_COUNT_TRUNCATED
10363 A C expression that is nonzero if on this machine the number of bits
10364 actually used for the count of a shift operation is equal to the number
10365 of bits needed to represent the size of the object being shifted. When
10366 this macro is nonzero, the compiler will assume that it is safe to omit
10367 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10368 truncates the count of a shift operation. On machines that have
10369 instructions that act on bit-fields at variable positions, which may
10370 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10371 also enables deletion of truncations of the values that serve as
10372 arguments to bit-field instructions.
10374 If both types of instructions truncate the count (for shifts) and
10375 position (for bit-field operations), or if no variable-position bit-field
10376 instructions exist, you should define this macro.
10378 However, on some machines, such as the 80386 and the 680x0, truncation
10379 only applies to shift operations and not the (real or pretended)
10380 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10381 such machines. Instead, add patterns to the @file{md} file that include
10382 the implied truncation of the shift instructions.
10384 You need not define this macro if it would always have the value of zero.
10387 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10388 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
10389 This function describes how the standard shift patterns for @var{mode}
10390 deal with shifts by negative amounts or by more than the width of the mode.
10391 @xref{shift patterns}.
10393 On many machines, the shift patterns will apply a mask @var{m} to the
10394 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10395 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10396 this is true for mode @var{mode}, the function should return @var{m},
10397 otherwise it should return 0. A return value of 0 indicates that no
10398 particular behavior is guaranteed.
10400 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10401 @emph{not} apply to general shift rtxes; it applies only to instructions
10402 that are generated by the named shift patterns.
10404 The default implementation of this function returns
10405 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10406 and 0 otherwise. This definition is always safe, but if
10407 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10408 nevertheless truncate the shift count, you may get better code
10412 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10413 A C expression which is nonzero if on this machine it is safe to
10414 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10415 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10416 operating on it as if it had only @var{outprec} bits.
10418 On many machines, this expression can be 1.
10420 @c rearranged this, removed the phrase "it is reported that". this was
10421 @c to fix an overfull hbox. --mew 10feb93
10422 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10423 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10424 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10425 such cases may improve things.
10428 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
10429 The representation of an integral mode can be such that the values
10430 are always extended to a wider integral mode. Return
10431 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10432 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10433 otherwise. (Currently, none of the targets use zero-extended
10434 representation this way so unlike @code{LOAD_EXTEND_OP},
10435 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10436 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10437 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10438 widest integral mode and currently we take advantage of this fact.)
10440 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10441 value even if the extension is not performed on certain hard registers
10442 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10443 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10445 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10446 describe two related properties. If you define
10447 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10448 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10451 In order to enforce the representation of @code{mode},
10452 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10456 @defmac STORE_FLAG_VALUE
10457 A C expression describing the value returned by a comparison operator
10458 with an integral mode and stored by a store-flag instruction
10459 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10460 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10461 comparison operators whose results have a @code{MODE_INT} mode.
10463 A value of 1 or @minus{}1 means that the instruction implementing the
10464 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10465 and 0 when the comparison is false. Otherwise, the value indicates
10466 which bits of the result are guaranteed to be 1 when the comparison is
10467 true. This value is interpreted in the mode of the comparison
10468 operation, which is given by the mode of the first operand in the
10469 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10470 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10473 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10474 generate code that depends only on the specified bits. It can also
10475 replace comparison operators with equivalent operations if they cause
10476 the required bits to be set, even if the remaining bits are undefined.
10477 For example, on a machine whose comparison operators return an
10478 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10479 @samp{0x80000000}, saying that just the sign bit is relevant, the
10483 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10487 can be converted to
10490 (ashift:SI @var{x} (const_int @var{n}))
10494 where @var{n} is the appropriate shift count to move the bit being
10495 tested into the sign bit.
10497 There is no way to describe a machine that always sets the low-order bit
10498 for a true value, but does not guarantee the value of any other bits,
10499 but we do not know of any machine that has such an instruction. If you
10500 are trying to port GCC to such a machine, include an instruction to
10501 perform a logical-and of the result with 1 in the pattern for the
10502 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10504 Often, a machine will have multiple instructions that obtain a value
10505 from a comparison (or the condition codes). Here are rules to guide the
10506 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10511 Use the shortest sequence that yields a valid definition for
10512 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10513 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10514 comparison operators to do so because there may be opportunities to
10515 combine the normalization with other operations.
