1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Target Description Macros and Functions
9 @cindex machine description macros
10 @cindex target description macros
11 @cindex macros, target description
12 @cindex @file{tm.h} macros
14 In addition to the file @file{@var{machine}.md}, a machine description
15 includes a C header file conventionally given the name
16 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
17 The header file defines numerous macros that convey the information
18 about the target machine that does not fit into the scheme of the
19 @file{.md} file. The file @file{tm.h} should be a link to
20 @file{@var{machine}.h}. The header file @file{config.h} includes
21 @file{tm.h} and most compiler source files include @file{config.h}. The
22 source file defines a variable @code{targetm}, which is a structure
23 containing pointers to functions and data relating to the target
24 machine. @file{@var{machine}.c} should also contain their definitions,
25 if they are not defined elsewhere in GCC, and other functions called
26 through the macros defined in the @file{.h} file.
29 * Target Structure:: The @code{targetm} variable.
30 * Driver:: Controlling how the driver runs the compilation passes.
31 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
32 * Per-Function Data:: Defining data structures for per-function information.
33 * Storage Layout:: Defining sizes and alignments of data.
34 * Type Layout:: Defining sizes and properties of basic user data types.
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
37 * Old Constraints:: The old way to define machine-specific constraints.
38 * Stack and Calling:: Defining which way the stack grows and by how much.
39 * Varargs:: Defining the varargs macros.
40 * Trampolines:: Code set up at run time to enter a nested function.
41 * Library Calls:: Controlling how library routines are implicitly called.
42 * Addressing Modes:: Defining addressing modes valid for memory operands.
43 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
44 * Condition Code:: Defining how insns update the condition code.
45 * Costs:: Defining relative costs of different operations.
46 * Scheduling:: Adjusting the behavior of the instruction scheduler.
47 * Sections:: Dividing storage into text, data, and other sections.
48 * PIC:: Macros for position independent code.
49 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
50 * Debugging Info:: Defining the format of debugging output.
51 * Floating Point:: Handling floating point for cross-compilers.
52 * Mode Switching:: Insertion of mode-switching instructions.
53 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
54 * Emulated TLS:: Emulated TLS support.
55 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
56 * PCH Target:: Validity checking for precompiled headers.
57 * C++ ABI:: Controlling C++ ABI changes.
58 * Named Address Spaces:: Adding support for named address spaces
59 * Misc:: Everything else.
62 @node Target Structure
63 @section The Global @code{targetm} Variable
65 @cindex target functions
67 @deftypevar {struct gcc_target} targetm
68 The target @file{.c} file must define the global @code{targetm} variable
69 which contains pointers to functions and data relating to the target
70 machine. The variable is declared in @file{target.h};
71 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
72 used to initialize the variable, and macros for the default initializers
73 for elements of the structure. The @file{.c} file should override those
74 macros for which the default definition is inappropriate. For example:
77 #include "target-def.h"
79 /* @r{Initialize the GCC target structure.} */
81 #undef TARGET_COMP_TYPE_ATTRIBUTES
82 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
84 struct gcc_target targetm = TARGET_INITIALIZER;
88 Where a macro should be defined in the @file{.c} file in this manner to
89 form part of the @code{targetm} structure, it is documented below as a
90 ``Target Hook'' with a prototype. Many macros will change in future
91 from being defined in the @file{.h} file to being part of the
92 @code{targetm} structure.
94 Similarly, there is a @code{targetcm} variable for hooks that are
95 specific to front ends for C-family languages, documented as ``C
96 Target Hook''. This is declared in @file{c-family/c-target.h}, the
97 initializer @code{TARGETCM_INITIALIZER} in
98 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
99 themselves, they should set @code{target_has_targetcm=yes} in
100 @file{config.gcc}; otherwise a default definition is used.
102 Similarly, there is a @code{targetm_common} variable for hooks that
103 are shared between the compiler driver and the compilers proper,
104 documented as ``Common Target Hook''. This is declared in
105 @file{common/common-target.h}, the initializer
106 @code{TARGETM_COMMON_INITIALIZER} in
107 @file{common/common-target-def.h}. If targets initialize
108 @code{targetm_common} themselves, they should set
109 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
110 default definition is used.
113 @section Controlling the Compilation Driver, @file{gcc}
115 @cindex controlling the compilation driver
117 @c prevent bad page break with this line
118 You can control the compilation driver.
120 @defmac DRIVER_SELF_SPECS
121 A list of specs for the driver itself. It should be a suitable
122 initializer for an array of strings, with no surrounding braces.
124 The driver applies these specs to its own command line between loading
125 default @file{specs} files (but not command-line specified ones) and
126 choosing the multilib directory or running any subcommands. It
127 applies them in the order given, so each spec can depend on the
128 options added by earlier ones. It is also possible to remove options
129 using @samp{%<@var{option}} in the usual way.
131 This macro can be useful when a port has several interdependent target
132 options. It provides a way of standardizing the command line so
133 that the other specs are easier to write.
135 Do not define this macro if it does not need to do anything.
138 @defmac OPTION_DEFAULT_SPECS
139 A list of specs used to support configure-time default options (i.e.@:
140 @option{--with} options) in the driver. It should be a suitable initializer
141 for an array of structures, each containing two strings, without the
142 outermost pair of surrounding braces.
144 The first item in the pair is the name of the default. This must match
145 the code in @file{config.gcc} for the target. The second item is a spec
146 to apply if a default with this name was specified. The string
147 @samp{%(VALUE)} in the spec will be replaced by the value of the default
148 everywhere it occurs.
150 The driver will apply these specs to its own command line between loading
151 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
152 the same mechanism as @code{DRIVER_SELF_SPECS}.
154 Do not define this macro if it does not need to do anything.
158 A C string constant that tells the GCC driver program options to
159 pass to CPP@. It can also specify how to translate options you
160 give to GCC into options for GCC to pass to the CPP@.
162 Do not define this macro if it does not need to do anything.
165 @defmac CPLUSPLUS_CPP_SPEC
166 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
167 than C@. If you do not define this macro, then the value of
168 @code{CPP_SPEC} (if any) will be used instead.
172 A C string constant that tells the GCC driver program options to
173 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
175 It can also specify how to translate options you give to GCC into options
176 for GCC to pass to front ends.
178 Do not define this macro if it does not need to do anything.
182 A C string constant that tells the GCC driver program options to
183 pass to @code{cc1plus}. It can also specify how to translate options you
184 give to GCC into options for GCC to pass to the @code{cc1plus}.
186 Do not define this macro if it does not need to do anything.
187 Note that everything defined in CC1_SPEC is already passed to
188 @code{cc1plus} so there is no need to duplicate the contents of
189 CC1_SPEC in CC1PLUS_SPEC@.
193 A C string constant that tells the GCC driver program options to
194 pass to the assembler. It can also specify how to translate options
195 you give to GCC into options for GCC to pass to the assembler.
196 See the file @file{sun3.h} for an example of this.
198 Do not define this macro if it does not need to do anything.
201 @defmac ASM_FINAL_SPEC
202 A C string constant that tells the GCC driver program how to
203 run any programs which cleanup after the normal assembler.
204 Normally, this is not needed. See the file @file{mips.h} for
207 Do not define this macro if it does not need to do anything.
210 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
211 Define this macro, with no value, if the driver should give the assembler
212 an argument consisting of a single dash, @option{-}, to instruct it to
213 read from its standard input (which will be a pipe connected to the
214 output of the compiler proper). This argument is given after any
215 @option{-o} option specifying the name of the output file.
217 If you do not define this macro, the assembler is assumed to read its
218 standard input if given no non-option arguments. If your assembler
219 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
220 see @file{mips.h} for instance.
224 A C string constant that tells the GCC driver program options to
225 pass to the linker. It can also specify how to translate options you
226 give to GCC into options for GCC to pass to the linker.
228 Do not define this macro if it does not need to do anything.
232 Another C string constant used much like @code{LINK_SPEC}. The difference
233 between the two is that @code{LIB_SPEC} is used at the end of the
234 command given to the linker.
236 If this macro is not defined, a default is provided that
237 loads the standard C library from the usual place. See @file{gcc.c}.
241 Another C string constant that tells the GCC driver program
242 how and when to place a reference to @file{libgcc.a} into the
243 linker command line. This constant is placed both before and after
244 the value of @code{LIB_SPEC}.
246 If this macro is not defined, the GCC driver provides a default that
247 passes the string @option{-lgcc} to the linker.
250 @defmac REAL_LIBGCC_SPEC
251 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
252 @code{LIBGCC_SPEC} is not directly used by the driver program but is
253 instead modified to refer to different versions of @file{libgcc.a}
254 depending on the values of the command line flags @option{-static},
255 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
256 targets where these modifications are inappropriate, define
257 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
258 driver how to place a reference to @file{libgcc} on the link command
259 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
262 @defmac USE_LD_AS_NEEDED
263 A macro that controls the modifications to @code{LIBGCC_SPEC}
264 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
265 generated that uses --as-needed and the shared libgcc in place of the
266 static exception handler library, when linking without any of
267 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
271 If defined, this C string constant is added to @code{LINK_SPEC}.
272 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
273 the modifications to @code{LIBGCC_SPEC} mentioned in
274 @code{REAL_LIBGCC_SPEC}.
277 @defmac STARTFILE_SPEC
278 Another C string constant used much like @code{LINK_SPEC}. The
279 difference between the two is that @code{STARTFILE_SPEC} is used at
280 the very beginning of the command given to the linker.
282 If this macro is not defined, a default is provided that loads the
283 standard C startup file from the usual place. See @file{gcc.c}.
287 Another C string constant used much like @code{LINK_SPEC}. The
288 difference between the two is that @code{ENDFILE_SPEC} is used at
289 the very end of the command given to the linker.
291 Do not define this macro if it does not need to do anything.
294 @defmac THREAD_MODEL_SPEC
295 GCC @code{-v} will print the thread model GCC was configured to use.
296 However, this doesn't work on platforms that are multilibbed on thread
297 models, such as AIX 4.3. On such platforms, define
298 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
299 blanks that names one of the recognized thread models. @code{%*}, the
300 default value of this macro, will expand to the value of
301 @code{thread_file} set in @file{config.gcc}.
304 @defmac SYSROOT_SUFFIX_SPEC
305 Define this macro to add a suffix to the target sysroot when GCC is
306 configured with a sysroot. This will cause GCC to search for usr/lib,
307 et al, within sysroot+suffix.
310 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
311 Define this macro to add a headers_suffix to the target sysroot when
312 GCC is configured with a sysroot. This will cause GCC to pass the
313 updated sysroot+headers_suffix to CPP, causing it to search for
314 usr/include, et al, within sysroot+headers_suffix.
318 Define this macro to provide additional specifications to put in the
319 @file{specs} file that can be used in various specifications like
322 The definition should be an initializer for an array of structures,
323 containing a string constant, that defines the specification name, and a
324 string constant that provides the specification.
326 Do not define this macro if it does not need to do anything.
328 @code{EXTRA_SPECS} is useful when an architecture contains several
329 related targets, which have various @code{@dots{}_SPECS} which are similar
330 to each other, and the maintainer would like one central place to keep
333 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
334 define either @code{_CALL_SYSV} when the System V calling sequence is
335 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
338 The @file{config/rs6000/rs6000.h} target file defines:
341 #define EXTRA_SPECS \
342 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
344 #define CPP_SYS_DEFAULT ""
347 The @file{config/rs6000/sysv.h} target file defines:
351 "%@{posix: -D_POSIX_SOURCE @} \
352 %@{mcall-sysv: -D_CALL_SYSV @} \
353 %@{!mcall-sysv: %(cpp_sysv_default) @} \
354 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
356 #undef CPP_SYSV_DEFAULT
357 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
360 while the @file{config/rs6000/eabiaix.h} target file defines
361 @code{CPP_SYSV_DEFAULT} as:
364 #undef CPP_SYSV_DEFAULT
365 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
369 @defmac LINK_LIBGCC_SPECIAL_1
370 Define this macro if the driver program should find the library
371 @file{libgcc.a}. If you do not define this macro, the driver program will pass
372 the argument @option{-lgcc} to tell the linker to do the search.
375 @defmac LINK_GCC_C_SEQUENCE_SPEC
376 The sequence in which libgcc and libc are specified to the linker.
377 By default this is @code{%G %L %G}.
380 @defmac LINK_COMMAND_SPEC
381 A C string constant giving the complete command line need to execute the
382 linker. When you do this, you will need to update your port each time a
383 change is made to the link command line within @file{gcc.c}. Therefore,
384 define this macro only if you need to completely redefine the command
385 line for invoking the linker and there is no other way to accomplish
386 the effect you need. Overriding this macro may be avoidable by overriding
387 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
390 @hook TARGET_ALWAYS_STRIP_DOTDOT
392 @defmac MULTILIB_DEFAULTS
393 Define this macro as a C expression for the initializer of an array of
394 string to tell the driver program which options are defaults for this
395 target and thus do not need to be handled specially when using
396 @code{MULTILIB_OPTIONS}.
398 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
399 the target makefile fragment or if none of the options listed in
400 @code{MULTILIB_OPTIONS} are set by default.
401 @xref{Target Fragment}.
404 @defmac RELATIVE_PREFIX_NOT_LINKDIR
405 Define this macro to tell @command{gcc} that it should only translate
406 a @option{-B} prefix into a @option{-L} linker option if the prefix
407 indicates an absolute file name.
410 @defmac MD_EXEC_PREFIX
411 If defined, this macro is an additional prefix to try after
412 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
413 when the compiler is built as a cross
414 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
415 to the list of directories used to find the assembler in @file{configure.in}.
418 @defmac STANDARD_STARTFILE_PREFIX
419 Define this macro as a C string constant if you wish to override the
420 standard choice of @code{libdir} as the default prefix to
421 try when searching for startup files such as @file{crt0.o}.
422 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
423 is built as a cross compiler.
426 @defmac STANDARD_STARTFILE_PREFIX_1
427 Define this macro as a C string constant if you wish to override the
428 standard choice of @code{/lib} as a prefix to try after the default prefix
429 when searching for startup files such as @file{crt0.o}.
430 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
431 is built as a cross compiler.
434 @defmac STANDARD_STARTFILE_PREFIX_2
435 Define this macro as a C string constant if you wish to override the
436 standard choice of @code{/lib} as yet another prefix to try after the
437 default prefix when searching for startup files such as @file{crt0.o}.
438 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
439 is built as a cross compiler.
442 @defmac MD_STARTFILE_PREFIX
443 If defined, this macro supplies an additional prefix to try after the
444 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
445 compiler is built as a cross compiler.
448 @defmac MD_STARTFILE_PREFIX_1
449 If defined, this macro supplies yet another prefix to try after the
450 standard prefixes. It is not searched when the compiler is built as a
454 @defmac INIT_ENVIRONMENT
455 Define this macro as a C string constant if you wish to set environment
456 variables for programs called by the driver, such as the assembler and
457 loader. The driver passes the value of this macro to @code{putenv} to
458 initialize the necessary environment variables.
461 @defmac LOCAL_INCLUDE_DIR
462 Define this macro as a C string constant if you wish to override the
463 standard choice of @file{/usr/local/include} as the default prefix to
464 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
465 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
466 @file{config.gcc}, normally @file{/usr/include}) in the search order.
468 Cross compilers do not search either @file{/usr/local/include} or its
472 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
473 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
474 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
475 If you do not define this macro, no component is used.
478 @defmac INCLUDE_DEFAULTS
479 Define this macro if you wish to override the entire default search path
480 for include files. For a native compiler, the default search path
481 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
482 @code{GPLUSPLUS_INCLUDE_DIR}, and
483 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
484 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
485 and specify private search areas for GCC@. The directory
486 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
488 The definition should be an initializer for an array of structures.
489 Each array element should have four elements: the directory name (a
490 string constant), the component name (also a string constant), a flag
491 for C++-only directories,
492 and a flag showing that the includes in the directory don't need to be
493 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
494 the array with a null element.
496 The component name denotes what GNU package the include file is part of,
497 if any, in all uppercase letters. For example, it might be @samp{GCC}
498 or @samp{BINUTILS}. If the package is part of a vendor-supplied
499 operating system, code the component name as @samp{0}.
501 For example, here is the definition used for VAX/VMS:
504 #define INCLUDE_DEFAULTS \
506 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
507 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
508 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
515 Here is the order of prefixes tried for exec files:
519 Any prefixes specified by the user with @option{-B}.
522 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
523 is not set and the compiler has not been installed in the configure-time
524 @var{prefix}, the location in which the compiler has actually been installed.
527 The directories specified by the environment variable @code{COMPILER_PATH}.
530 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
531 in the configured-time @var{prefix}.
534 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
537 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
540 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
544 Here is the order of prefixes tried for startfiles:
548 Any prefixes specified by the user with @option{-B}.
551 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
552 value based on the installed toolchain location.
555 The directories specified by the environment variable @code{LIBRARY_PATH}
556 (or port-specific name; native only, cross compilers do not use this).
559 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
560 in the configured @var{prefix} or 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 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
571 native compiler, or we have a target system root.
574 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
575 native compiler, or we have a target system root.
578 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
579 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
580 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
583 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
584 compiler, or we have a target system root. The default for this macro is
588 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
589 compiler, or we have a target system root. The default for this macro is
593 @node Run-time Target
594 @section Run-time Target Specification
595 @cindex run-time target specification
596 @cindex predefined macros
597 @cindex target specifications
599 @c prevent bad page break with this line
600 Here are run-time target specifications.
602 @defmac TARGET_CPU_CPP_BUILTINS ()
603 This function-like macro expands to a block of code that defines
604 built-in preprocessor macros and assertions for the target CPU, using
605 the functions @code{builtin_define}, @code{builtin_define_std} and
606 @code{builtin_assert}. When the front end
607 calls this macro it provides a trailing semicolon, and since it has
608 finished command line option processing your code can use those
611 @code{builtin_assert} takes a string in the form you pass to the
612 command-line option @option{-A}, such as @code{cpu=mips}, and creates
613 the assertion. @code{builtin_define} takes a string in the form
614 accepted by option @option{-D} and unconditionally defines the macro.
616 @code{builtin_define_std} takes a string representing the name of an
617 object-like macro. If it doesn't lie in the user's namespace,
618 @code{builtin_define_std} defines it unconditionally. Otherwise, it
619 defines a version with two leading underscores, and another version
620 with two leading and trailing underscores, and defines the original
621 only if an ISO standard was not requested on the command line. For
622 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
623 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
624 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
625 defines only @code{_ABI64}.
627 You can also test for the C dialect being compiled. The variable
628 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
629 or @code{clk_objective_c}. Note that if we are preprocessing
630 assembler, this variable will be @code{clk_c} but the function-like
631 macro @code{preprocessing_asm_p()} will return true, so you might want
632 to check for that first. If you need to check for strict ANSI, the
633 variable @code{flag_iso} can be used. The function-like macro
634 @code{preprocessing_trad_p()} can be used to check for traditional
638 @defmac TARGET_OS_CPP_BUILTINS ()
639 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
640 and is used for the target operating system instead.
643 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
644 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
645 and is used for the target object format. @file{elfos.h} uses this
646 macro to define @code{__ELF__}, so you probably do not need to define
650 @deftypevar {extern int} target_flags
651 This variable is declared in @file{options.h}, which is included before
652 any target-specific headers.
655 @hook TARGET_DEFAULT_TARGET_FLAGS
656 This variable specifies the initial value of @code{target_flags}.
657 Its default setting is 0.
660 @cindex optional hardware or system features
661 @cindex features, optional, in system conventions
663 @hook TARGET_HANDLE_OPTION
664 This hook is called whenever the user specifies one of the
665 target-specific options described by the @file{.opt} definition files
666 (@pxref{Options}). It has the opportunity to do some option-specific
667 processing and should return true if the option is valid. The default
668 definition does nothing but return true.
670 @var{decoded} specifies the option and its arguments. @var{opts} and
671 @var{opts_set} are the @code{gcc_options} structures to be used for
672 storing option state, and @var{loc} is the location at which the
673 option was passed (@code{UNKNOWN_LOCATION} except for options passed
677 @hook TARGET_HANDLE_C_OPTION
678 This target hook is called whenever the user specifies one of the
679 target-specific C language family options described by the @file{.opt}
680 definition files(@pxref{Options}). It has the opportunity to do some
681 option-specific processing and should return true if the option is
682 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
683 default definition does nothing but return false.
685 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
686 options. However, if processing an option requires routines that are
687 only available in the C (and related language) front ends, then you
688 should use @code{TARGET_HANDLE_C_OPTION} instead.
691 @hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT
693 @hook TARGET_STRING_OBJECT_REF_TYPE_P
695 @hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG
697 @hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE
698 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
699 but is called when the optimize level is changed via an attribute or
700 pragma or when it is reset at the end of the code affected by the
701 attribute or pragma. It is not called at the beginning of compilation
702 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
703 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
704 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
707 @defmac C_COMMON_OVERRIDE_OPTIONS
708 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
709 but is only used in the C
710 language frontends (C, Objective-C, C++, Objective-C++) and so can be
711 used to alter option flag variables which only exist in those
715 @hook TARGET_OPTION_OPTIMIZATION_TABLE
716 Some machines may desire to change what optimizations are performed for
717 various optimization levels. This variable, if defined, describes
718 options to enable at particular sets of optimization levels. These
719 options are processed once
720 just after the optimization level is determined and before the remainder
721 of the command options have been parsed, so may be overridden by other
722 options passed explicitly.
724 This processing is run once at program startup and when the optimization
725 options are changed via @code{#pragma GCC optimize} or by using the
726 @code{optimize} attribute.
729 @hook TARGET_OPTION_INIT_STRUCT
731 @hook TARGET_OPTION_DEFAULT_PARAMS
733 @defmac SWITCHABLE_TARGET
734 Some targets need to switch between substantially different subtargets
735 during compilation. For example, the MIPS target has one subtarget for
736 the traditional MIPS architecture and another for MIPS16. Source code
737 can switch between these two subarchitectures using the @code{mips16}
738 and @code{nomips16} attributes.
740 Such subtargets can differ in things like the set of available
741 registers, the set of available instructions, the costs of various
742 operations, and so on. GCC caches a lot of this type of information
743 in global variables, and recomputing them for each subtarget takes a
744 significant amount of time. The compiler therefore provides a facility
745 for maintaining several versions of the global variables and quickly
746 switching between them; see @file{target-globals.h} for details.
748 Define this macro to 1 if your target needs this facility. The default
752 @node Per-Function Data
753 @section Defining data structures for per-function information.
754 @cindex per-function data
755 @cindex data structures
757 If the target needs to store information on a per-function basis, GCC
758 provides a macro and a couple of variables to allow this. Note, just
759 using statics to store the information is a bad idea, since GCC supports
760 nested functions, so you can be halfway through encoding one function
761 when another one comes along.
763 GCC defines a data structure called @code{struct function} which
764 contains all of the data specific to an individual function. This
765 structure contains a field called @code{machine} whose type is
766 @code{struct machine_function *}, which can be used by targets to point
767 to their own specific data.
769 If a target needs per-function specific data it should define the type
770 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
771 This macro should be used to initialize the function pointer
772 @code{init_machine_status}. This pointer is explained below.
774 One typical use of per-function, target specific data is to create an
775 RTX to hold the register containing the function's return address. This
776 RTX can then be used to implement the @code{__builtin_return_address}
777 function, for level 0.
779 Note---earlier implementations of GCC used a single data area to hold
780 all of the per-function information. Thus when processing of a nested
781 function began the old per-function data had to be pushed onto a
782 stack, and when the processing was finished, it had to be popped off the
783 stack. GCC used to provide function pointers called
784 @code{save_machine_status} and @code{restore_machine_status} to handle
785 the saving and restoring of the target specific information. Since the
786 single data area approach is no longer used, these pointers are no
789 @defmac INIT_EXPANDERS
790 Macro called to initialize any target specific information. This macro
791 is called once per function, before generation of any RTL has begun.
792 The intention of this macro is to allow the initialization of the
793 function pointer @code{init_machine_status}.
796 @deftypevar {void (*)(struct function *)} init_machine_status
797 If this function pointer is non-@code{NULL} it will be called once per
798 function, before function compilation starts, in order to allow the
799 target to perform any target specific initialization of the
800 @code{struct function} structure. It is intended that this would be
801 used to initialize the @code{machine} of that structure.
803 @code{struct machine_function} structures are expected to be freed by GC@.
804 Generally, any memory that they reference must be allocated by using
805 GC allocation, including the structure itself.
809 @section Storage Layout
810 @cindex storage layout
812 Note that the definitions of the macros in this table which are sizes or
813 alignments measured in bits do not need to be constant. They can be C
814 expressions that refer to static variables, such as the @code{target_flags}.
815 @xref{Run-time Target}.
817 @defmac BITS_BIG_ENDIAN
818 Define this macro to have the value 1 if the most significant bit in a
819 byte has the lowest number; otherwise define it to have the value zero.
820 This means that bit-field instructions count from the most significant
821 bit. If the machine has no bit-field instructions, then this must still
822 be defined, but it doesn't matter which value it is defined to. This
823 macro need not be a constant.
825 This macro does not affect the way structure fields are packed into
826 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
829 @defmac BYTES_BIG_ENDIAN
830 Define this macro to have the value 1 if the most significant byte in a
831 word has the lowest number. This macro need not be a constant.
834 @defmac WORDS_BIG_ENDIAN
835 Define this macro to have the value 1 if, in a multiword object, the
836 most significant word has the lowest number. This applies to both
837 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
838 order of words in memory is not the same as the order in registers. This
839 macro need not be a constant.
842 @defmac REG_WORDS_BIG_ENDIAN
843 On some machines, the order of words in a multiword object differs between
844 registers in memory. In such a situation, define this macro to describe
845 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
846 the order of words in memory.
849 @defmac FLOAT_WORDS_BIG_ENDIAN
850 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
851 @code{TFmode} floating point numbers are stored in memory with the word
852 containing the sign bit at the lowest address; otherwise define it to
853 have the value 0. This macro need not be a constant.
855 You need not define this macro if the ordering is the same as for
859 @defmac BITS_PER_UNIT
860 Define this macro to be the number of bits in an addressable storage
861 unit (byte). If you do not define this macro the default is 8.
864 @defmac BITS_PER_WORD
865 Number of bits in a word. If you do not define this macro, the default
866 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
869 @defmac MAX_BITS_PER_WORD
870 Maximum number of bits in a word. If this is undefined, the default is
871 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
872 largest value that @code{BITS_PER_WORD} can have at run-time.
875 @defmac UNITS_PER_WORD
876 Number of storage units in a word; normally the size of a general-purpose
877 register, a power of two from 1 or 8.
880 @defmac MIN_UNITS_PER_WORD
881 Minimum number of units in a word. If this is undefined, the default is
882 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
883 smallest value that @code{UNITS_PER_WORD} can have at run-time.
887 Width of a pointer, in bits. You must specify a value no wider than the
888 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
889 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
890 a value the default is @code{BITS_PER_WORD}.
893 @defmac POINTERS_EXTEND_UNSIGNED
894 A C expression that determines how pointers should be extended from
895 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
896 greater than zero if pointers should be zero-extended, zero if they
897 should be sign-extended, and negative if some other sort of conversion
898 is needed. In the last case, the extension is done by the target's
899 @code{ptr_extend} instruction.
901 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
902 and @code{word_mode} are all the same width.
905 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
906 A macro to update @var{m} and @var{unsignedp} when an object whose type
907 is @var{type} and which has the specified mode and signedness is to be
908 stored in a register. This macro is only called when @var{type} is a
911 On most RISC machines, which only have operations that operate on a full
912 register, define this macro to set @var{m} to @code{word_mode} if
913 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
914 cases, only integer modes should be widened because wider-precision
915 floating-point operations are usually more expensive than their narrower
918 For most machines, the macro definition does not change @var{unsignedp}.
919 However, some machines, have instructions that preferentially handle
920 either signed or unsigned quantities of certain modes. For example, on
921 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
922 sign-extend the result to 64 bits. On such machines, set
923 @var{unsignedp} according to which kind of extension is more efficient.
925 Do not define this macro if it would never modify @var{m}.
928 @hook TARGET_PROMOTE_FUNCTION_MODE
929 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
930 function return values. The target hook should return the new mode
931 and possibly change @code{*@var{punsignedp}} if the promotion should
932 change signedness. This function is called only for scalar @emph{or
935 @var{for_return} allows to distinguish the promotion of arguments and
936 return values. If it is @code{1}, a return value is being promoted and
937 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
938 If it is @code{2}, the returned mode should be that of the register in
939 which an incoming parameter is copied, or the outgoing result is computed;
940 then the hook should return the same mode as @code{promote_mode}, though
941 the signedness may be different.
943 @var{type} can be NULL when promoting function arguments of libcalls.
945 The default is to not promote arguments and return values. You can
946 also define the hook to @code{default_promote_function_mode_always_promote}
947 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
950 @defmac PARM_BOUNDARY
951 Normal alignment required for function parameters on the stack, in
952 bits. All stack parameters receive at least this much alignment
953 regardless of data type. On most machines, this is the same as the
957 @defmac STACK_BOUNDARY
958 Define this macro to the minimum alignment enforced by hardware for the
959 stack pointer on this machine. The definition is a C expression for the
960 desired alignment (measured in bits). This value is used as a default
961 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
962 this should be the same as @code{PARM_BOUNDARY}.
965 @defmac PREFERRED_STACK_BOUNDARY
966 Define this macro if you wish to preserve a certain alignment for the
967 stack pointer, greater than what the hardware enforces. The definition
968 is a C expression for the desired alignment (measured in bits). This
969 macro must evaluate to a value equal to or larger than
970 @code{STACK_BOUNDARY}.
973 @defmac INCOMING_STACK_BOUNDARY
974 Define this macro if the incoming stack boundary may be different
975 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
976 to a value equal to or larger than @code{STACK_BOUNDARY}.
979 @defmac FUNCTION_BOUNDARY
980 Alignment required for a function entry point, in bits.
983 @defmac BIGGEST_ALIGNMENT
984 Biggest alignment that any data type can require on this machine, in
985 bits. Note that this is not the biggest alignment that is supported,
986 just the biggest alignment that, when violated, may cause a fault.
989 @defmac MALLOC_ABI_ALIGNMENT
990 Alignment, in bits, a C conformant malloc implementation has to
991 provide. If not defined, the default value is @code{BITS_PER_WORD}.
994 @defmac ATTRIBUTE_ALIGNED_VALUE
995 Alignment used by the @code{__attribute__ ((aligned))} construct. If
996 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
999 @defmac MINIMUM_ATOMIC_ALIGNMENT
1000 If defined, the smallest alignment, in bits, that can be given to an
1001 object that can be referenced in one operation, without disturbing any
1002 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1003 on machines that don't have byte or half-word store operations.
1006 @defmac BIGGEST_FIELD_ALIGNMENT
1007 Biggest alignment that any structure or union field can require on this
1008 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1009 structure and union fields only, unless the field alignment has been set
1010 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1013 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1014 An expression for the alignment of a structure field @var{field} if the
1015 alignment computed in the usual way (including applying of
1016 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1017 alignment) is @var{computed}. It overrides alignment only if the
1018 field alignment has not been set by the
1019 @code{__attribute__ ((aligned (@var{n})))} construct.
1022 @defmac MAX_STACK_ALIGNMENT
1023 Biggest stack alignment guaranteed by the backend. Use this macro
1024 to specify the maximum alignment of a variable on stack.
1026 If not defined, the default value is @code{STACK_BOUNDARY}.
1028 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1029 @c But the fix for PR 32893 indicates that we can only guarantee
1030 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1031 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1034 @defmac MAX_OFILE_ALIGNMENT
1035 Biggest alignment supported by the object file format of this machine.
1036 Use this macro to limit the alignment which can be specified using the
1037 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1038 the default value is @code{BIGGEST_ALIGNMENT}.
1040 On systems that use ELF, the default (in @file{config/elfos.h}) is
1041 the largest supported 32-bit ELF section alignment representable on
1042 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1043 On 32-bit ELF the largest supported section alignment in bits is
1044 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1047 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1048 If defined, a C expression to compute the alignment for a variable in
1049 the static store. @var{type} is the data type, and @var{basic-align} is
1050 the alignment that the object would ordinarily have. The value of this
1051 macro is used instead of that alignment to align the object.
1053 If this macro is not defined, then @var{basic-align} is used.
1056 One use of this macro is to increase alignment of medium-size data to
1057 make it all fit in fewer cache lines. Another is to cause character
1058 arrays to be word-aligned so that @code{strcpy} calls that copy
1059 constants to character arrays can be done inline.
1062 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1063 If defined, a C expression to compute the alignment given to a constant
1064 that is being placed in memory. @var{constant} is the constant and
1065 @var{basic-align} is the alignment that the object would ordinarily
1066 have. The value of this macro is used instead of that alignment to
1069 If this macro is not defined, then @var{basic-align} is used.
1071 The typical use of this macro is to increase alignment for string
1072 constants to be word aligned so that @code{strcpy} calls that copy
1073 constants can be done inline.
1076 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1077 If defined, a C expression to compute the alignment for a variable in
1078 the local store. @var{type} is the data type, and @var{basic-align} is
1079 the alignment that the object would ordinarily have. The value of this
1080 macro is used instead of that alignment to align the object.
1082 If this macro is not defined, then @var{basic-align} is used.
1084 One use of this macro is to increase alignment of medium-size data to
1085 make it all fit in fewer cache lines.
1087 If the value of this macro has a type, it should be an unsigned type.
1090 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1091 If defined, a C expression to compute the alignment for stack slot.
1092 @var{type} is the data type, @var{mode} is the widest mode available,
1093 and @var{basic-align} is the alignment that the slot would ordinarily
1094 have. The value of this macro is used instead of that alignment to
1097 If this macro is not defined, then @var{basic-align} is used when
1098 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1101 This macro is to set alignment of stack slot to the maximum alignment
1102 of all possible modes which the slot may have.
1104 If the value of this macro has a type, it should be an unsigned type.
1107 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1108 If defined, a C expression to compute the alignment for a local
1109 variable @var{decl}.
1111 If this macro is not defined, then
1112 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1115 One use of this macro is to increase alignment of medium-size data to
1116 make it all fit in fewer cache lines.
1118 If the value of this macro has a type, it should be an unsigned type.
1121 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1122 If defined, a C expression to compute the minimum required alignment
1123 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1124 @var{mode}, assuming normal alignment @var{align}.
1126 If this macro is not defined, then @var{align} will be used.
1129 @defmac EMPTY_FIELD_BOUNDARY
1130 Alignment in bits to be given to a structure bit-field that follows an
1131 empty field such as @code{int : 0;}.
1133 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1136 @defmac STRUCTURE_SIZE_BOUNDARY
1137 Number of bits which any structure or union's size must be a multiple of.
1138 Each structure or union's size is rounded up to a multiple of this.
1140 If you do not define this macro, the default is the same as
1141 @code{BITS_PER_UNIT}.