10518 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10519 slightly preferred on machines with expensive jumps and 1 preferred on
10523 As a second choice, choose a value of @samp{0x80000001} if instructions
10524 exist that set both the sign and low-order bits but do not define the
10528 Otherwise, use a value of @samp{0x80000000}.
10531 Many machines can produce both the value chosen for
10532 @code{STORE_FLAG_VALUE} and its negation in the same number of
10533 instructions. On those machines, you should also define a pattern for
10534 those cases, e.g., one matching
10537 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10540 Some machines can also perform @code{and} or @code{plus} operations on
10541 condition code values with less instructions than the corresponding
10542 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10543 machines, define the appropriate patterns. Use the names @code{incscc}
10544 and @code{decscc}, respectively, for the patterns which perform
10545 @code{plus} or @code{minus} operations on condition code values. See
10546 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10547 find such instruction sequences on other machines.
10549 If this macro is not defined, the default value, 1, is used. You need
10550 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10551 instructions, or if the value generated by these instructions is 1.
10554 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10555 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10556 returned when comparison operators with floating-point results are true.
10557 Define this macro on machines that have comparison operations that return
10558 floating-point values. If there are no such operations, do not define
10562 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10563 A C expression that gives a rtx representing the nonzero true element
10564 for vector comparisons. The returned rtx should be valid for the inner
10565 mode of @var{mode} which is guaranteed to be a vector mode. Define
10566 this macro on machines that have vector comparison operations that
10567 return a vector result. If there are no such operations, do not define
10568 this macro. Typically, this macro is defined as @code{const1_rtx} or
10569 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10570 the compiler optimizing such vector comparison operations for the
10574 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10575 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10576 A C expression that indicates whether the architecture defines a value
10577 for @code{clz} or @code{ctz} with a zero operand.
10578 A result of @code{0} indicates the value is undefined.
10579 If the value is defined for only the RTL expression, the macro should
10580 evaluate to @code{1}; if the value applies also to the corresponding optab
10581 entry (which is normally the case if it expands directly into
10582 the corresponding RTL), then the macro should evaluate to @code{2}.
10583 In the cases where the value is defined, @var{value} should be set to
10586 If this macro is not defined, the value of @code{clz} or
10587 @code{ctz} at zero is assumed to be undefined.
10589 This macro must be defined if the target's expansion for @code{ffs}
10590 relies on a particular value to get correct results. Otherwise it
10591 is not necessary, though it may be used to optimize some corner cases, and
10592 to provide a default expansion for the @code{ffs} optab.
10594 Note that regardless of this macro the ``definedness'' of @code{clz}
10595 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10596 visible to the user. Thus one may be free to adjust the value at will
10597 to match the target expansion of these operations without fear of
10602 An alias for the machine mode for pointers. On most machines, define
10603 this to be the integer mode corresponding to the width of a hardware
10604 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10605 On some machines you must define this to be one of the partial integer
10606 modes, such as @code{PSImode}.
10608 The width of @code{Pmode} must be at least as large as the value of
10609 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10610 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10614 @defmac FUNCTION_MODE
10615 An alias for the machine mode used for memory references to functions
10616 being called, in @code{call} RTL expressions. On most CISC machines,
10617 where an instruction can begin at any byte address, this should be
10618 @code{QImode}. On most RISC machines, where all instructions have fixed
10619 size and alignment, this should be a mode with the same size and alignment
10620 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10623 @defmac STDC_0_IN_SYSTEM_HEADERS
10624 In normal operation, the preprocessor expands @code{__STDC__} to the
10625 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10626 hosts, like Solaris, the system compiler uses a different convention,
10627 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10628 strict conformance to the C Standard.
10630 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10631 convention when processing system header files, but when processing user
10632 files @code{__STDC__} will always expand to 1.
10635 @defmac NO_IMPLICIT_EXTERN_C
10636 Define this macro if the system header files support C++ as well as C@.
10637 This macro inhibits the usual method of using system header files in
10638 C++, which is to pretend that the file's contents are enclosed in
10639 @samp{extern "C" @{@dots{}@}}.
10644 @defmac REGISTER_TARGET_PRAGMAS ()
10645 Define this macro if you want to implement any target-specific pragmas.
10646 If defined, it is a C expression which makes a series of calls to
10647 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10648 for each pragma. The macro may also do any
10649 setup required for the pragmas.