1144 @defmac STRICT_ALIGNMENT
1145 Define this macro to be the value 1 if instructions will fail to work
1146 if given data not on the nominal alignment. If instructions will merely
1147 go slower in that case, define this macro as 0.
1150 @defmac PCC_BITFIELD_TYPE_MATTERS
1151 Define this if you wish to imitate the way many other C compilers handle
1152 alignment of bit-fields and the structures that contain them.
1154 The behavior is that the type written for a named bit-field (@code{int},
1155 @code{short}, or other integer type) imposes an alignment for the entire
1156 structure, as if the structure really did contain an ordinary field of
1157 that type. In addition, the bit-field is placed within the structure so
1158 that it would fit within such a field, not crossing a boundary for it.
1160 Thus, on most machines, a named bit-field whose type is written as
1161 @code{int} would not cross a four-byte boundary, and would force
1162 four-byte alignment for the whole structure. (The alignment used may
1163 not be four bytes; it is controlled by the other alignment parameters.)
1165 An unnamed bit-field will not affect the alignment of the containing
1168 If the macro is defined, its definition should be a C expression;
1169 a nonzero value for the expression enables this behavior.
1171 Note that if this macro is not defined, or its value is zero, some
1172 bit-fields may cross more than one alignment boundary. The compiler can
1173 support such references if there are @samp{insv}, @samp{extv}, and
1174 @samp{extzv} insns that can directly reference memory.
1176 The other known way of making bit-fields work is to define
1177 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1178 Then every structure can be accessed with fullwords.
1180 Unless the machine has bit-field instructions or you define
1181 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1182 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1184 If your aim is to make GCC use the same conventions for laying out
1185 bit-fields as are used by another compiler, here is how to investigate
1186 what the other compiler does. Compile and run this program:
1205 printf ("Size of foo1 is %d\n",
1206 sizeof (struct foo1));
1207 printf ("Size of foo2 is %d\n",
1208 sizeof (struct foo2));
1213 If this prints 2 and 5, then the compiler's behavior is what you would
1214 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1217 @defmac BITFIELD_NBYTES_LIMITED
1218 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1219 to aligning a bit-field within the structure.
1222 @hook TARGET_ALIGN_ANON_BITFIELD
1223 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1224 whether unnamed bitfields affect the alignment of the containing
1225 structure. The hook should return true if the structure should inherit
1226 the alignment requirements of an unnamed bitfield's type.
1229 @hook TARGET_NARROW_VOLATILE_BITFIELD
1230 This target hook should return @code{true} if accesses to volatile bitfields
1231 should use the narrowest mode possible. It should return @code{false} if
1232 these accesses should use the bitfield container type.
1234 The default is @code{!TARGET_STRICT_ALIGN}.
1237 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1238 Return 1 if a structure or array containing @var{field} should be accessed using
1241 If @var{field} is the only field in the structure, @var{mode} is its
1242 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1243 case where structures of one field would require the structure's mode to
1244 retain the field's mode.
1246 Normally, this is not needed.
1249 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1250 Define this macro as an expression for the alignment of a type (given
1251 by @var{type} as a tree node) if the alignment computed in the usual
1252 way is @var{computed} and the alignment explicitly specified was
1255 The default is to use @var{specified} if it is larger; otherwise, use
1256 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1259 @defmac MAX_FIXED_MODE_SIZE
1260 An integer expression for the size in bits of the largest integer
1261 machine mode that should actually be used. All integer machine modes of
1262 this size or smaller can be used for structures and unions with the
1263 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1264 (DImode)} is assumed.
1267 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1268 If defined, an expression of type @code{enum machine_mode} that
1269 specifies the mode of the save area operand of a
1270 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1271 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1272 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1273 having its mode specified.
1275 You need not define this macro if it always returns @code{Pmode}. You
1276 would most commonly define this macro if the
1277 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1281 @defmac STACK_SIZE_MODE
1282 If defined, an expression of type @code{enum machine_mode} that
1283 specifies the mode of the size increment operand of an
1284 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1286 You need not define this macro if it always returns @code{word_mode}.
1287 You would most commonly define this macro if the @code{allocate_stack}
1288 pattern needs to support both a 32- and a 64-bit mode.
1291 @hook TARGET_LIBGCC_CMP_RETURN_MODE
1292 This target hook should return the mode to be used for the return value
1293 of compare instructions expanded to libgcc calls. If not defined
1294 @code{word_mode} is returned which is the right choice for a majority of
1298 @hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1299 This target hook should return the mode to be used for the shift count operand
1300 of shift instructions expanded to libgcc calls. If not defined
1301 @code{word_mode} is returned which is the right choice for a majority of
1305 @hook TARGET_UNWIND_WORD_MODE
1306 Return machine mode to be used for @code{_Unwind_Word} type.
1307 The default is to use @code{word_mode}.
1310 @defmac ROUND_TOWARDS_ZERO
1311 If defined, this macro should be true if the prevailing rounding
1312 mode is towards zero.
1314 Defining this macro only affects the way @file{libgcc.a} emulates
1315 floating-point arithmetic.
1317 Not defining this macro is equivalent to returning zero.
1320 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1321 This macro should return true if floats with @var{size}
1322 bits do not have a NaN or infinity representation, but use the largest
1323 exponent for normal numbers instead.
1325 Defining this macro only affects the way @file{libgcc.a} emulates
1326 floating-point arithmetic.
1328 The default definition of this macro returns false for all sizes.
1331 @hook TARGET_MS_BITFIELD_LAYOUT_P
1332 This target hook returns @code{true} if bit-fields in the given
1333 @var{record_type} are to be laid out following the rules of Microsoft
1334 Visual C/C++, namely: (i) a bit-field won't share the same storage
1335 unit with the previous bit-field if their underlying types have
1336 different sizes, and the bit-field will be aligned to the highest
1337 alignment of the underlying types of itself and of the previous
1338 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1339 the whole enclosing structure, even if it is unnamed; except that
1340 (iii) a zero-sized bit-field will be disregarded unless it follows
1341 another bit-field of nonzero size. If this hook returns @code{true},
1342 other macros that control bit-field layout are ignored.
1344 When a bit-field is inserted into a packed record, the whole size
1345 of the underlying type is used by one or more same-size adjacent
1346 bit-fields (that is, if its long:3, 32 bits is used in the record,
1347 and any additional adjacent long bit-fields are packed into the same
1348 chunk of 32 bits. However, if the size changes, a new field of that
1349 size is allocated). In an unpacked record, this is the same as using
1350 alignment, but not equivalent when packing.
1352 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1353 the latter will take precedence. If @samp{__attribute__((packed))} is
1354 used on a single field when MS bit-fields are in use, it will take
1355 precedence for that field, but the alignment of the rest of the structure
1356 may affect its placement.
1359 @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1360 Returns true if the target supports decimal floating point.
1363 @hook TARGET_FIXED_POINT_SUPPORTED_P
1364 Returns true if the target supports fixed-point arithmetic.
1367 @hook TARGET_EXPAND_TO_RTL_HOOK
1368 This hook is called just before expansion into rtl, allowing the target
1369 to perform additional initializations or analysis before the expansion.
1370 For example, the rs6000 port uses it to allocate a scratch stack slot
1371 for use in copying SDmode values between memory and floating point
1372 registers whenever the function being expanded has any SDmode
1376 @hook TARGET_INSTANTIATE_DECLS
1377 This hook allows the backend to perform additional instantiations on rtl
1378 that are not actually in any insns yet, but will be later.
1381 @hook TARGET_MANGLE_TYPE
1382 If your target defines any fundamental types, or any types your target
1383 uses should be mangled differently from the default, define this hook
1384 to return the appropriate encoding for these types as part of a C++
1385 mangled name. The @var{type} argument is the tree structure representing
1386 the type to be mangled. The hook may be applied to trees which are
1387 not target-specific fundamental types; it should return @code{NULL}
1388 for all such types, as well as arguments it does not recognize. If the
1389 return value is not @code{NULL}, it must point to a statically-allocated
1392 Target-specific fundamental types might be new fundamental types or
1393 qualified versions of ordinary fundamental types. Encode new
1394 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1395 is the name used for the type in source code, and @var{n} is the
1396 length of @var{name} in decimal. Encode qualified versions of
1397 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1398 @var{name} is the name used for the type qualifier in source code,
1399 @var{n} is the length of @var{name} as above, and @var{code} is the
1400 code used to represent the unqualified version of this type. (See
1401 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1402 codes.) In both cases the spaces are for clarity; do not include any
1403 spaces in your string.
1405 This hook is applied to types prior to typedef resolution. If the mangled
1406 name for a particular type depends only on that type's main variant, you
1407 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1410 The default version of this hook always returns @code{NULL}, which is
1411 appropriate for a target that does not define any new fundamental
1416 @section Layout of Source Language Data Types
1418 These macros define the sizes and other characteristics of the standard
1419 basic data types used in programs being compiled. Unlike the macros in
1420 the previous section, these apply to specific features of C and related
1421 languages, rather than to fundamental aspects of storage layout.
1423 @defmac INT_TYPE_SIZE
1424 A C expression for the size in bits of the type @code{int} on the
1425 target machine. If you don't define this, the default is one word.
1428 @defmac SHORT_TYPE_SIZE
1429 A C expression for the size in bits of the type @code{short} on the
1430 target machine. If you don't define this, the default is half a word.
1431 (If this would be less than one storage unit, it is rounded up to one
1435 @defmac LONG_TYPE_SIZE
1436 A C expression for the size in bits of the type @code{long} on the
1437 target machine. If you don't define this, the default is one word.
1440 @defmac ADA_LONG_TYPE_SIZE
1441 On some machines, the size used for the Ada equivalent of the type
1442 @code{long} by a native Ada compiler differs from that used by C@. In
1443 that situation, define this macro to be a C expression to be used for
1444 the size of that type. If you don't define this, the default is the
1445 value of @code{LONG_TYPE_SIZE}.
1448 @defmac LONG_LONG_TYPE_SIZE
1449 A C expression for the size in bits of the type @code{long long} on the
1450 target machine. If you don't define this, the default is two
1451 words. If you want to support GNU Ada on your machine, the value of this
1452 macro must be at least 64.
1455 @defmac CHAR_TYPE_SIZE
1456 A C expression for the size in bits of the type @code{char} on the
1457 target machine. If you don't define this, the default is
1458 @code{BITS_PER_UNIT}.
1461 @defmac BOOL_TYPE_SIZE
1462 A C expression for the size in bits of the C++ type @code{bool} and
1463 C99 type @code{_Bool} on the target machine. If you don't define
1464 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1467 @defmac FLOAT_TYPE_SIZE
1468 A C expression for the size in bits of the type @code{float} on the
1469 target machine. If you don't define this, the default is one word.
1472 @defmac DOUBLE_TYPE_SIZE
1473 A C expression for the size in bits of the type @code{double} on the
1474 target machine. If you don't define this, the default is two
1478 @defmac LONG_DOUBLE_TYPE_SIZE
1479 A C expression for the size in bits of the type @code{long double} on
1480 the target machine. If you don't define this, the default is two
1484 @defmac SHORT_FRACT_TYPE_SIZE
1485 A C expression for the size in bits of the type @code{short _Fract} on
1486 the target machine. If you don't define this, the default is
1487 @code{BITS_PER_UNIT}.
1490 @defmac FRACT_TYPE_SIZE
1491 A C expression for the size in bits of the type @code{_Fract} on
1492 the target machine. If you don't define this, the default is
1493 @code{BITS_PER_UNIT * 2}.
1496 @defmac LONG_FRACT_TYPE_SIZE
1497 A C expression for the size in bits of the type @code{long _Fract} on
1498 the target machine. If you don't define this, the default is
1499 @code{BITS_PER_UNIT * 4}.
1502 @defmac LONG_LONG_FRACT_TYPE_SIZE
1503 A C expression for the size in bits of the type @code{long long _Fract} on
1504 the target machine. If you don't define this, the default is
1505 @code{BITS_PER_UNIT * 8}.
1508 @defmac SHORT_ACCUM_TYPE_SIZE
1509 A C expression for the size in bits of the type @code{short _Accum} on
1510 the target machine. If you don't define this, the default is
1511 @code{BITS_PER_UNIT * 2}.
1514 @defmac ACCUM_TYPE_SIZE
1515 A C expression for the size in bits of the type @code{_Accum} on
1516 the target machine. If you don't define this, the default is
1517 @code{BITS_PER_UNIT * 4}.
1520 @defmac LONG_ACCUM_TYPE_SIZE
1521 A C expression for the size in bits of the type @code{long _Accum} on
1522 the target machine. If you don't define this, the default is
1523 @code{BITS_PER_UNIT * 8}.
1526 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1527 A C expression for the size in bits of the type @code{long long _Accum} on
1528 the target machine. If you don't define this, the default is
1529 @code{BITS_PER_UNIT * 16}.
1532 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1533 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1534 if you want routines in @file{libgcc2.a} for a size other than
1535 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1536 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1539 @defmac LIBGCC2_HAS_DF_MODE
1540 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1541 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1542 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1543 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1544 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1548 @defmac LIBGCC2_HAS_XF_MODE
1549 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1550 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1551 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1552 is 80 then the default is 1, otherwise it is 0.
1555 @defmac LIBGCC2_HAS_TF_MODE
1556 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1557 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1558 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1559 is 128 then the default is 1, otherwise it is 0.
1562 @defmac LIBGCC2_GNU_PREFIX
1563 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1564 hook and should be defined if that hook is overriden to be true. It
1565 causes function names in libgcc to be changed to use a @code{__gnu_}
1566 prefix for their name rather than the default @code{__}. A port which
1567 uses this macro should also arrange to use @file{t-gnu-prefix} in
1568 the libgcc @file{config.host}.
1575 Define these macros to be the size in bits of the mantissa of
1576 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1577 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1578 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1579 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1580 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1581 @code{DOUBLE_TYPE_SIZE} or
1582 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1585 @defmac TARGET_FLT_EVAL_METHOD
1586 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1587 assuming, if applicable, that the floating-point control word is in its
1588 default state. If you do not define this macro the value of
1589 @code{FLT_EVAL_METHOD} will be zero.
1592 @defmac WIDEST_HARDWARE_FP_SIZE
1593 A C expression for the size in bits of the widest floating-point format
1594 supported by the hardware. If you define this macro, you must specify a
1595 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1596 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1600 @defmac DEFAULT_SIGNED_CHAR
1601 An expression whose value is 1 or 0, according to whether the type
1602 @code{char} should be signed or unsigned by default. The user can
1603 always override this default with the options @option{-fsigned-char}
1604 and @option{-funsigned-char}.
1607 @hook TARGET_DEFAULT_SHORT_ENUMS
1608 This target hook should return true if the compiler should give an
1609 @code{enum} type only as many bytes as it takes to represent the range
1610 of possible values of that type. It should return false if all
1611 @code{enum} types should be allocated like @code{int}.
1613 The default is to return false.
1617 A C expression for a string describing the name of the data type to use
1618 for size values. The typedef name @code{size_t} is defined using the
1619 contents of the string.
1621 The string can contain more than one keyword. If so, separate them with
1622 spaces, and write first any length keyword, then @code{unsigned} if
1623 appropriate, and finally @code{int}. The string must exactly match one
1624 of the data type names defined in the function
1625 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1626 You may not omit @code{int} or change the order---that would cause the
1627 compiler to crash on startup.
1629 If you don't define this macro, the default is @code{"long unsigned
1634 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1635 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1636 dealing with size. This macro is a C expression for a string describing
1637 the name of the data type from which the precision of @code{sizetype}
1640 The string has the same restrictions as @code{SIZE_TYPE} string.
1642 If you don't define this macro, the default is @code{SIZE_TYPE}.
1645 @defmac PTRDIFF_TYPE
1646 A C expression for a string describing the name of the data type to use
1647 for the result of subtracting two pointers. The typedef name
1648 @code{ptrdiff_t} is defined using the contents of the string. See
1649 @code{SIZE_TYPE} above for more information.
1651 If you don't define this macro, the default is @code{"long int"}.
1655 A C expression for a string describing the name of the data type to use
1656 for wide characters. The typedef name @code{wchar_t} is defined using
1657 the contents of the string. See @code{SIZE_TYPE} above for more
1660 If you don't define this macro, the default is @code{"int"}.
1663 @defmac WCHAR_TYPE_SIZE
1664 A C expression for the size in bits of the data type for wide
1665 characters. This is used in @code{cpp}, which cannot make use of
1670 A C expression for a string describing the name of the data type to
1671 use for wide characters passed to @code{printf} and returned from
1672 @code{getwc}. The typedef name @code{wint_t} is defined using the
1673 contents of the string. See @code{SIZE_TYPE} above for more
1676 If you don't define this macro, the default is @code{"unsigned int"}.
1680 A C expression for a string describing the name of the data type that
1681 can represent any value of any standard or extended signed integer type.
1682 The typedef name @code{intmax_t} is defined using the contents of the
1683 string. See @code{SIZE_TYPE} above for more information.
1685 If you don't define this macro, the default is the first of
1686 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1687 much precision as @code{long long int}.
1690 @defmac UINTMAX_TYPE
1691 A C expression for a string describing the name of the data type that
1692 can represent any value of any standard or extended unsigned integer
1693 type. The typedef name @code{uintmax_t} is defined using the contents
1694 of the string. See @code{SIZE_TYPE} above for more information.
1696 If you don't define this macro, the default is the first of
1697 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1698 unsigned int"} that has as much precision as @code{long long unsigned
1702 @defmac SIG_ATOMIC_TYPE
1708 @defmacx UINT16_TYPE
1709 @defmacx UINT32_TYPE
1710 @defmacx UINT64_TYPE
1711 @defmacx INT_LEAST8_TYPE
1712 @defmacx INT_LEAST16_TYPE
1713 @defmacx INT_LEAST32_TYPE
1714 @defmacx INT_LEAST64_TYPE
1715 @defmacx UINT_LEAST8_TYPE
1716 @defmacx UINT_LEAST16_TYPE
1717 @defmacx UINT_LEAST32_TYPE
1718 @defmacx UINT_LEAST64_TYPE
1719 @defmacx INT_FAST8_TYPE
1720 @defmacx INT_FAST16_TYPE
1721 @defmacx INT_FAST32_TYPE
1722 @defmacx INT_FAST64_TYPE
1723 @defmacx UINT_FAST8_TYPE
1724 @defmacx UINT_FAST16_TYPE
1725 @defmacx UINT_FAST32_TYPE
1726 @defmacx UINT_FAST64_TYPE
1727 @defmacx INTPTR_TYPE
1728 @defmacx UINTPTR_TYPE
1729 C expressions for the standard types @code{sig_atomic_t},
1730 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1731 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1732 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1733 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1734 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1735 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1736 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1737 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1738 @code{SIZE_TYPE} above for more information.
1740 If any of these macros evaluates to a null pointer, the corresponding
1741 type is not supported; if GCC is configured to provide
1742 @code{<stdint.h>} in such a case, the header provided may not conform
1743 to C99, depending on the type in question. The defaults for all of
1744 these macros are null pointers.
1747 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1748 The C++ compiler represents a pointer-to-member-function with a struct
1755 ptrdiff_t vtable_index;
1762 The C++ compiler must use one bit to indicate whether the function that
1763 will be called through a pointer-to-member-function is virtual.
1764 Normally, we assume that the low-order bit of a function pointer must
1765 always be zero. Then, by ensuring that the vtable_index is odd, we can
1766 distinguish which variant of the union is in use. But, on some
1767 platforms function pointers can be odd, and so this doesn't work. In
1768 that case, we use the low-order bit of the @code{delta} field, and shift
1769 the remainder of the @code{delta} field to the left.
1771 GCC will automatically make the right selection about where to store
1772 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1773 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1774 set such that functions always start at even addresses, but the lowest
1775 bit of pointers to functions indicate whether the function at that
1776 address is in ARM or Thumb mode. If this is the case of your
1777 architecture, you should define this macro to
1778 @code{ptrmemfunc_vbit_in_delta}.
1780 In general, you should not have to define this macro. On architectures
1781 in which function addresses are always even, according to
1782 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1783 @code{ptrmemfunc_vbit_in_pfn}.
1786 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1787 Normally, the C++ compiler uses function pointers in vtables. This
1788 macro allows the target to change to use ``function descriptors''
1789 instead. Function descriptors are found on targets for whom a
1790 function pointer is actually a small data structure. Normally the
1791 data structure consists of the actual code address plus a data
1792 pointer to which the function's data is relative.
1794 If vtables are used, the value of this macro should be the number
1795 of words that the function descriptor occupies.
1798 @defmac TARGET_VTABLE_ENTRY_ALIGN
1799 By default, the vtable entries are void pointers, the so the alignment
1800 is the same as pointer alignment. The value of this macro specifies
1801 the alignment of the vtable entry in bits. It should be defined only
1802 when special alignment is necessary. */
1805 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1806 There are a few non-descriptor entries in the vtable at offsets below
1807 zero. If these entries must be padded (say, to preserve the alignment
1808 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1809 of words in each data entry.
1813 @section Register Usage
1814 @cindex register usage
1816 This section explains how to describe what registers the target machine
1817 has, and how (in general) they can be used.
1819 The description of which registers a specific instruction can use is
1820 done with register classes; see @ref{Register Classes}. For information
1821 on using registers to access a stack frame, see @ref{Frame Registers}.
1822 For passing values in registers, see @ref{Register Arguments}.
1823 For returning values in registers, see @ref{Scalar Return}.
1826 * Register Basics:: Number and kinds of registers.
1827 * Allocation Order:: Order in which registers are allocated.
1828 * Values in Registers:: What kinds of values each reg can hold.
1829 * Leaf Functions:: Renumbering registers for leaf functions.
1830 * Stack Registers:: Handling a register stack such as 80387.
1833 @node Register Basics
1834 @subsection Basic Characteristics of Registers
1836 @c prevent bad page break with this line
1837 Registers have various characteristics.
1839 @defmac FIRST_PSEUDO_REGISTER
1840 Number of hardware registers known to the compiler. They receive
1841 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1842 pseudo register's number really is assigned the number
1843 @code{FIRST_PSEUDO_REGISTER}.
1846 @defmac FIXED_REGISTERS
1847 @cindex fixed register
1848 An initializer that says which registers are used for fixed purposes
1849 all throughout the compiled code and are therefore not available for
1850 general allocation. These would include the stack pointer, the frame
1851 pointer (except on machines where that can be used as a general
1852 register when no frame pointer is needed), the program counter on
1853 machines where that is considered one of the addressable registers,
1854 and any other numbered register with a standard use.
1856 This information is expressed as a sequence of numbers, separated by
1857 commas and surrounded by braces. The @var{n}th number is 1 if
1858 register @var{n} is fixed, 0 otherwise.
1860 The table initialized from this macro, and the table initialized by
1861 the following one, may be overridden at run time either automatically,
1862 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1863 the user with the command options @option{-ffixed-@var{reg}},
1864 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1867 @defmac CALL_USED_REGISTERS
1868 @cindex call-used register
1869 @cindex call-clobbered register
1870 @cindex call-saved register
1871 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1872 clobbered (in general) by function calls as well as for fixed
1873 registers. This macro therefore identifies the registers that are not
1874 available for general allocation of values that must live across
1877 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1878 automatically saves it on function entry and restores it on function
1879 exit, if the register is used within the function.
1882 @defmac CALL_REALLY_USED_REGISTERS
1883 @cindex call-used register
1884 @cindex call-clobbered register
1885 @cindex call-saved register
1886 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1887 that the entire set of @code{FIXED_REGISTERS} be included.
1888 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1889 This macro is optional. If not specified, it defaults to the value
1890 of @code{CALL_USED_REGISTERS}.
1893 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1894 @cindex call-used register
1895 @cindex call-clobbered register
1896 @cindex call-saved register
1897 A C expression that is nonzero if it is not permissible to store a
1898 value of mode @var{mode} in hard register number @var{regno} across a
1899 call without some part of it being clobbered. For most machines this
1900 macro need not be defined. It is only required for machines that do not
1901 preserve the entire contents of a register across a call.
1905 @findex call_used_regs
1908 @findex reg_class_contents
1909 @hook TARGET_CONDITIONAL_REGISTER_USAGE
1910 This hook may conditionally modify five variables
1911 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1912 @code{reg_names}, and @code{reg_class_contents}, to take into account
1913 any dependence of these register sets on target flags. The first three
1914 of these are of type @code{char []} (interpreted as Boolean vectors).
1915 @code{global_regs} is a @code{const char *[]}, and
1916 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1917 called, @code{fixed_regs}, @code{call_used_regs},
1918 @code{reg_class_contents}, and @code{reg_names} have been initialized
1919 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1920 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1921 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1922 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1923 command options have been applied.
1925 @cindex disabling certain registers
1926 @cindex controlling register usage
1927 If the usage of an entire class of registers depends on the target
1928 flags, you may indicate this to GCC by using this macro to modify
1929 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1930 registers in the classes which should not be used by GCC@. Also define
1931 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1932 to return @code{NO_REGS} if it
1933 is called with a letter for a class that shouldn't be used.
1935 (However, if this class is not included in @code{GENERAL_REGS} and all
1936 of the insn patterns whose constraints permit this class are
1937 controlled by target switches, then GCC will automatically avoid using
1938 these registers when the target switches are opposed to them.)
1941 @defmac INCOMING_REGNO (@var{out})
1942 Define this macro if the target machine has register windows. This C
1943 expression returns the register number as seen by the called function
1944 corresponding to the register number @var{out} as seen by the calling
1945 function. Return @var{out} if register number @var{out} is not an
1949 @defmac OUTGOING_REGNO (@var{in})
1950 Define this macro if the target machine has register windows. This C
1951 expression returns the register number as seen by the calling function
1952 corresponding to the register number @var{in} as seen by the called
1953 function. Return @var{in} if register number @var{in} is not an inbound
1957 @defmac LOCAL_REGNO (@var{regno})
1958 Define this macro if the target machine has register windows. This C
1959 expression returns true if the register is call-saved but is in the
1960 register window. Unlike most call-saved registers, such registers
1961 need not be explicitly restored on function exit or during non-local
1966 If the program counter has a register number, define this as that
1967 register number. Otherwise, do not define it.
1970 @node Allocation Order
1971 @subsection Order of Allocation of Registers
1972 @cindex order of register allocation
1973 @cindex register allocation order
1975 @c prevent bad page break with this line
1976 Registers are allocated in order.
1978 @defmac REG_ALLOC_ORDER
1979 If defined, an initializer for a vector of integers, containing the
1980 numbers of hard registers in the order in which GCC should prefer
1981 to use them (from most preferred to least).
1983 If this macro is not defined, registers are used lowest numbered first
1984 (all else being equal).
1986 One use of this macro is on machines where the highest numbered
1987 registers must always be saved and the save-multiple-registers
1988 instruction supports only sequences of consecutive registers. On such
1989 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1990 the highest numbered allocable register first.
1993 @defmac ADJUST_REG_ALLOC_ORDER
1994 A C statement (sans semicolon) to choose the order in which to allocate
1995 hard registers for pseudo-registers local to a basic block.
1997 Store the desired register order in the array @code{reg_alloc_order}.
1998 Element 0 should be the register to allocate first; element 1, the next
1999 register; and so on.
2001 The macro body should not assume anything about the contents of
2002 @code{reg_alloc_order} before execution of the macro.
2004 On most machines, it is not necessary to define this macro.
2007 @defmac HONOR_REG_ALLOC_ORDER
2008 Normally, IRA tries to estimate the costs for saving a register in the
2009 prologue and restoring it in the epilogue. This discourages it from
2010 using call-saved registers. If a machine wants to ensure that IRA
2011 allocates registers in the order given by REG_ALLOC_ORDER even if some
2012 call-saved registers appear earlier than call-used ones, this macro
2016 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2017 In some case register allocation order is not enough for the
2018 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2019 If this macro is defined, it should return a floating point value
2020 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2021 be increased by approximately the pseudo's usage frequency times the
2022 value returned by this macro. Not defining this macro is equivalent
2023 to having it always return @code{0.0}.
2025 On most machines, it is not necessary to define this macro.
2028 @node Values in Registers
2029 @subsection How Values Fit in Registers
2031 This section discusses the macros that describe which kinds of values
2032 (specifically, which machine modes) each register can hold, and how many
2033 consecutive registers are needed for a given mode.
2035 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2036 A C expression for the number of consecutive hard registers, starting
2037 at register number @var{regno}, required to hold a value of mode
2038 @var{mode}. This macro must never return zero, even if a register
2039 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2040 and/or CANNOT_CHANGE_MODE_CLASS instead.
2042 On a machine where all registers are exactly one word, a suitable
2043 definition of this macro is
2046 #define HARD_REGNO_NREGS(REGNO, MODE) \
2047 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2052 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2053 A C expression that is nonzero if a value of mode @var{mode}, stored
2054 in memory, ends with padding that causes it to take up more space than
2055 in registers starting at register number @var{regno} (as determined by
2056 multiplying GCC's notion of the size of the register when containing
2057 this mode by the number of registers returned by
2058 @code{HARD_REGNO_NREGS}). By default this is zero.
2060 For example, if a floating-point value is stored in three 32-bit
2061 registers but takes up 128 bits in memory, then this would be
2064 This macros only needs to be defined if there are cases where
2065 @code{subreg_get_info}
2066 would otherwise wrongly determine that a @code{subreg} can be
2067 represented by an offset to the register number, when in fact such a
2068 @code{subreg} would contain some of the padding not stored in
2069 registers and so not be representable.
2072 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2073 For values of @var{regno} and @var{mode} for which
2074 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2075 returning the greater number of registers required to hold the value
2076 including any padding. In the example above, the value would be four.
2079 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2080 Define this macro if the natural size of registers that hold values
2081 of mode @var{mode} is not the word size. It is a C expression that
2082 should give the natural size in bytes for the specified mode. It is
2083 used by the register allocator to try to optimize its results. This
2084 happens for example on SPARC 64-bit where the natural size of
2085 floating-point registers is still 32-bit.
2088 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2089 A C expression that is nonzero if it is permissible to store a value
2090 of mode @var{mode} in hard register number @var{regno} (or in several
2091 registers starting with that one). For a machine where all registers
2092 are equivalent, a suitable definition is
2095 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2098 You need not include code to check for the numbers of fixed registers,
2099 because the allocation mechanism considers them to be always occupied.
2101 @cindex register pairs
2102 On some machines, double-precision values must be kept in even/odd
2103 register pairs. You can implement that by defining this macro to reject
2104 odd register numbers for such modes.
2106 The minimum requirement for a mode to be OK in a register is that the
2107 @samp{mov@var{mode}} instruction pattern support moves between the
2108 register and other hard register in the same class and that moving a
2109 value into the register and back out not alter it.
2111 Since the same instruction used to move @code{word_mode} will work for
2112 all narrower integer modes, it is not necessary on any machine for
2113 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2114 you define patterns @samp{movhi}, etc., to take advantage of this. This
2115 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2116 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2119 Many machines have special registers for floating point arithmetic.
2120 Often people assume that floating point machine modes are allowed only
2121 in floating point registers. This is not true. Any registers that
2122 can hold integers can safely @emph{hold} a floating point machine
2123 mode, whether or not floating arithmetic can be done on it in those
2124 registers. Integer move instructions can be used to move the values.
2126 On some machines, though, the converse is true: fixed-point machine
2127 modes may not go in floating registers. This is true if the floating
2128 registers normalize any value stored in them, because storing a
2129 non-floating value there would garble it. In this case,
2130 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2131 floating registers. But if the floating registers do not automatically
2132 normalize, if you can store any bit pattern in one and retrieve it
2133 unchanged without a trap, then any machine mode may go in a floating
2134 register, so you can define this macro to say so.
2136 The primary significance of special floating registers is rather that
2137 they are the registers acceptable in floating point arithmetic
2138 instructions. However, this is of no concern to
2139 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2140 constraints for those instructions.
2142 On some machines, the floating registers are especially slow to access,
2143 so that it is better to store a value in a stack frame than in such a
2144 register if floating point arithmetic is not being done. As long as the
2145 floating registers are not in class @code{GENERAL_REGS}, they will not
2146 be used unless some pattern's constraint asks for one.
2149 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2150 A C expression that is nonzero if it is OK to rename a hard register
2151 @var{from} to another hard register @var{to}.
2153 One common use of this macro is to prevent renaming of a register to
2154 another register that is not saved by a prologue in an interrupt
2157 The default is always nonzero.
2160 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2161 A C expression that is nonzero if a value of mode
2162 @var{mode1} is accessible in mode @var{mode2} without copying.
2164 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2165 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2166 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2167 should be nonzero. If they differ for any @var{r}, you should define
2168 this macro to return zero unless some other mechanism ensures the
2169 accessibility of the value in a narrower mode.
2171 You should define this macro to return nonzero in as many cases as
2172 possible since doing so will allow GCC to perform better register
2176 @hook TARGET_HARD_REGNO_SCRATCH_OK
2177 This target hook should return @code{true} if it is OK to use a hard register
2178 @var{regno} as scratch reg in peephole2.
2180 One common use of this macro is to prevent using of a register that
2181 is not saved by a prologue in an interrupt handler.
2183 The default version of this hook always returns @code{true}.
2186 @defmac AVOID_CCMODE_COPIES
2187 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2188 registers. You should only define this macro if support for copying to/from
2189 @code{CCmode} is incomplete.
2192 @node Leaf Functions
2193 @subsection Handling Leaf Functions
2195 @cindex leaf functions
2196 @cindex functions, leaf
2197 On some machines, a leaf function (i.e., one which makes no calls) can run
2198 more efficiently if it does not make its own register window. Often this
2199 means it is required to receive its arguments in the registers where they
2200 are passed by the caller, instead of the registers where they would
2203 The special treatment for leaf functions generally applies only when
2204 other conditions are met; for example, often they may use only those
2205 registers for its own variables and temporaries. We use the term ``leaf
2206 function'' to mean a function that is suitable for this special
2207 handling, so that functions with no calls are not necessarily ``leaf
2210 GCC assigns register numbers before it knows whether the function is
2211 suitable for leaf function treatment. So it needs to renumber the
2212 registers in order to output a leaf function. The following macros
2215 @defmac LEAF_REGISTERS
2216 Name of a char vector, indexed by hard register number, which
2217 contains 1 for a register that is allowable in a candidate for leaf
2220 If leaf function treatment involves renumbering the registers, then the
2221 registers marked here should be the ones before renumbering---those that
2222 GCC would ordinarily allocate. The registers which will actually be
2223 used in the assembler code, after renumbering, should not be marked with 1
2226 Define this macro only if the target machine offers a way to optimize
2227 the treatment of leaf functions.
2230 @defmac LEAF_REG_REMAP (@var{regno})
2231 A C expression whose value is the register number to which @var{regno}
2232 should be renumbered, when a function is treated as a leaf function.
2234 If @var{regno} is a register number which should not appear in a leaf
2235 function before renumbering, then the expression should yield @minus{}1, which
2236 will cause the compiler to abort.