10651 The primary reason to define this macro is to provide compatibility with
10652 other compilers for the same target. In general, we discourage
10653 definition of target-specific pragmas for GCC@.
10655 If the pragma can be implemented by attributes then you should consider
10656 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10658 Preprocessor macros that appear on pragma lines are not expanded. All
10659 @samp{#pragma} directives that do not match any registered pragma are
10660 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10663 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10664 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10666 Each call to @code{c_register_pragma} or
10667 @code{c_register_pragma_with_expansion} establishes one pragma. The
10668 @var{callback} routine will be called when the preprocessor encounters a
10672 #pragma [@var{space}] @var{name} @dots{}
10675 @var{space} is the case-sensitive namespace of the pragma, or
10676 @code{NULL} to put the pragma in the global namespace. The callback
10677 routine receives @var{pfile} as its first argument, which can be passed
10678 on to cpplib's functions if necessary. You can lex tokens after the
10679 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10680 callback will be silently ignored. The end of the line is indicated by
10681 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10682 arguments of pragmas registered with
10683 @code{c_register_pragma_with_expansion} but not on the arguments of
10684 pragmas registered with @code{c_register_pragma}.
10686 Note that the use of @code{pragma_lex} is specific to the C and C++
10687 compilers. It will not work in the Java or Fortran compilers, or any
10688 other language compilers for that matter. Thus if @code{pragma_lex} is going
10689 to be called from target-specific code, it must only be done so when
10690 building the C and C++ compilers. This can be done by defining the
10691 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10692 target entry in the @file{config.gcc} file. These variables should name
10693 the target-specific, language-specific object file which contains the
10694 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10695 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10696 how to build this object file.
10699 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10700 Define this macro if macros should be expanded in the
10701 arguments of @samp{#pragma pack}.
10704 @deftypevr {Target Hook} bool TARGET_HANDLE_PRAGMA_EXTERN_PREFIX
10705 True if @code{#pragma extern_prefix} is to be supported.
10708 @defmac TARGET_DEFAULT_PACK_STRUCT
10709 If your target requires a structure packing default other than 0 (meaning
10710 the machine default), define this macro to the necessary value (in bytes).
10711 This must be a value that would also be valid to use with
10712 @samp{#pragma pack()} (that is, a small power of two).
10715 @defmac DOLLARS_IN_IDENTIFIERS
10716 Define this macro to control use of the character @samp{$} in
10717 identifier names for the C family of languages. 0 means @samp{$} is
10718 not allowed by default; 1 means it is allowed. 1 is the default;
10719 there is no need to define this macro in that case.
10722 @defmac NO_DOLLAR_IN_LABEL
10723 Define this macro if the assembler does not accept the character
10724 @samp{$} in label names. By default constructors and destructors in
10725 G++ have @samp{$} in the identifiers. If this macro is defined,
10726 @samp{.} is used instead.
10729 @defmac NO_DOT_IN_LABEL
10730 Define this macro if the assembler does not accept the character
10731 @samp{.} in label names. By default constructors and destructors in G++
10732 have names that use @samp{.}. If this macro is defined, these names
10733 are rewritten to avoid @samp{.}.
10736 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10737 Define this macro as a C expression that is nonzero if it is safe for the
10738 delay slot scheduler to place instructions in the delay slot of @var{insn},
10739 even if they appear to use a resource set or clobbered in @var{insn}.
10740 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10741 every @code{call_insn} has this behavior. On machines where some @code{insn}
10742 or @code{jump_insn} is really a function call and hence has this behavior,
10743 you should define this macro.
10745 You need not define this macro if it would always return zero.
10748 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10749 Define this macro as a C expression that is nonzero if it is safe for the
10750 delay slot scheduler to place instructions in the delay slot of @var{insn},
10751 even if they appear to set or clobber a resource referenced in @var{insn}.
10752 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10753 some @code{insn} or @code{jump_insn} is really a function call and its operands
10754 are registers whose use is actually in the subroutine it calls, you should
10755 define this macro. Doing so allows the delay slot scheduler to move
10756 instructions which copy arguments into the argument registers into the delay
10757 slot of @var{insn}.
10759 You need not define this macro if it would always return zero.
10762 @defmac MULTIPLE_SYMBOL_SPACES
10763 Define this macro as a C expression that is nonzero if, in some cases,
10764 global symbols from one translation unit may not be bound to undefined
10765 symbols in another translation unit without user intervention. For
10766 instance, under Microsoft Windows symbols must be explicitly imported
10767 from shared libraries (DLLs).