2238 Define this macro only if the target machine offers a way to optimize the
2239 treatment of leaf functions, and registers need to be renumbered to do
2243 @findex current_function_is_leaf
2244 @findex current_function_uses_only_leaf_regs
2245 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2246 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2247 specially. They can test the C variable @code{current_function_is_leaf}
2248 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2249 set prior to local register allocation and is valid for the remaining
2250 compiler passes. They can also test the C variable
2251 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2252 functions which only use leaf registers.
2253 @code{current_function_uses_only_leaf_regs} is valid after all passes
2254 that modify the instructions have been run and is only useful if
2255 @code{LEAF_REGISTERS} is defined.
2256 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2257 @c of the next paragraph?! --mew 2feb93
2259 @node Stack Registers
2260 @subsection Registers That Form a Stack
2262 There are special features to handle computers where some of the
2263 ``registers'' form a stack. Stack registers are normally written by
2264 pushing onto the stack, and are numbered relative to the top of the
2267 Currently, GCC can only handle one group of stack-like registers, and
2268 they must be consecutively numbered. Furthermore, the existing
2269 support for stack-like registers is specific to the 80387 floating
2270 point coprocessor. If you have a new architecture that uses
2271 stack-like registers, you will need to do substantial work on
2272 @file{reg-stack.c} and write your machine description to cooperate
2273 with it, as well as defining these macros.
2276 Define this if the machine has any stack-like registers.
2279 @defmac STACK_REG_COVER_CLASS
2280 This is a cover class containing the stack registers. Define this if
2281 the machine has any stack-like registers.
2284 @defmac FIRST_STACK_REG
2285 The number of the first stack-like register. This one is the top
2289 @defmac LAST_STACK_REG
2290 The number of the last stack-like register. This one is the bottom of
2294 @node Register Classes
2295 @section Register Classes
2296 @cindex register class definitions
2297 @cindex class definitions, register
2299 On many machines, the numbered registers are not all equivalent.
2300 For example, certain registers may not be allowed for indexed addressing;
2301 certain registers may not be allowed in some instructions. These machine
2302 restrictions are described to the compiler using @dfn{register classes}.
2304 You define a number of register classes, giving each one a name and saying
2305 which of the registers belong to it. Then you can specify register classes
2306 that are allowed as operands to particular instruction patterns.
2310 In general, each register will belong to several classes. In fact, one
2311 class must be named @code{ALL_REGS} and contain all the registers. Another
2312 class must be named @code{NO_REGS} and contain no registers. Often the
2313 union of two classes will be another class; however, this is not required.
2315 @findex GENERAL_REGS
2316 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2317 terribly special about the name, but the operand constraint letters
2318 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2319 the same as @code{ALL_REGS}, just define it as a macro which expands
2322 Order the classes so that if class @var{x} is contained in class @var{y}
2323 then @var{x} has a lower class number than @var{y}.
2325 The way classes other than @code{GENERAL_REGS} are specified in operand
2326 constraints is through machine-dependent operand constraint letters.
2327 You can define such letters to correspond to various classes, then use
2328 them in operand constraints.
2330 You must define the narrowest register classes for allocatable
2331 registers, so that each class either has no subclasses, or that for
2332 some mode, the move cost between registers within the class is
2333 cheaper than moving a register in the class to or from memory
2336 You should define a class for the union of two classes whenever some
2337 instruction allows both classes. For example, if an instruction allows
2338 either a floating point (coprocessor) register or a general register for a
2339 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2340 which includes both of them. Otherwise you will get suboptimal code,
2341 or even internal compiler errors when reload cannot find a register in the
2342 class computed via @code{reg_class_subunion}.
2344 You must also specify certain redundant information about the register
2345 classes: for each class, which classes contain it and which ones are
2346 contained in it; for each pair of classes, the largest class contained
2349 When a value occupying several consecutive registers is expected in a
2350 certain class, all the registers used must belong to that class.
2351 Therefore, register classes cannot be used to enforce a requirement for
2352 a register pair to start with an even-numbered register. The way to
2353 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2355 Register classes used for input-operands of bitwise-and or shift
2356 instructions have a special requirement: each such class must have, for
2357 each fixed-point machine mode, a subclass whose registers can transfer that
2358 mode to or from memory. For example, on some machines, the operations for
2359 single-byte values (@code{QImode}) are limited to certain registers. When
2360 this is so, each register class that is used in a bitwise-and or shift
2361 instruction must have a subclass consisting of registers from which
2362 single-byte values can be loaded or stored. This is so that
2363 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2365 @deftp {Data type} {enum reg_class}
2366 An enumerated type that must be defined with all the register class names
2367 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2368 must be the last register class, followed by one more enumerated value,
2369 @code{LIM_REG_CLASSES}, which is not a register class but rather
2370 tells how many classes there are.
2372 Each register class has a number, which is the value of casting
2373 the class name to type @code{int}. The number serves as an index
2374 in many of the tables described below.
2377 @defmac N_REG_CLASSES
2378 The number of distinct register classes, defined as follows:
2381 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2385 @defmac REG_CLASS_NAMES
2386 An initializer containing the names of the register classes as C string
2387 constants. These names are used in writing some of the debugging dumps.
2390 @defmac REG_CLASS_CONTENTS
2391 An initializer containing the contents of the register classes, as integers
2392 which are bit masks. The @var{n}th integer specifies the contents of class
2393 @var{n}. The way the integer @var{mask} is interpreted is that
2394 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2396 When the machine has more than 32 registers, an integer does not suffice.
2397 Then the integers are replaced by sub-initializers, braced groupings containing
2398 several integers. Each sub-initializer must be suitable as an initializer
2399 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2400 In this situation, the first integer in each sub-initializer corresponds to
2401 registers 0 through 31, the second integer to registers 32 through 63, and
2405 @defmac REGNO_REG_CLASS (@var{regno})
2406 A C expression whose value is a register class containing hard register
2407 @var{regno}. In general there is more than one such class; choose a class
2408 which is @dfn{minimal}, meaning that no smaller class also contains the
2412 @defmac BASE_REG_CLASS
2413 A macro whose definition is the name of the class to which a valid
2414 base register must belong. A base register is one used in an address
2415 which is the register value plus a displacement.
2418 @defmac MODE_BASE_REG_CLASS (@var{mode})
2419 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2420 the selection of a base register in a mode dependent manner. If
2421 @var{mode} is VOIDmode then it should return the same value as
2422 @code{BASE_REG_CLASS}.
2425 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2426 A C expression whose value is the register class to which a valid
2427 base register must belong in order to be used in a base plus index
2428 register address. You should define this macro if base plus index
2429 addresses have different requirements than other base register uses.
2432 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2433 A C expression whose value is the register class to which a valid
2434 base register for a memory reference in mode @var{mode} to address
2435 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2436 define the context in which the base register occurs. @var{outer_code} is
2437 the code of the immediately enclosing expression (@code{MEM} for the top level
2438 of an address, @code{ADDRESS} for something that occurs in an
2439 @code{address_operand}). @var{index_code} is the code of the corresponding
2440 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2443 @defmac INDEX_REG_CLASS
2444 A macro whose definition is the name of the class to which a valid
2445 index register must belong. An index register is one used in an
2446 address where its value is either multiplied by a scale factor or
2447 added to another register (as well as added to a displacement).
2450 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2451 A C expression which is nonzero if register number @var{num} is
2452 suitable for use as a base register in operand addresses.
2455 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2456 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2457 that expression may examine the mode of the memory reference in
2458 @var{mode}. You should define this macro if the mode of the memory
2459 reference affects whether a register may be used as a base register. If
2460 you define this macro, the compiler will use it instead of
2461 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2462 addresses that appear outside a @code{MEM}, i.e., as an
2463 @code{address_operand}.
2466 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2467 A C expression which is nonzero if register number @var{num} is suitable for
2468 use as a base register in base plus index operand addresses, accessing
2469 memory in mode @var{mode}. It may be either a suitable hard register or a
2470 pseudo register that has been allocated such a hard register. You should
2471 define this macro if base plus index addresses have different requirements
2472 than other base register uses.
2474 Use of this macro is deprecated; please use the more general
2475 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2478 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2479 A C expression which is nonzero if register number @var{num} is
2480 suitable for use as a base register in operand addresses, accessing
2481 memory in mode @var{mode} in address space @var{address_space}.
2482 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2483 that that expression may examine the context in which the register
2484 appears in the memory reference. @var{outer_code} is the code of the
2485 immediately enclosing expression (@code{MEM} if at the top level of the
2486 address, @code{ADDRESS} for something that occurs in an
2487 @code{address_operand}). @var{index_code} is the code of the
2488 corresponding index expression if @var{outer_code} is @code{PLUS};
2489 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2490 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2493 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2494 A C expression which is nonzero if register number @var{num} is
2495 suitable for use as an index register in operand addresses. It may be
2496 either a suitable hard register or a pseudo register that has been
2497 allocated such a hard register.
2499 The difference between an index register and a base register is that
2500 the index register may be scaled. If an address involves the sum of
2501 two registers, neither one of them scaled, then either one may be
2502 labeled the ``base'' and the other the ``index''; but whichever
2503 labeling is used must fit the machine's constraints of which registers
2504 may serve in each capacity. The compiler will try both labelings,
2505 looking for one that is valid, and will reload one or both registers
2506 only if neither labeling works.
2509 @hook TARGET_PREFERRED_RENAME_CLASS
2511 @hook TARGET_PREFERRED_RELOAD_CLASS
2512 A target hook that places additional restrictions on the register class
2513 to use when it is necessary to copy value @var{x} into a register in class
2514 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2515 another, smaller class.
2517 The default version of this hook always returns value of @code{rclass} argument.
2519 Sometimes returning a more restrictive class makes better code. For
2520 example, on the 68000, when @var{x} is an integer constant that is in range
2521 for a @samp{moveq} instruction, the value of this macro is always
2522 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2523 Requiring a data register guarantees that a @samp{moveq} will be used.
2525 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2526 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2527 loaded into some register class. By returning @code{NO_REGS} you can
2528 force @var{x} into a memory location. For example, rs6000 can load
2529 immediate values into general-purpose registers, but does not have an
2530 instruction for loading an immediate value into a floating-point
2531 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2532 @var{x} is a floating-point constant. If the constant can't be loaded
2533 into any kind of register, code generation will be better if
2534 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2535 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2537 If an insn has pseudos in it after register allocation, reload will go
2538 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2539 to find the best one. Returning @code{NO_REGS}, in this case, makes
2540 reload add a @code{!} in front of the constraint: the x86 back-end uses
2541 this feature to discourage usage of 387 registers when math is done in
2542 the SSE registers (and vice versa).
2545 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2546 A C expression that places additional restrictions on the register class
2547 to use when it is necessary to copy value @var{x} into a register in class
2548 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2549 another, smaller class. On many machines, the following definition is
2553 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2556 Sometimes returning a more restrictive class makes better code. For
2557 example, on the 68000, when @var{x} is an integer constant that is in range
2558 for a @samp{moveq} instruction, the value of this macro is always
2559 @code{DATA_REGS} as long as @var{class} includes the data registers.
2560 Requiring a data register guarantees that a @samp{moveq} will be used.
2562 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2563 @var{class} is if @var{x} is a legitimate constant which cannot be
2564 loaded into some register class. By returning @code{NO_REGS} you can
2565 force @var{x} into a memory location. For example, rs6000 can load
2566 immediate values into general-purpose registers, but does not have an
2567 instruction for loading an immediate value into a floating-point
2568 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2569 @var{x} is a floating-point constant. If the constant can't be loaded
2570 into any kind of register, code generation will be better if
2571 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2572 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2574 If an insn has pseudos in it after register allocation, reload will go
2575 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2576 to find the best one. Returning @code{NO_REGS}, in this case, makes
2577 reload add a @code{!} in front of the constraint: the x86 back-end uses
2578 this feature to discourage usage of 387 registers when math is done in
2579 the SSE registers (and vice versa).
2582 @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2583 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2586 The default version of this hook always returns value of @code{rclass}
2589 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2590 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2593 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2594 A C expression that places additional restrictions on the register class
2595 to use when it is necessary to be able to hold a value of mode
2596 @var{mode} in a reload register for which class @var{class} would
2599 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2600 there are certain modes that simply can't go in certain reload classes.
2602 The value is a register class; perhaps @var{class}, or perhaps another,
2605 Don't define this macro unless the target machine has limitations which
2606 require the macro to do something nontrivial.
2609 @hook TARGET_SECONDARY_RELOAD
2610 Many machines have some registers that cannot be copied directly to or
2611 from memory or even from other types of registers. An example is the
2612 @samp{MQ} register, which on most machines, can only be copied to or
2613 from general registers, but not memory. Below, we shall be using the
2614 term 'intermediate register' when a move operation cannot be performed
2615 directly, but has to be done by copying the source into the intermediate
2616 register first, and then copying the intermediate register to the
2617 destination. An intermediate register always has the same mode as
2618 source and destination. Since it holds the actual value being copied,
2619 reload might apply optimizations to re-use an intermediate register
2620 and eliding the copy from the source when it can determine that the
2621 intermediate register still holds the required value.
2623 Another kind of secondary reload is required on some machines which
2624 allow copying all registers to and from memory, but require a scratch
2625 register for stores to some memory locations (e.g., those with symbolic
2626 address on the RT, and those with certain symbolic address on the SPARC
2627 when compiling PIC)@. Scratch registers need not have the same mode
2628 as the value being copied, and usually hold a different value than
2629 that being copied. Special patterns in the md file are needed to
2630 describe how the copy is performed with the help of the scratch register;
2631 these patterns also describe the number, register class(es) and mode(s)
2632 of the scratch register(s).
2634 In some cases, both an intermediate and a scratch register are required.
2636 For input reloads, this target hook is called with nonzero @var{in_p},
2637 and @var{x} is an rtx that needs to be copied to a register of class
2638 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2639 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2640 needs to be copied to rtx @var{x} in @var{reload_mode}.
2642 If copying a register of @var{reload_class} from/to @var{x} requires
2643 an intermediate register, the hook @code{secondary_reload} should
2644 return the register class required for this intermediate register.
2645 If no intermediate register is required, it should return NO_REGS.
2646 If more than one intermediate register is required, describe the one
2647 that is closest in the copy chain to the reload register.
2649 If scratch registers are needed, you also have to describe how to
2650 perform the copy from/to the reload register to/from this
2651 closest intermediate register. Or if no intermediate register is
2652 required, but still a scratch register is needed, describe the
2653 copy from/to the reload register to/from the reload operand @var{x}.
2655 You do this by setting @code{sri->icode} to the instruction code of a pattern
2656 in the md file which performs the move. Operands 0 and 1 are the output
2657 and input of this copy, respectively. Operands from operand 2 onward are
2658 for scratch operands. These scratch operands must have a mode, and a
2659 single-register-class
2660 @c [later: or memory]
2663 When an intermediate register is used, the @code{secondary_reload}
2664 hook will be called again to determine how to copy the intermediate
2665 register to/from the reload operand @var{x}, so your hook must also
2666 have code to handle the register class of the intermediate operand.
2668 @c [For later: maybe we'll allow multi-alternative reload patterns -
2669 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2670 @c and match the constraints of input and output to determine the required
2671 @c alternative. A restriction would be that constraints used to match
2672 @c against reloads registers would have to be written as register class
2673 @c constraints, or we need a new target macro / hook that tells us if an
2674 @c arbitrary constraint can match an unknown register of a given class.
2675 @c Such a macro / hook would also be useful in other places.]
2678 @var{x} might be a pseudo-register or a @code{subreg} of a
2679 pseudo-register, which could either be in a hard register or in memory.
2680 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2681 in memory and the hard register number if it is in a register.
2683 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2684 currently not supported. For the time being, you will have to continue
2685 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2687 @code{copy_cost} also uses this target hook to find out how values are
2688 copied. If you want it to include some extra cost for the need to allocate
2689 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2690 Or if two dependent moves are supposed to have a lower cost than the sum
2691 of the individual moves due to expected fortuitous scheduling and/or special
2692 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2695 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2696 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2697 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2698 These macros are obsolete, new ports should use the target hook
2699 @code{TARGET_SECONDARY_RELOAD} instead.
2701 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2702 target hook. Older ports still define these macros to indicate to the
2703 reload phase that it may
2704 need to allocate at least one register for a reload in addition to the
2705 register to contain the data. Specifically, if copying @var{x} to a
2706 register @var{class} in @var{mode} requires an intermediate register,
2707 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2708 largest register class all of whose registers can be used as
2709 intermediate registers or scratch registers.
2711 If copying a register @var{class} in @var{mode} to @var{x} requires an
2712 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2713 was supposed to be defined be defined to return the largest register
2714 class required. If the
2715 requirements for input and output reloads were the same, the macro
2716 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2719 The values returned by these macros are often @code{GENERAL_REGS}.
2720 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2721 can be directly copied to or from a register of @var{class} in
2722 @var{mode} without requiring a scratch register. Do not define this
2723 macro if it would always return @code{NO_REGS}.
2725 If a scratch register is required (either with or without an
2726 intermediate register), you were supposed to define patterns for
2727 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2728 (@pxref{Standard Names}. These patterns, which were normally
2729 implemented with a @code{define_expand}, should be similar to the
2730 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2733 These patterns need constraints for the reload register and scratch
2735 contain a single register class. If the original reload register (whose
2736 class is @var{class}) can meet the constraint given in the pattern, the
2737 value returned by these macros is used for the class of the scratch
2738 register. Otherwise, two additional reload registers are required.