10769 You need not define this macro if it would always evaluate to zero.
10772 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10773 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10774 any hard regs the port wishes to automatically clobber for an asm.
10775 It should return the result of the last @code{tree_cons} used to add a
10776 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10777 corresponding parameters to the asm and may be inspected to avoid
10778 clobbering a register that is an input or output of the asm. You can use
10779 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10780 for overlap with regards to asm-declared registers.
10783 @defmac MATH_LIBRARY
10784 Define this macro as a C string constant for the linker argument to link
10785 in the system math library, minus the initial @samp{"-l"}, or
10786 @samp{""} if the target does not have a
10787 separate math library.
10789 You need only define this macro if the default of @samp{"m"} is wrong.
10792 @defmac LIBRARY_PATH_ENV
10793 Define this macro as a C string constant for the environment variable that
10794 specifies where the linker should look for libraries.
10796 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10800 @defmac TARGET_POSIX_IO
10801 Define this macro if the target supports the following POSIX@ file
10802 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10803 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10804 to use file locking when exiting a program, which avoids race conditions
10805 if the program has forked. It will also create directories at run-time
10806 for cross-profiling.
10809 @defmac MAX_CONDITIONAL_EXECUTE
10811 A C expression for the maximum number of instructions to execute via
10812 conditional execution instructions instead of a branch. A value of
10813 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10814 1 if it does use cc0.
10817 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10818 Used if the target needs to perform machine-dependent modifications on the
10819 conditionals used for turning basic blocks into conditionally executed code.
10820 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10821 contains information about the currently processed blocks. @var{true_expr}
10822 and @var{false_expr} are the tests that are used for converting the
10823 then-block and the else-block, respectively. Set either @var{true_expr} or
10824 @var{false_expr} to a null pointer if the tests cannot be converted.
10827 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10828 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10829 if-statements into conditions combined by @code{and} and @code{or} operations.
10830 @var{bb} contains the basic block that contains the test that is currently
10831 being processed and about to be turned into a condition.
10834 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10835 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10836 be converted to conditional execution format. @var{ce_info} points to
10837 a data structure, @code{struct ce_if_block}, which contains information
10838 about the currently processed blocks.
10841 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10842 A C expression to perform any final machine dependent modifications in
10843 converting code to conditional execution. The involved basic blocks
10844 can be found in the @code{struct ce_if_block} structure that is pointed
10845 to by @var{ce_info}.
10848 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10849 A C expression to cancel any machine dependent modifications in
10850 converting code to conditional execution. The involved basic blocks
10851 can be found in the @code{struct ce_if_block} structure that is pointed
10852 to by @var{ce_info}.
10855 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10856 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10857 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10860 @defmac IFCVT_EXTRA_FIELDS
10861 If defined, it should expand to a set of field declarations that will be
10862 added to the @code{struct ce_if_block} structure. These should be initialized
10863 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10866 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
10867 If non-null, this hook performs a target-specific pass over the
10868 instruction stream. The compiler will run it at all optimization levels,
10869 just before the point at which it normally does delayed-branch scheduling.
10871 The exact purpose of the hook varies from target to target. Some use
10872 it to do transformations that are necessary for correctness, such as
10873 laying out in-function constant pools or avoiding hardware hazards.
10874 Others use it as an opportunity to do some machine-dependent optimizations.
10876 You need not implement the hook if it has nothing to do. The default
10877 definition is null.
10880 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
10881 Define this hook if you have any machine-specific built-in functions
10882 that need to be defined. It should be a function that performs the
10885 Machine specific built-in functions can be useful to expand special machine
10886 instructions that would otherwise not normally be generated because
10887 they have no equivalent in the source language (for example, SIMD vector
10888 instructions or prefetch instructions).
10890 To create a built-in function, call the function
10891 @code{lang_hooks.builtin_function}
10892 which is defined by the language front end. You can use any type nodes set
10893 up by @code{build_common_tree_nodes};
10894 only language front ends that use those two functions will call
10895 @samp{TARGET_INIT_BUILTINS}.
10898 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
10899 Define this hook if you have any machine-specific built-in functions
10900 that need to be defined. It should be a function that returns the
10901 builtin function declaration for the builtin function code @var{code}.
10902 If there is no such builtin and it cannot be initialized at this time
10903 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10904 If @var{code} is out of range the function should return
10905 @code{error_mark_node}.
10908 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10910 Expand a call to a machine specific built-in function that was set up by
10911 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10912 function call; the result should go to @var{target} if that is
10913 convenient, and have mode @var{mode} if that is convenient.
10914 @var{subtarget} may be used as the target for computing one of
10915 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10916 ignored. This function should return the result of the call to the
10920 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
10921 Select a replacement for a machine specific built-in function that
10922 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10923 @emph{before} regular type checking, and so allows the target to
10924 implement a crude form of function overloading. @var{fndecl} is the
10925 declaration of the built-in function. @var{arglist} is the list of
10926 arguments passed to the built-in function. The result is a
10927 complete expression that implements the operation, usually
10928 another @code{CALL_EXPR}.
10929 @var{arglist} really has type @samp{VEC(tree,gc)*}
10932 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
10933 Fold a call to a machine specific built-in function that was set up by
10934 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10935 built-in function. @var{n_args} is the number of arguments passed to
10936 the function; the arguments themselves are pointed to by @var{argp}.
10937 The result is another tree containing a simplified expression for the
10938 call's result. If @var{ignore} is true the value will be ignored.
10941 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const_rtx @var{insn})
10943 Take an instruction in @var{insn} and return NULL if it is valid within a
10944 low-overhead loop, otherwise return a string explaining why doloop
10945 could not be applied.
10947 Many targets use special registers for low-overhead looping. For any
10948 instruction that clobbers these this function should return a string indicating
10949 the reason why the doloop could not be applied.
10950 By default, the RTL loop optimizer does not use a present doloop pattern for
10951 loops containing function calls or branch on table instructions.
10954 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10956 Take a branch insn in @var{branch1} and another in @var{branch2}.
10957 Return true if redirecting @var{branch1} to the destination of
10958 @var{branch2} is possible.
10960 On some targets, branches may have a limited range. Optimizing the
10961 filling of delay slots can result in branches being redirected, and this
10962 may in turn cause a branch offset to overflow.
10965 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
10966 This target hook returns @code{true} if @var{x} is considered to be commutative.
10967 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10968 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10969 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10972 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
10974 When the initial value of a hard register has been copied in a pseudo
10975 register, it is often not necessary to actually allocate another register
10976 to this pseudo register, because the original hard register or a stack slot
10977 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10978 is called at the start of register allocation once for each hard register
10979 that had its initial value copied by using
10980 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10981 Possible values are @code{NULL_RTX}, if you don't want
10982 to do any special allocation, a @code{REG} rtx---that would typically be
10983 the hard register itself, if it is known not to be clobbered---or a
10985 If you are returning a @code{MEM}, this is only a hint for the allocator;
10986 it might decide to use another register anyways.
10987 You may use @code{current_function_leaf_function} in the hook, functions
10988 that use @code{REG_N_SETS}, to determine if the hard
10989 register in question will not be clobbered.
10990 The default value of this hook is @code{NULL}, which disables any special
10994 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
10995 This target hook returns nonzero if @var{x}, an @code{unspec} or
10996 @code{unspec_volatile} operation, might cause a trap. Targets can use
10997 this hook to enhance precision of analysis for @code{unspec} and
10998 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10999 to analyze inner elements of @var{x} in which case @var{flags} should be
11003 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
11004 The compiler invokes this hook whenever it changes its current function
11005 context (@code{cfun}). You can define this function if
11006 the back end needs to perform any initialization or reset actions on a
11007 per-function basis. For example, it may be used to implement function
11008 attributes that affect register usage or code generation patterns.
11009 The argument @var{decl} is the declaration for the new function context,
11010 and may be null to indicate that the compiler has left a function context
11011 and is returning to processing at the top level.
11012 The default hook function does nothing.
11014 GCC sets @code{cfun} to a dummy function context during initialization of
11015 some parts of the back end. The hook function is not invoked in this
11016 situation; you need not worry about the hook being invoked recursively,
11017 or when the back end is in a partially-initialized state.
11018 @code{cfun} might be @code{NULL} to indicate processing at top level,
11019 outside of any function scope.
11022 @defmac TARGET_OBJECT_SUFFIX
11023 Define this macro to be a C string representing the suffix for object
11024 files on your target machine. If you do not define this macro, GCC will
11025 use @samp{.o} as the suffix for object files.