2739 Their classes are obtained from the constraints in the insn pattern.
2741 @var{x} might be a pseudo-register or a @code{subreg} of a
2742 pseudo-register, which could either be in a hard register or in memory.
2743 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2744 in memory and the hard register number if it is in a register.
2746 These macros should not be used in the case where a particular class of
2747 registers can only be copied to memory and not to another class of
2748 registers. In that case, secondary reload registers are not needed and
2749 would not be helpful. Instead, a stack location must be used to perform
2750 the copy and the @code{mov@var{m}} pattern should use memory as an
2751 intermediate storage. This case often occurs between floating-point and
2755 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2756 Certain machines have the property that some registers cannot be copied
2757 to some other registers without using memory. Define this macro on
2758 those machines to be a C expression that is nonzero if objects of mode
2759 @var{m} in registers of @var{class1} can only be copied to registers of
2760 class @var{class2} by storing a register of @var{class1} into memory
2761 and loading that memory location into a register of @var{class2}.
2763 Do not define this macro if its value would always be zero.
2766 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2767 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2768 allocates a stack slot for a memory location needed for register copies.
2769 If this macro is defined, the compiler instead uses the memory location
2770 defined by this macro.
2772 Do not define this macro if you do not define
2773 @code{SECONDARY_MEMORY_NEEDED}.
2776 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2777 When the compiler needs a secondary memory location to copy between two
2778 registers of mode @var{mode}, it normally allocates sufficient memory to
2779 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2780 load operations in a mode that many bits wide and whose class is the
2781 same as that of @var{mode}.
2783 This is right thing to do on most machines because it ensures that all
2784 bits of the register are copied and prevents accesses to the registers
2785 in a narrower mode, which some machines prohibit for floating-point
2788 However, this default behavior is not correct on some machines, such as
2789 the DEC Alpha, that store short integers in floating-point registers
2790 differently than in integer registers. On those machines, the default
2791 widening will not work correctly and you must define this macro to
2792 suppress that widening in some cases. See the file @file{alpha.h} for
2795 Do not define this macro if you do not define
2796 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2797 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2800 @hook TARGET_CLASS_LIKELY_SPILLED_P
2801 A target hook which returns @code{true} if pseudos that have been assigned
2802 to registers of class @var{rclass} would likely be spilled because
2803 registers of @var{rclass} are needed for spill registers.
2805 The default version of this target hook returns @code{true} if @var{rclass}
2806 has exactly one register and @code{false} otherwise. On most machines, this
2807 default should be used. Only use this target hook to some other expression
2808 if pseudos allocated by @file{local-alloc.c} end up in memory because their
2809 hard registers were needed for spill registers. If this target hook returns
2810 @code{false} for those classes, those pseudos will only be allocated by
2811 @file{global.c}, which knows how to reallocate the pseudo to another
2812 register. If there would not be another register available for reallocation,
2813 you should not change the implementation of this target hook since
2814 the only effect of such implementation would be to slow down register
2818 @hook TARGET_CLASS_MAX_NREGS
2819 A target hook returns the maximum number of consecutive registers
2820 of class @var{rclass} needed to hold a value of mode @var{mode}.
2822 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2823 the value returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass},
2824 @var{mode})} target hook should be the maximum value of
2825 @code{HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno}
2826 values in the class @var{rclass}.
2828 This target hook helps control the handling of multiple-word values
2831 The default version of this target hook returns the size of @var{mode}
2835 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2836 A C expression for the maximum number of consecutive registers
2837 of class @var{class} needed to hold a value of mode @var{mode}.
2839 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2840 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2841 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2842 @var{mode})} for all @var{regno} values in the class @var{class}.
2844 This macro helps control the handling of multiple-word values
2848 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2849 If defined, a C expression that returns nonzero for a @var{class} for which
2850 a change from mode @var{from} to mode @var{to} is invalid.
2852 For the example, loading 32-bit integer or floating-point objects into
2853 floating-point registers on the Alpha extends them to 64 bits.
2854 Therefore loading a 64-bit object and then storing it as a 32-bit object
2855 does not store the low-order 32 bits, as would be the case for a normal
2856 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2860 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2861 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2862 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2866 @node Old Constraints
2867 @section Obsolete Macros for Defining Constraints
2868 @cindex defining constraints, obsolete method
2869 @cindex constraints, defining, obsolete method
2871 Machine-specific constraints can be defined with these macros instead
2872 of the machine description constructs described in @ref{Define
2873 Constraints}. This mechanism is obsolete. New ports should not use
2874 it; old ports should convert to the new mechanism.
2876 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2877 For the constraint at the start of @var{str}, which starts with the letter
2878 @var{c}, return the length. This allows you to have register class /
2879 constant / extra constraints that are longer than a single letter;
2880 you don't need to define this macro if you can do with single-letter
2881 constraints only. The definition of this macro should use
2882 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2883 to handle specially.
2884 There are some sanity checks in genoutput.c that check the constraint lengths
2885 for the md file, so you can also use this macro to help you while you are
2886 transitioning from a byzantine single-letter-constraint scheme: when you
2887 return a negative length for a constraint you want to re-use, genoutput
2888 will complain about every instance where it is used in the md file.
2891 @defmac REG_CLASS_FROM_LETTER (@var{char})
2892 A C expression which defines the machine-dependent operand constraint
2893 letters for register classes. If @var{char} is such a letter, the
2894 value should be the register class corresponding to it. Otherwise,
2895 the value should be @code{NO_REGS}. The register letter @samp{r},
2896 corresponding to class @code{GENERAL_REGS}, will not be passed
2897 to this macro; you do not need to handle it.
2900 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2901 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2902 passed in @var{str}, so that you can use suffixes to distinguish between
2906 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2907 A C expression that defines the machine-dependent operand constraint
2908 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2909 particular ranges of integer values. If @var{c} is one of those
2910 letters, the expression should check that @var{value}, an integer, is in
2911 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2912 not one of those letters, the value should be 0 regardless of
2916 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2917 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2918 string passed in @var{str}, so that you can use suffixes to distinguish
2919 between different variants.
2922 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2923 A C expression that defines the machine-dependent operand constraint
2924 letters that specify particular ranges of @code{const_double} values
2925 (@samp{G} or @samp{H}).
2927 If @var{c} is one of those letters, the expression should check that
2928 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2929 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2930 letters, the value should be 0 regardless of @var{value}.
2932 @code{const_double} is used for all floating-point constants and for
2933 @code{DImode} fixed-point constants. A given letter can accept either
2934 or both kinds of values. It can use @code{GET_MODE} to distinguish
2935 between these kinds.
2938 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2939 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2940 string passed in @var{str}, so that you can use suffixes to distinguish
2941 between different variants.
2944 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2945 A C expression that defines the optional machine-dependent constraint
2946 letters that can be used to segregate specific types of operands, usually
2947 memory references, for the target machine. Any letter that is not
2948 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2949 @code{REG_CLASS_FROM_CONSTRAINT}
2950 may be used. Normally this macro will not be defined.
2952 If it is required for a particular target machine, it should return 1
2953 if @var{value} corresponds to the operand type represented by the
2954 constraint letter @var{c}. If @var{c} is not defined as an extra
2955 constraint, the value returned should be 0 regardless of @var{value}.
2957 For example, on the ROMP, load instructions cannot have their output
2958 in r0 if the memory reference contains a symbolic address. Constraint
2959 letter @samp{Q} is defined as representing a memory address that does
2960 @emph{not} contain a symbolic address. An alternative is specified with
2961 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2962 alternative specifies @samp{m} on the input and a register class that
2963 does not include r0 on the output.
2966 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2967 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2968 in @var{str}, so that you can use suffixes to distinguish between different
2972 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2973 A C expression that defines the optional machine-dependent constraint
2974 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2975 be treated like memory constraints by the reload pass.
2977 It should return 1 if the operand type represented by the constraint
2978 at the start of @var{str}, the first letter of which is the letter @var{c},
2979 comprises a subset of all memory references including
2980 all those whose address is simply a base register. This allows the reload
2981 pass to reload an operand, if it does not directly correspond to the operand
2982 type of @var{c}, by copying its address into a base register.
2984 For example, on the S/390, some instructions do not accept arbitrary
2985 memory references, but only those that do not make use of an index
2986 register. The constraint letter @samp{Q} is defined via
2987 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2988 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2989 a @samp{Q} constraint can handle any memory operand, because the
2990 reload pass knows it can be reloaded by copying the memory address
2991 into a base register if required. This is analogous to the way
2992 an @samp{o} constraint can handle any memory operand.
2995 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2996 A C expression that defines the optional machine-dependent constraint
2997 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2998 @code{EXTRA_CONSTRAINT_STR}, that should
2999 be treated like address constraints by the reload pass.
3001 It should return 1 if the operand type represented by the constraint
3002 at the start of @var{str}, which starts with the letter @var{c}, comprises
3003 a subset of all memory addresses including
3004 all those that consist of just a base register. This allows the reload
3005 pass to reload an operand, if it does not directly correspond to the operand
3006 type of @var{str}, by copying it into a base register.
3008 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3009 be used with the @code{address_operand} predicate. It is treated
3010 analogously to the @samp{p} constraint.
3013 @node Stack and Calling
3014 @section Stack Layout and Calling Conventions
3015 @cindex calling conventions
3017 @c prevent bad page break with this line
3018 This describes the stack layout and calling conventions.
3022 * Exception Handling::
3027 * Register Arguments::
3029 * Aggregate Return::
3034 * Stack Smashing Protection::
3038 @subsection Basic Stack Layout
3039 @cindex stack frame layout
3040 @cindex frame layout
3042 @c prevent bad page break with this line
3043 Here is the basic stack layout.
3045 @defmac STACK_GROWS_DOWNWARD
3046 Define this macro if pushing a word onto the stack moves the stack
3047 pointer to a smaller address.
3049 When we say, ``define this macro if @dots{}'', it means that the
3050 compiler checks this macro only with @code{#ifdef} so the precise
3051 definition used does not matter.
3054 @defmac STACK_PUSH_CODE
3055 This macro defines the operation used when something is pushed
3056 on the stack. In RTL, a push operation will be
3057 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3059 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3060 and @code{POST_INC}. Which of these is correct depends on
3061 the stack direction and on whether the stack pointer points
3062 to the last item on the stack or whether it points to the
3063 space for the next item on the stack.
3065 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3066 defined, which is almost always right, and @code{PRE_INC} otherwise,
3067 which is often wrong.
3070 @defmac FRAME_GROWS_DOWNWARD
3071 Define this macro to nonzero value if the addresses of local variable slots
3072 are at negative offsets from the frame pointer.
3075 @defmac ARGS_GROW_DOWNWARD
3076 Define this macro if successive arguments to a function occupy decreasing
3077 addresses on the stack.
3080 @defmac STARTING_FRAME_OFFSET
3081 Offset from the frame pointer to the first local variable slot to be allocated.
3083 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3084 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3085 Otherwise, it is found by adding the length of the first slot to the
3086 value @code{STARTING_FRAME_OFFSET}.
3087 @c i'm not sure if the above is still correct.. had to change it to get
3088 @c rid of an overfull. --mew 2feb93
3091 @defmac STACK_ALIGNMENT_NEEDED
3092 Define to zero to disable final alignment of the stack during reload.
3093 The nonzero default for this macro is suitable for most ports.
3095 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3096 is a register save block following the local block that doesn't require
3097 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3098 stack alignment and do it in the backend.
3101 @defmac STACK_POINTER_OFFSET
3102 Offset from the stack pointer register to the first location at which
3103 outgoing arguments are placed. If not specified, the default value of
3104 zero is used. This is the proper value for most machines.
3106 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3107 the first location at which outgoing arguments are placed.
3110 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3111 Offset from the argument pointer register to the first argument's
3112 address. On some machines it may depend on the data type of the
3115 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3116 the first argument's address.
3119 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3120 Offset from the stack pointer register to an item dynamically allocated
3121 on the stack, e.g., by @code{alloca}.
3123 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3124 length of the outgoing arguments. The default is correct for most
3125 machines. See @file{function.c} for details.
3128 @defmac INITIAL_FRAME_ADDRESS_RTX
3129 A C expression whose value is RTL representing the address of the initial
3130 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3131 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3132 default value will be used. Define this macro in order to make frame pointer
3133 elimination work in the presence of @code{__builtin_frame_address (count)} and
3134 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3137 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3138 A C expression whose value is RTL representing the address in a stack
3139 frame where the pointer to the caller's frame is stored. Assume that
3140 @var{frameaddr} is an RTL expression for the address of the stack frame
3143 If you don't define this macro, the default is to return the value
3144 of @var{frameaddr}---that is, the stack frame address is also the
3145 address of the stack word that points to the previous frame.
3148 @defmac SETUP_FRAME_ADDRESSES
3149 If defined, a C expression that produces the machine-specific code to
3150 setup the stack so that arbitrary frames can be accessed. For example,
3151 on the SPARC, we must flush all of the register windows to the stack
3152 before we can access arbitrary stack frames. You will seldom need to
3156 @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
3157 This target hook should return an rtx that is used to store
3158 the address of the current frame into the built in @code{setjmp} buffer.
3159 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3160 machines. One reason you may need to define this target hook is if
3161 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3164 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3165 A C expression whose value is RTL representing the value of the frame
3166 address for the current frame. @var{frameaddr} is the frame pointer
3167 of the current frame. This is used for __builtin_frame_address.
3168 You need only define this macro if the frame address is not the same
3169 as the frame pointer. Most machines do not need to define it.
3172 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3173 A C expression whose value is RTL representing the value of the return
3174 address for the frame @var{count} steps up from the current frame, after
3175 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3176 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3177 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3179 The value of the expression must always be the correct address when
3180 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3181 determine the return address of other frames.
3184 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3185 Define this if the return address of a particular stack frame is accessed
3186 from the frame pointer of the previous stack frame.
3189 @defmac INCOMING_RETURN_ADDR_RTX
3190 A C expression whose value is RTL representing the location of the
3191 incoming return address at the beginning of any function, before the
3192 prologue. This RTL is either a @code{REG}, indicating that the return
3193 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3196 You only need to define this macro if you want to support call frame
3197 debugging information like that provided by DWARF 2.
3199 If this RTL is a @code{REG}, you should also define
3200 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3203 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3204 A C expression whose value is an integer giving a DWARF 2 column
3205 number that may be used as an alternative return column. The column
3206 must not correspond to any gcc hard register (that is, it must not
3207 be in the range of @code{DWARF_FRAME_REGNUM}).
3209 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3210 general register, but an alternative column needs to be used for signal
3211 frames. Some targets have also used different frame return columns
3215 @defmac DWARF_ZERO_REG
3216 A C expression whose value is an integer giving a DWARF 2 register
3217 number that is considered to always have the value zero. This should
3218 only be defined if the target has an architected zero register, and
3219 someone decided it was a good idea to use that register number to
3220 terminate the stack backtrace. New ports should avoid this.
3223 @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
3224 This target hook allows the backend to emit frame-related insns that
3225 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3226 info engine will invoke it on insns of the form
3228 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3232 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3234 to let the backend emit the call frame instructions. @var{label} is
3235 the CFI label attached to the insn, @var{pattern} is the pattern of
3236 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3239 @defmac INCOMING_FRAME_SP_OFFSET
3240 A C expression whose value is an integer giving the offset, in bytes,
3241 from the value of the stack pointer register to the top of the stack
3242 frame at the beginning of any function, before the prologue. The top of
3243 the frame is defined to be the value of the stack pointer in the
3244 previous frame, just before the call instruction.
3246 You only need to define this macro if you want to support call frame
3247 debugging information like that provided by DWARF 2.
3250 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3251 A C expression whose value is an integer giving the offset, in bytes,
3252 from the argument pointer to the canonical frame address (cfa). The
3253 final value should coincide with that calculated by
3254 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3255 during virtual register instantiation.
3257 The default value for this macro is
3258 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3259 which is correct for most machines; in general, the arguments are found
3260 immediately before the stack frame. Note that this is not the case on
3261 some targets that save registers into the caller's frame, such as SPARC
3262 and rs6000, and so such targets need to define this macro.
3264 You only need to define this macro if the default is incorrect, and you
3265 want to support call frame debugging information like that provided by
3269 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3270 If defined, a C expression whose value is an integer giving the offset
3271 in bytes from the frame pointer to the canonical frame address (cfa).
3272 The final value should coincide with that calculated by
3273 @code{INCOMING_FRAME_SP_OFFSET}.
3275 Normally the CFA is calculated as an offset from the argument pointer,
3276 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3277 variable due to the ABI, this may not be possible. If this macro is
3278 defined, it implies that the virtual register instantiation should be
3279 based on the frame pointer instead of the argument pointer. Only one
3280 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3284 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3285 If defined, a C expression whose value is an integer giving the offset
3286 in bytes from the canonical frame address (cfa) to the frame base used
3287 in DWARF 2 debug information. The default is zero. A different value
3288 may reduce the size of debug information on some ports.
3291 @node Exception Handling
3292 @subsection Exception Handling Support
3293 @cindex exception handling
3295 @defmac EH_RETURN_DATA_REGNO (@var{N})
3296 A C expression whose value is the @var{N}th register number used for
3297 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3298 @var{N} registers are usable.
3300 The exception handling library routines communicate with the exception
3301 handlers via a set of agreed upon registers. Ideally these registers
3302 should be call-clobbered; it is possible to use call-saved registers,
3303 but may negatively impact code size. The target must support at least
3304 2 data registers, but should define 4 if there are enough free registers.
3306 You must define this macro if you want to support call frame exception
3307 handling like that provided by DWARF 2.