11028 @defmac TARGET_EXECUTABLE_SUFFIX
11029 Define this macro to be a C string representing the suffix to be
11030 automatically added to executable files on your target machine. If you
11031 do not define this macro, GCC will use the null string as the suffix for
11035 @defmac COLLECT_EXPORT_LIST
11036 If defined, @code{collect2} will scan the individual object files
11037 specified on its command line and create an export list for the linker.
11038 Define this macro for systems like AIX, where the linker discards
11039 object files that are not referenced from @code{main} and uses export
11043 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11044 Define this macro to a C expression representing a variant of the
11045 method call @var{mdecl}, if Java Native Interface (JNI) methods
11046 must be invoked differently from other methods on your target.
11047 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11048 the @code{stdcall} calling convention and this macro is then
11049 defined as this expression:
11052 build_type_attribute_variant (@var{mdecl},
11054 (get_identifier ("stdcall"),
11059 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11060 This target hook returns @code{true} past the point in which new jump
11061 instructions could be created. On machines that require a register for
11062 every jump such as the SHmedia ISA of SH5, this point would typically be
11063 reload, so this target hook should be defined to a function such as:
11067 cannot_modify_jumps_past_reload_p ()
11069 return (reload_completed || reload_in_progress);
11074 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11075 This target hook returns a register class for which branch target register
11076 optimizations should be applied. All registers in this class should be
11077 usable interchangeably. After reload, registers in this class will be
11078 re-allocated and loads will be hoisted out of loops and be subjected
11079 to inter-block scheduling.
11082 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11083 Branch target register optimization will by default exclude callee-saved
11085 that are not already live during the current function; if this target hook
11086 returns true, they will be included. The target code must than make sure
11087 that all target registers in the class returned by
11088 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11089 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11090 epilogues have already been generated. Note, even if you only return
11091 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11092 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11093 to reserve space for caller-saved target registers.
11096 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11097 This target hook returns true if the target supports conditional execution.
11098 This target hook is required only when the target has several different
11099 modes and they have different conditional execution capability, such as ARM.
11102 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop})
11103 This target hook returns a new value for the number of times @var{loop}
11104 should be unrolled. The parameter @var{nunroll} is the number of times
11105 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11106 the loop, which is going to be checked for unrolling. This target hook
11107 is required only when the target has special constraints like maximum
11108 number of memory accesses.
11111 @defmac POWI_MAX_MULTS
11112 If defined, this macro is interpreted as a signed integer C expression
11113 that specifies the maximum number of floating point multiplications
11114 that should be emitted when expanding exponentiation by an integer
11115 constant inline. When this value is defined, exponentiation requiring
11116 more than this number of multiplications is implemented by calling the
11117 system library's @code{pow}, @code{powf} or @code{powl} routines.
11118 The default value places no upper bound on the multiplication count.
11121 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11122 This target hook should register any extra include files for the
11123 target. The parameter @var{stdinc} indicates if normal include files
11124 are present. The parameter @var{sysroot} is the system root directory.
11125 The parameter @var{iprefix} is the prefix for the gcc directory.
11128 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11129 This target hook should register any extra include files for the
11130 target before any standard headers. The parameter @var{stdinc}
11131 indicates if normal include files are present. The parameter
11132 @var{sysroot} is the system root directory. The parameter
11133 @var{iprefix} is the prefix for the gcc directory.
11136 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11137 This target hook should register special include paths for the target.
11138 The parameter @var{path} is the include to register. On Darwin
11139 systems, this is used for Framework includes, which have semantics
11140 that are different from @option{-I}.
11143 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11144 This target macro returns @code{true} if it is safe to use a local alias
11145 for a virtual function @var{fndecl} when constructing thunks,
11146 @code{false} otherwise. By default, the macro returns @code{true} for all
11147 functions, if a target supports aliases (i.e.@: defines
11148 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11151 @defmac TARGET_FORMAT_TYPES
11152 If defined, this macro is the name of a global variable containing
11153 target-specific format checking information for the @option{-Wformat}
11154 option. The default is to have no target-specific format checks.
11157 @defmac TARGET_N_FORMAT_TYPES
11158 If defined, this macro is the number of entries in
11159 @code{TARGET_FORMAT_TYPES}.
11162 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11163 If defined, this macro is the name of a global variable containing
11164 target-specific format overrides for the @option{-Wformat} option. The
11165 default is to have no target-specific format overrides. If defined,
11166 @code{TARGET_FORMAT_TYPES} must be defined, too.