3310 @defmac EH_RETURN_STACKADJ_RTX
3311 A C expression whose value is RTL representing a location in which
3312 to store a stack adjustment to be applied before function return.
3313 This is used to unwind the stack to an exception handler's call frame.
3314 It will be assigned zero on code paths that return normally.
3316 Typically this is a call-clobbered hard register that is otherwise
3317 untouched by the epilogue, but could also be a stack slot.
3319 Do not define this macro if the stack pointer is saved and restored
3320 by the regular prolog and epilog code in the call frame itself; in
3321 this case, the exception handling library routines will update the
3322 stack location to be restored in place. Otherwise, you must define
3323 this macro if you want to support call frame exception handling like
3324 that provided by DWARF 2.
3327 @defmac EH_RETURN_HANDLER_RTX
3328 A C expression whose value is RTL representing a location in which
3329 to store the address of an exception handler to which we should
3330 return. It will not be assigned on code paths that return normally.
3332 Typically this is the location in the call frame at which the normal
3333 return address is stored. For targets that return by popping an
3334 address off the stack, this might be a memory address just below
3335 the @emph{target} call frame rather than inside the current call
3336 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3337 been assigned, so it may be used to calculate the location of the
3340 Some targets have more complex requirements than storing to an
3341 address calculable during initial code generation. In that case
3342 the @code{eh_return} instruction pattern should be used instead.
3344 If you want to support call frame exception handling, you must
3345 define either this macro or the @code{eh_return} instruction pattern.
3348 @defmac RETURN_ADDR_OFFSET
3349 If defined, an integer-valued C expression for which rtl will be generated
3350 to add it to the exception handler address before it is searched in the
3351 exception handling tables, and to subtract it again from the address before
3352 using it to return to the exception handler.
3355 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3356 This macro chooses the encoding of pointers embedded in the exception
3357 handling sections. If at all possible, this should be defined such
3358 that the exception handling section will not require dynamic relocations,
3359 and so may be read-only.
3361 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3362 @var{global} is true if the symbol may be affected by dynamic relocations.
3363 The macro should return a combination of the @code{DW_EH_PE_*} defines
3364 as found in @file{dwarf2.h}.
3366 If this macro is not defined, pointers will not be encoded but
3367 represented directly.
3370 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3371 This macro allows the target to emit whatever special magic is required
3372 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3373 Generic code takes care of pc-relative and indirect encodings; this must
3374 be defined if the target uses text-relative or data-relative encodings.
3376 This is a C statement that branches to @var{done} if the format was
3377 handled. @var{encoding} is the format chosen, @var{size} is the number
3378 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3382 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3383 This macro allows the target to add CPU and operating system specific
3384 code to the call-frame unwinder for use when there is no unwind data
3385 available. The most common reason to implement this macro is to unwind
3386 through signal frames.
3388 This macro is called from @code{uw_frame_state_for} in
3389 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3390 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3391 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3392 for the address of the code being executed and @code{context->cfa} for
3393 the stack pointer value. If the frame can be decoded, the register
3394 save addresses should be updated in @var{fs} and the macro should
3395 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3396 the macro should evaluate to @code{_URC_END_OF_STACK}.
3398 For proper signal handling in Java this macro is accompanied by
3399 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3402 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3403 This macro allows the target to add operating system specific code to the
3404 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3405 usually used for signal or interrupt frames.
3407 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3408 @var{context} is an @code{_Unwind_Context};
3409 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3410 for the abi and context in the @code{.unwabi} directive. If the
3411 @code{.unwabi} directive can be handled, the register save addresses should
3412 be updated in @var{fs}.
3415 @defmac TARGET_USES_WEAK_UNWIND_INFO
3416 A C expression that evaluates to true if the target requires unwind
3417 info to be given comdat linkage. Define it to be @code{1} if comdat
3418 linkage is necessary. The default is @code{0}.
3421 @node Stack Checking
3422 @subsection Specifying How Stack Checking is Done
3424 GCC will check that stack references are within the boundaries of the
3425 stack, if the option @option{-fstack-check} is specified, in one of
3430 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3431 will assume that you have arranged for full stack checking to be done
3432 at appropriate places in the configuration files. GCC will not do
3433 other special processing.
3436 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3437 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3438 that you have arranged for static stack checking (checking of the
3439 static stack frame of functions) to be done at appropriate places
3440 in the configuration files. GCC will only emit code to do dynamic
3441 stack checking (checking on dynamic stack allocations) using the third
3445 If neither of the above are true, GCC will generate code to periodically
3446 ``probe'' the stack pointer using the values of the macros defined below.
3449 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3450 GCC will change its allocation strategy for large objects if the option
3451 @option{-fstack-check} is specified: they will always be allocated
3452 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3454 @defmac STACK_CHECK_BUILTIN
3455 A nonzero value if stack checking is done by the configuration files in a
3456 machine-dependent manner. You should define this macro if stack checking
3457 is required by the ABI of your machine or if you would like to do stack
3458 checking in some more efficient way than the generic approach. The default
3459 value of this macro is zero.
3462 @defmac STACK_CHECK_STATIC_BUILTIN
3463 A nonzero value if static stack checking is done by the configuration files
3464 in a machine-dependent manner. You should define this macro if you would
3465 like to do static stack checking in some more efficient way than the generic
3466 approach. The default value of this macro is zero.
3469 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3470 An integer specifying the interval at which GCC must generate stack probe
3471 instructions, defined as 2 raised to this integer. You will normally
3472 define this macro so that the interval be no larger than the size of
3473 the ``guard pages'' at the end of a stack area. The default value
3474 of 12 (4096-byte interval) is suitable for most systems.
3477 @defmac STACK_CHECK_MOVING_SP
3478 An integer which is nonzero if GCC should move the stack pointer page by page
3479 when doing probes. This can be necessary on systems where the stack pointer
3480 contains the bottom address of the memory area accessible to the executing
3481 thread at any point in time. In this situation an alternate signal stack
3482 is required in order to be able to recover from a stack overflow. The
3483 default value of this macro is zero.
3486 @defmac STACK_CHECK_PROTECT
3487 The number of bytes of stack needed to recover from a stack overflow, for
3488 languages where such a recovery is supported. The default value of 75 words
3489 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3490 8192 bytes with other exception handling mechanisms should be adequate for
3494 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3495 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3496 in the opposite case.
3498 @defmac STACK_CHECK_MAX_FRAME_SIZE
3499 The maximum size of a stack frame, in bytes. GCC will generate probe
3500 instructions in non-leaf functions to ensure at least this many bytes of
3501 stack are available. If a stack frame is larger than this size, stack
3502 checking will not be reliable and GCC will issue a warning. The
3503 default is chosen so that GCC only generates one instruction on most
3504 systems. You should normally not change the default value of this macro.
3507 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3508 GCC uses this value to generate the above warning message. It
3509 represents the amount of fixed frame used by a function, not including
3510 space for any callee-saved registers, temporaries and user variables.
3511 You need only specify an upper bound for this amount and will normally
3512 use the default of four words.
3515 @defmac STACK_CHECK_MAX_VAR_SIZE
3516 The maximum size, in bytes, of an object that GCC will place in the
3517 fixed area of the stack frame when the user specifies
3518 @option{-fstack-check}.
3519 GCC computed the default from the values of the above macros and you will
3520 normally not need to override that default.
3524 @node Frame Registers
3525 @subsection Registers That Address the Stack Frame
3527 @c prevent bad page break with this line
3528 This discusses registers that address the stack frame.
3530 @defmac STACK_POINTER_REGNUM
3531 The register number of the stack pointer register, which must also be a
3532 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3533 the hardware determines which register this is.
3536 @defmac FRAME_POINTER_REGNUM
3537 The register number of the frame pointer register, which is used to
3538 access automatic variables in the stack frame. On some machines, the
3539 hardware determines which register this is. On other machines, you can
3540 choose any register you wish for this purpose.
3543 @defmac HARD_FRAME_POINTER_REGNUM
3544 On some machines the offset between the frame pointer and starting
3545 offset of the automatic variables is not known until after register
3546 allocation has been done (for example, because the saved registers are
3547 between these two locations). On those machines, define
3548 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3549 be used internally until the offset is known, and define
3550 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3551 used for the frame pointer.
3553 You should define this macro only in the very rare circumstances when it
3554 is not possible to calculate the offset between the frame pointer and
3555 the automatic variables until after register allocation has been
3556 completed. When this macro is defined, you must also indicate in your
3557 definition of @code{ELIMINABLE_REGS} how to eliminate
3558 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3559 or @code{STACK_POINTER_REGNUM}.
3561 Do not define this macro if it would be the same as
3562 @code{FRAME_POINTER_REGNUM}.
3565 @defmac ARG_POINTER_REGNUM
3566 The register number of the arg pointer register, which is used to access
3567 the function's argument list. On some machines, this is the same as the
3568 frame pointer register. On some machines, the hardware determines which
3569 register this is. On other machines, you can choose any register you
3570 wish for this purpose. If this is not the same register as the frame
3571 pointer register, then you must mark it as a fixed register according to
3572 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3573 (@pxref{Elimination}).
3576 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3577 Define this to a preprocessor constant that is nonzero if
3578 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3579 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3580 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3581 definition is not suitable for use in preprocessor conditionals.
3584 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3585 Define this to a preprocessor constant that is nonzero if
3586 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3587 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3588 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3589 definition is not suitable for use in preprocessor conditionals.
3592 @defmac RETURN_ADDRESS_POINTER_REGNUM
3593 The register number of the return address pointer register, which is used to
3594 access the current function's return address from the stack. On some
3595 machines, the return address is not at a fixed offset from the frame
3596 pointer or stack pointer or argument pointer. This register can be defined
3597 to point to the return address on the stack, and then be converted by
3598 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3600 Do not define this macro unless there is no other way to get the return
3601 address from the stack.
3604 @defmac STATIC_CHAIN_REGNUM
3605 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3606 Register numbers used for passing a function's static chain pointer. If
3607 register windows are used, the register number as seen by the called
3608 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3609 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3610 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3613 The static chain register need not be a fixed register.
3615 If the static chain is passed in memory, these macros should not be
3616 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3619 @hook TARGET_STATIC_CHAIN
3620 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3621 targets that may use different static chain locations for different
3622 nested functions. This may be required if the target has function
3623 attributes that affect the calling conventions of the function and
3624 those calling conventions use different static chain locations.
3626 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3628 If the static chain is passed in memory, this hook should be used to
3629 provide rtx giving @code{mem} expressions that denote where they are stored.
3630 Often the @code{mem} expression as seen by the caller will be at an offset
3631 from the stack pointer and the @code{mem} expression as seen by the callee
3632 will be at an offset from the frame pointer.
3633 @findex stack_pointer_rtx
3634 @findex frame_pointer_rtx
3635 @findex arg_pointer_rtx
3636 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3637 @code{arg_pointer_rtx} will have been initialized and should be used
3638 to refer to those items.
3641 @defmac DWARF_FRAME_REGISTERS
3642 This macro specifies the maximum number of hard registers that can be
3643 saved in a call frame. This is used to size data structures used in
3644 DWARF2 exception handling.
3646 Prior to GCC 3.0, this macro was needed in order to establish a stable
3647 exception handling ABI in the face of adding new hard registers for ISA
3648 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3649 in the number of hard registers. Nevertheless, this macro can still be
3650 used to reduce the runtime memory requirements of the exception handling
3651 routines, which can be substantial if the ISA contains a lot of
3652 registers that are not call-saved.
3654 If this macro is not defined, it defaults to
3655 @code{FIRST_PSEUDO_REGISTER}.
3658 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3660 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3661 for backward compatibility in pre GCC 3.0 compiled code.
3663 If this macro is not defined, it defaults to
3664 @code{DWARF_FRAME_REGISTERS}.
3667 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3669 Define this macro if the target's representation for dwarf registers
3670 is different than the internal representation for unwind column.
3671 Given a dwarf register, this macro should return the internal unwind
3672 column number to use instead.
3674 See the PowerPC's SPE target for an example.
3677 @defmac DWARF_FRAME_REGNUM (@var{regno})
3679 Define this macro if the target's representation for dwarf registers
3680 used in .eh_frame or .debug_frame is different from that used in other
3681 debug info sections. Given a GCC hard register number, this macro
3682 should return the .eh_frame register number. The default is
3683 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3687 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3689 Define this macro to map register numbers held in the call frame info
3690 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3691 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3692 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3693 return @code{@var{regno}}.
3697 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3699 Define this macro if the target stores register values as
3700 @code{_Unwind_Word} type in unwind context. It should be defined if
3701 target register size is larger than the size of @code{void *}. The
3702 default is to store register values as @code{void *} type.
3706 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3708 Define this macro to be 1 if the target always uses extended unwind
3709 context with version, args_size and by_value fields. If it is undefined,
3710 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3711 defined and 0 otherwise.
3716 @subsection Eliminating Frame Pointer and Arg Pointer
3718 @c prevent bad page break with this line
3719 This is about eliminating the frame pointer and arg pointer.
3721 @hook TARGET_FRAME_POINTER_REQUIRED
3722 This target hook should return @code{true} if a function must have and use
3723 a frame pointer. This target hook is called in the reload pass. If its return
3724 value is @code{true} the function will have a frame pointer.
3726 This target hook can in principle examine the current function and decide
3727 according to the facts, but on most machines the constant @code{false} or the
3728 constant @code{true} suffices. Use @code{false} when the machine allows code
3729 to be generated with no frame pointer, and doing so saves some time or space.
3730 Use @code{true} when there is no possible advantage to avoiding a frame
3733 In certain cases, the compiler does not know how to produce valid code
3734 without a frame pointer. The compiler recognizes those cases and
3735 automatically gives the function a frame pointer regardless of what
3736 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3739 In a function that does not require a frame pointer, the frame pointer
3740 register can be allocated for ordinary usage, unless you mark it as a
3741 fixed register. See @code{FIXED_REGISTERS} for more information.
3743 Default return value is @code{false}.
3746 @findex get_frame_size
3747 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3748 A C statement to store in the variable @var{depth-var} the difference
3749 between the frame pointer and the stack pointer values immediately after
3750 the function prologue. The value would be computed from information
3751 such as the result of @code{get_frame_size ()} and the tables of
3752 registers @code{regs_ever_live} and @code{call_used_regs}.
3754 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3755 need not be defined. Otherwise, it must be defined even if
3756 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3757 case, you may set @var{depth-var} to anything.
3760 @defmac ELIMINABLE_REGS
3761 If defined, this macro specifies a table of register pairs used to
3762 eliminate unneeded registers that point into the stack frame. If it is not
3763 defined, the only elimination attempted by the compiler is to replace
3764 references to the frame pointer with references to the stack pointer.
3766 The definition of this macro is a list of structure initializations, each
3767 of which specifies an original and replacement register.
3769 On some machines, the position of the argument pointer is not known until
3770 the compilation is completed. In such a case, a separate hard register
3771 must be used for the argument pointer. This register can be eliminated by
3772 replacing it with either the frame pointer or the argument pointer,
3773 depending on whether or not the frame pointer has been eliminated.
3775 In this case, you might specify:
3777 #define ELIMINABLE_REGS \
3778 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3779 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3780 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3783 Note that the elimination of the argument pointer with the stack pointer is
3784 specified first since that is the preferred elimination.
3787 @hook TARGET_CAN_ELIMINATE
3788 This target hook should returns @code{true} if the compiler is allowed to
3789 try to replace register number @var{from_reg} with register number
3790 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3791 is defined, and will usually be @code{true}, since most of the cases
3792 preventing register elimination are things that the compiler already
3795 Default return value is @code{true}.
3798 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3799 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3800 specifies the initial difference between the specified pair of
3801 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3805 @node Stack Arguments
3806 @subsection Passing Function Arguments on the Stack
3807 @cindex arguments on stack
3808 @cindex stack arguments
3810 The macros in this section control how arguments are passed
3811 on the stack. See the following section for other macros that
3812 control passing certain arguments in registers.
3814 @hook TARGET_PROMOTE_PROTOTYPES
3815 This target hook returns @code{true} if an argument declared in a
3816 prototype as an integral type smaller than @code{int} should actually be
3817 passed as an @code{int}. In addition to avoiding errors in certain
3818 cases of mismatch, it also makes for better code on certain machines.
3819 The default is to not promote prototypes.
3823 A C expression. If nonzero, push insns will be used to pass
3825 If the target machine does not have a push instruction, set it to zero.
3826 That directs GCC to use an alternate strategy: to
3827 allocate the entire argument block and then store the arguments into
3828 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3831 @defmac PUSH_ARGS_REVERSED
3832 A C expression. If nonzero, function arguments will be evaluated from
3833 last to first, rather than from first to last. If this macro is not
3834 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3835 and args grow in opposite directions, and 0 otherwise.
3838 @defmac PUSH_ROUNDING (@var{npushed})
3839 A C expression that is the number of bytes actually pushed onto the
3840 stack when an instruction attempts to push @var{npushed} bytes.
3842 On some machines, the definition
3845 #define PUSH_ROUNDING(BYTES) (BYTES)
3849 will suffice. But on other machines, instructions that appear
3850 to push one byte actually push two bytes in an attempt to maintain
3851 alignment. Then the definition should be
3854 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3857 If the value of this macro has a type, it should be an unsigned type.
3860 @findex current_function_outgoing_args_size
3861 @defmac ACCUMULATE_OUTGOING_ARGS
3862 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3863 will be computed and placed into the variable
3864 @code{current_function_outgoing_args_size}. No space will be pushed
3865 onto the stack for each call; instead, the function prologue should
3866 increase the stack frame size by this amount.