11169 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11170 If defined, this macro specifies the number of entries in
11171 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11174 @defmac TARGET_OVERRIDES_FORMAT_INIT
11175 If defined, this macro specifies the optional initialization
11176 routine for target specific customizations of the system printf
11177 and scanf formatter settings.
11180 @deftypevr {Target Hook} bool TARGET_RELAXED_ORDERING
11181 If set to @code{true}, means that the target's memory model does not
11182 guarantee that loads which do not depend on one another will access
11183 main memory in the order of the instruction stream; if ordering is
11184 important, an explicit memory barrier must be used. This is true of
11185 many recent processors which implement a policy of ``relaxed,''
11186 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11187 and ia64. The default is @code{false}.
11190 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11191 If defined, this macro returns the diagnostic message when it is
11192 illegal to pass argument @var{val} to function @var{funcdecl}
11193 with prototype @var{typelist}.
11196 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11197 If defined, this macro returns the diagnostic message when it is
11198 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11199 if validity should be determined by the front end.
11202 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11203 If defined, this macro returns the diagnostic message when it is
11204 invalid to apply operation @var{op} (where unary plus is denoted by
11205 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11206 if validity should be determined by the front end.
11209 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11210 If defined, this macro returns the diagnostic message when it is
11211 invalid to apply operation @var{op} to operands of types @var{type1}
11212 and @var{type2}, or @code{NULL} if validity should be determined by
11216 @deftypefn {Target Hook} {const char *} TARGET_INVALID_PARAMETER_TYPE (const_tree @var{type})
11217 If defined, this macro returns the diagnostic message when it is
11218 invalid for functions to include parameters of type @var{type},
11219 or @code{NULL} if validity should be determined by
11220 the front end. This is currently used only by the C and C++ front ends.
11223 @deftypefn {Target Hook} {const char *} TARGET_INVALID_RETURN_TYPE (const_tree @var{type})
11224 If defined, this macro returns the diagnostic message when it is
11225 invalid for functions to have return type @var{type},
11226 or @code{NULL} if validity should be determined by
11227 the front end. This is currently used only by the C and C++ front ends.
11230 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11231 If defined, this target hook returns the type to which values of
11232 @var{type} should be promoted when they appear in expressions,
11233 analogous to the integer promotions, or @code{NULL_TREE} to use the
11234 front end's normal promotion rules. This hook is useful when there are
11235 target-specific types with special promotion rules.
11236 This is currently used only by the C and C++ front ends.
11239 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11240 If defined, this hook returns the result of converting @var{expr} to
11241 @var{type}. It should return the converted expression,
11242 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11243 This hook is useful when there are target-specific types with special
11245 This is currently used only by the C and C++ front ends.
11248 @defmac TARGET_USE_JCR_SECTION
11249 This macro determines whether to use the JCR section to register Java
11250 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11251 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11255 This macro determines the size of the objective C jump buffer for the
11256 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11259 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11260 Define this macro if any target-specific attributes need to be attached
11261 to the functions in @file{libgcc} that provide low-level support for
11262 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11263 and the associated definitions of those functions.
11266 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11267 Define this macro to update the current function stack boundary if
11271 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11272 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11273 different argument pointer register is needed to access the function's
11274 argument list due to stack realignment. Return @code{NULL} if no DRAP
11278 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11279 When optimization is disabled, this hook indicates whether or not
11280 arguments should be allocated to stack slots. Normally, GCC allocates
11281 stacks slots for arguments when not optimizing in order to make
11282 debugging easier. However, when a function is declared with
11283 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11284 cannot safely move arguments from the registers in which they are passed
11285 to the stack. Therefore, this hook should return true in general, but
11286 false for naked functions. The default implementation always returns true.
11289 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11290 On some architectures it can take multiple instructions to synthesize
11291 a constant. If there is another constant already in a register that
11292 is close enough in value then it is preferable that the new constant
11293 is computed from this register using immediate addition or
11294 subtraction. We accomplish this through CSE. Besides the value of
11295 the constant we also add a lower and an upper constant anchor to the
11296 available expressions. These are then queried when encountering new
11297 constants. The anchors are computed by rounding the constant up and
11298 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11299 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11300 accepted by immediate-add plus one. We currently assume that the
11301 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11302 MIPS, where add-immediate takes a 16-bit signed value,
11303 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11304 is zero, which disables this optimization. @end deftypevr