3868 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3872 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3873 Define this macro if functions should assume that stack space has been
3874 allocated for arguments even when their values are passed in
3877 The value of this macro is the size, in bytes, of the area reserved for
3878 arguments passed in registers for the function represented by @var{fndecl},
3879 which can be zero if GCC is calling a library function.
3880 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3883 This space can be allocated by the caller, or be a part of the
3884 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3887 @c above is overfull. not sure what to do. --mew 5feb93 did
3888 @c something, not sure if it looks good. --mew 10feb93
3890 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3891 Define this to a nonzero value if it is the responsibility of the
3892 caller to allocate the area reserved for arguments passed in registers
3893 when calling a function of @var{fntype}. @var{fntype} may be NULL
3894 if the function called is a library function.
3896 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3897 whether the space for these arguments counts in the value of
3898 @code{current_function_outgoing_args_size}.
3901 @defmac STACK_PARMS_IN_REG_PARM_AREA
3902 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3903 stack parameters don't skip the area specified by it.
3904 @c i changed this, makes more sens and it should have taken care of the
3905 @c overfull.. not as specific, tho. --mew 5feb93
3907 Normally, when a parameter is not passed in registers, it is placed on the
3908 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3909 suppresses this behavior and causes the parameter to be passed on the
3910 stack in its natural location.
3913 @hook TARGET_RETURN_POPS_ARGS
3914 This target hook returns the number of bytes of its own arguments that
3915 a function pops on returning, or 0 if the function pops no arguments
3916 and the caller must therefore pop them all after the function returns.
3918 @var{fundecl} is a C variable whose value is a tree node that describes
3919 the function in question. Normally it is a node of type
3920 @code{FUNCTION_DECL} that describes the declaration of the function.
3921 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3923 @var{funtype} is a C variable whose value is a tree node that
3924 describes the function in question. Normally it is a node of type
3925 @code{FUNCTION_TYPE} that describes the data type of the function.
3926 From this it is possible to obtain the data types of the value and
3927 arguments (if known).
3929 When a call to a library function is being considered, @var{fundecl}
3930 will contain an identifier node for the library function. Thus, if
3931 you need to distinguish among various library functions, you can do so
3932 by their names. Note that ``library function'' in this context means
3933 a function used to perform arithmetic, whose name is known specially
3934 in the compiler and was not mentioned in the C code being compiled.
3936 @var{size} is the number of bytes of arguments passed on the
3937 stack. If a variable number of bytes is passed, it is zero, and
3938 argument popping will always be the responsibility of the calling function.
3940 On the VAX, all functions always pop their arguments, so the definition
3941 of this macro is @var{size}. On the 68000, using the standard
3942 calling convention, no functions pop their arguments, so the value of
3943 the macro is always 0 in this case. But an alternative calling
3944 convention is available in which functions that take a fixed number of
3945 arguments pop them but other functions (such as @code{printf}) pop
3946 nothing (the caller pops all). When this convention is in use,
3947 @var{funtype} is examined to determine whether a function takes a fixed
3948 number of arguments.
3951 @defmac CALL_POPS_ARGS (@var{cum})
3952 A C expression that should indicate the number of bytes a call sequence
3953 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3954 when compiling a function call.
3956 @var{cum} is the variable in which all arguments to the called function
3957 have been accumulated.
3959 On certain architectures, such as the SH5, a call trampoline is used
3960 that pops certain registers off the stack, depending on the arguments
3961 that have been passed to the function. Since this is a property of the
3962 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3966 @node Register Arguments
3967 @subsection Passing Arguments in Registers
3968 @cindex arguments in registers
3969 @cindex registers arguments
3971 This section describes the macros which let you control how various
3972 types of arguments are passed in registers or how they are arranged in
3975 @hook TARGET_FUNCTION_ARG
3976 Return an RTX indicating whether a function argument is passed in a
3977 register and if so, which register.
3979 The arguments are @var{ca}, which summarizes all the previous
3980 arguments; @var{mode}, the machine mode of the argument; @var{type},
3981 the data type of the argument as a tree node or 0 if that is not known
3982 (which happens for C support library functions); and @var{named},
3983 which is @code{true} for an ordinary argument and @code{false} for
3984 nameless arguments that correspond to @samp{@dots{}} in the called
3985 function's prototype. @var{type} can be an incomplete type if a
3986 syntax error has previously occurred.
3988 The return value is usually either a @code{reg} RTX for the hard
3989 register in which to pass the argument, or zero to pass the argument
3992 The value of the expression can also be a @code{parallel} RTX@. This is
3993 used when an argument is passed in multiple locations. The mode of the
3994 @code{parallel} should be the mode of the entire argument. The
3995 @code{parallel} holds any number of @code{expr_list} pairs; each one
3996 describes where part of the argument is passed. In each
3997 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3998 register in which to pass this part of the argument, and the mode of the
3999 register RTX indicates how large this part of the argument is. The
4000 second operand of the @code{expr_list} is a @code{const_int} which gives
4001 the offset in bytes into the entire argument of where this part starts.
4002 As a special exception the first @code{expr_list} in the @code{parallel}
4003 RTX may have a first operand of zero. This indicates that the entire
4004 argument is also stored on the stack.
4006 The last time this hook is called, it is called with @code{MODE ==
4007 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
4008 pattern as operands 2 and 3 respectively.
4010 @cindex @file{stdarg.h} and register arguments
4011 The usual way to make the ISO library @file{stdarg.h} work on a
4012 machine where some arguments are usually passed in registers, is to
4013 cause nameless arguments to be passed on the stack instead. This is
4014 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
4015 @var{named} is @code{false}.
4017 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
4018 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
4019 You may use the hook @code{targetm.calls.must_pass_in_stack}
4020 in the definition of this macro to determine if this argument is of a
4021 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4022 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
4023 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4024 defined, the argument will be computed in the stack and then loaded into
4028 @hook TARGET_MUST_PASS_IN_STACK
4029 This target hook should return @code{true} if we should not pass @var{type}
4030 solely in registers. The file @file{expr.h} defines a
4031 definition that is usually appropriate, refer to @file{expr.h} for additional
4035 @hook TARGET_FUNCTION_INCOMING_ARG
4036 Define this hook if the target machine has ``register windows'', so
4037 that the register in which a function sees an arguments is not
4038 necessarily the same as the one in which the caller passed the
4041 For such machines, @code{TARGET_FUNCTION_ARG} computes the register in
4042 which the caller passes the value, and
4043 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4044 fashion to tell the function being called where the arguments will
4047 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4048 @code{TARGET_FUNCTION_ARG} serves both purposes.
4051 @hook TARGET_ARG_PARTIAL_BYTES
4052 This target hook returns the number of bytes at the beginning of an
4053 argument that must be put in registers. The value must be zero for
4054 arguments that are passed entirely in registers or that are entirely
4055 pushed on the stack.
4057 On some machines, certain arguments must be passed partially in
4058 registers and partially in memory. On these machines, typically the
4059 first few words of arguments are passed in registers, and the rest
4060 on the stack. If a multi-word argument (a @code{double} or a
4061 structure) crosses that boundary, its first few words must be passed
4062 in registers and the rest must be pushed. This macro tells the
4063 compiler when this occurs, and how many bytes should go in registers.
4065 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4066 register to be used by the caller for this argument; likewise
4067 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4070 @hook TARGET_PASS_BY_REFERENCE
4071 This target hook should return @code{true} if an argument at the
4072 position indicated by @var{cum} should be passed by reference. This
4073 predicate is queried after target independent reasons for being
4074 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4076 If the hook returns true, a copy of that argument is made in memory and a
4077 pointer to the argument is passed instead of the argument itself.
4078 The pointer is passed in whatever way is appropriate for passing a pointer
4082 @hook TARGET_CALLEE_COPIES
4083 The function argument described by the parameters to this hook is
4084 known to be passed by reference. The hook should return true if the
4085 function argument should be copied by the callee instead of copied
4088 For any argument for which the hook returns true, if it can be
4089 determined that the argument is not modified, then a copy need
4092 The default version of this hook always returns false.
4095 @defmac CUMULATIVE_ARGS
4096 A C type for declaring a variable that is used as the first argument
4097 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4098 target machines, the type @code{int} suffices and can hold the number
4099 of bytes of argument so far.
4101 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4102 arguments that have been passed on the stack. The compiler has other
4103 variables to keep track of that. For target machines on which all
4104 arguments are passed on the stack, there is no need to store anything in
4105 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4106 should not be empty, so use @code{int}.
4109 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4110 If defined, this macro is called before generating any code for a
4111 function, but after the @var{cfun} descriptor for the function has been
4112 created. The back end may use this macro to update @var{cfun} to
4113 reflect an ABI other than that which would normally be used by default.
4114 If the compiler is generating code for a compiler-generated function,
4115 @var{fndecl} may be @code{NULL}.
4118 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4119 A C statement (sans semicolon) for initializing the variable
4120 @var{cum} for the state at the beginning of the argument list. The
4121 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4122 is the tree node for the data type of the function which will receive
4123 the args, or 0 if the args are to a compiler support library function.
4124 For direct calls that are not libcalls, @var{fndecl} contain the
4125 declaration node of the function. @var{fndecl} is also set when
4126 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4127 being compiled. @var{n_named_args} is set to the number of named
4128 arguments, including a structure return address if it is passed as a
4129 parameter, when making a call. When processing incoming arguments,
4130 @var{n_named_args} is set to @minus{}1.
4132 When processing a call to a compiler support library function,
4133 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4134 contains the name of the function, as a string. @var{libname} is 0 when
4135 an ordinary C function call is being processed. Thus, each time this
4136 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4137 never both of them at once.
4140 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4141 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4142 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4143 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4144 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4145 0)} is used instead.
4148 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4149 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4150 finding the arguments for the function being compiled. If this macro is
4151 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4153 The value passed for @var{libname} is always 0, since library routines
4154 with special calling conventions are never compiled with GCC@. The
4155 argument @var{libname} exists for symmetry with
4156 @code{INIT_CUMULATIVE_ARGS}.
4157 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4158 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4161 @hook TARGET_FUNCTION_ARG_ADVANCE
4162 This hook updates the summarizer variable pointed to by @var{ca} to
4163 advance past an argument in the argument list. The values @var{mode},
4164 @var{type} and @var{named} describe that argument. Once this is done,
4165 the variable @var{cum} is suitable for analyzing the @emph{following}
4166 argument with @code{TARGET_FUNCTION_ARG}, etc.
4168 This hook need not do anything if the argument in question was passed
4169 on the stack. The compiler knows how to track the amount of stack space
4170 used for arguments without any special help.
4173 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4174 If defined, a C expression that is the number of bytes to add to the
4175 offset of the argument passed in memory. This is needed for the SPU,
4176 which passes @code{char} and @code{short} arguments in the preferred
4177 slot that is in the middle of the quad word instead of starting at the
4181 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4182 If defined, a C expression which determines whether, and in which direction,
4183 to pad out an argument with extra space. The value should be of type
4184 @code{enum direction}: either @code{upward} to pad above the argument,
4185 @code{downward} to pad below, or @code{none} to inhibit padding.
4187 The @emph{amount} of padding is not controlled by this macro, but by the
4188 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
4189 always just enough to reach the next multiple of that boundary.
4191 This macro has a default definition which is right for most systems.
4192 For little-endian machines, the default is to pad upward. For
4193 big-endian machines, the default is to pad downward for an argument of
4194 constant size shorter than an @code{int}, and upward otherwise.
4197 @defmac PAD_VARARGS_DOWN
4198 If defined, a C expression which determines whether the default
4199 implementation of va_arg will attempt to pad down before reading the
4200 next argument, if that argument is smaller than its aligned space as
4201 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4202 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4205 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4206 Specify padding for the last element of a block move between registers and
4207 memory. @var{first} is nonzero if this is the only element. Defining this
4208 macro allows better control of register function parameters on big-endian
4209 machines, without using @code{PARALLEL} rtl. In particular,
4210 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4211 registers, as there is no longer a "wrong" part of a register; For example,
4212 a three byte aggregate may be passed in the high part of a register if so
4216 @hook TARGET_FUNCTION_ARG_BOUNDARY
4217 This hook returns the alignment boundary, in bits, of an argument
4218 with the specified mode and type. The default hook returns
4219 @code{PARM_BOUNDARY} for all arguments.
4222 @hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY
4224 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4225 A C expression that is nonzero if @var{regno} is the number of a hard
4226 register in which function arguments are sometimes passed. This does
4227 @emph{not} include implicit arguments such as the static chain and
4228 the structure-value address. On many machines, no registers can be
4229 used for this purpose since all function arguments are pushed on the
4233 @hook TARGET_SPLIT_COMPLEX_ARG
4234 This hook should return true if parameter of type @var{type} are passed
4235 as two scalar parameters. By default, GCC will attempt to pack complex
4236 arguments into the target's word size. Some ABIs require complex arguments
4237 to be split and treated as their individual components. For example, on
4238 AIX64, complex floats should be passed in a pair of floating point
4239 registers, even though a complex float would fit in one 64-bit floating
4242 The default value of this hook is @code{NULL}, which is treated as always
4246 @hook TARGET_BUILD_BUILTIN_VA_LIST
4247 This hook returns a type node for @code{va_list} for the target.
4248 The default version of the hook returns @code{void*}.
4251 @hook TARGET_ENUM_VA_LIST_P
4252 This target hook is used in function @code{c_common_nodes_and_builtins}
4253 to iterate through the target specific builtin types for va_list. The
4254 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4255 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4257 The arguments @var{pname} and @var{ptree} are used to store the result of
4258 this macro and are set to the name of the va_list builtin type and its
4260 If the return value of this macro is zero, then there is no more element.
4261 Otherwise the @var{IDX} should be increased for the next call of this
4262 macro to iterate through all types.
4265 @hook TARGET_FN_ABI_VA_LIST
4266 This hook returns the va_list type of the calling convention specified by
4268 The default version of this hook returns @code{va_list_type_node}.
4271 @hook TARGET_CANONICAL_VA_LIST_TYPE
4272 This hook returns the va_list type of the calling convention specified by the
4273 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4277 @hook TARGET_GIMPLIFY_VA_ARG_EXPR
4278 This hook performs target-specific gimplification of
4279 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4280 arguments to @code{va_arg}; the latter two are as in
4281 @code{gimplify.c:gimplify_expr}.
4284 @hook TARGET_VALID_POINTER_MODE
4285 Define this to return nonzero if the port can handle pointers
4286 with machine mode @var{mode}. The default version of this
4287 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4290 @hook TARGET_REF_MAY_ALIAS_ERRNO
4292 @hook TARGET_SCALAR_MODE_SUPPORTED_P
4293 Define this to return nonzero if the port is prepared to handle
4294 insns involving scalar mode @var{mode}. For a scalar mode to be
4295 considered supported, all the basic arithmetic and comparisons
4298 The default version of this hook returns true for any mode
4299 required to handle the basic C types (as defined by the port).
4300 Included here are the double-word arithmetic supported by the
4301 code in @file{optabs.c}.
4304 @hook TARGET_VECTOR_MODE_SUPPORTED_P
4305 Define this to return nonzero if the port is prepared to handle
4306 insns involving vector mode @var{mode}. At the very least, it
4307 must have move patterns for this mode.
4310 @hook TARGET_ARRAY_MODE_SUPPORTED_P
4312 @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
4313 Define this to return nonzero for machine modes for which the port has
4314 small register classes. If this target hook returns nonzero for a given
4315 @var{mode}, the compiler will try to minimize the lifetime of registers
4316 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4317 In this case, the hook is expected to return nonzero if it returns nonzero
4320 On some machines, it is risky to let hard registers live across arbitrary
4321 insns. Typically, these machines have instructions that require values
4322 to be in specific registers (like an accumulator), and reload will fail
4323 if the required hard register is used for another purpose across such an
4326 Passes before reload do not know which hard registers will be used
4327 in an instruction, but the machine modes of the registers set or used in
4328 the instruction are already known. And for some machines, register
4329 classes are small for, say, integer registers but not for floating point
4330 registers. For example, the AMD x86-64 architecture requires specific
4331 registers for the legacy x86 integer instructions, but there are many
4332 SSE registers for floating point operations. On such targets, a good
4333 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4334 machine modes but zero for the SSE register classes.
4336 The default version of this hook returns false for any mode. It is always
4337 safe to redefine this hook to return with a nonzero value. But if you
4338 unnecessarily define it, you will reduce the amount of optimizations
4339 that can be performed in some cases. If you do not define this hook
4340 to return a nonzero value when it is required, the compiler will run out
4341 of spill registers and print a fatal error message.
4344 @hook TARGET_FLAGS_REGNUM
4347 @subsection How Scalar Function Values Are Returned
4348 @cindex return values in registers
4349 @cindex values, returned by functions
4350 @cindex scalars, returned as values
4352 This section discusses the macros that control returning scalars as
4353 values---values that can fit in registers.
4355 @hook TARGET_FUNCTION_VALUE
4357 Define this to return an RTX representing the place where a function
4358 returns or receives a value of data type @var{ret_type}, a tree node
4359 representing a data type. @var{fn_decl_or_type} is a tree node
4360 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4361 function being called. If @var{outgoing} is false, the hook should
4362 compute the register in which the caller will see the return value.
4363 Otherwise, the hook should return an RTX representing the place where
4364 a function returns a value.
4366 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4367 (Actually, on most machines, scalar values are returned in the same
4368 place regardless of mode.) The value of the expression is usually a
4369 @code{reg} RTX for the hard register where the return value is stored.
4370 The value can also be a @code{parallel} RTX, if the return value is in
4371 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4372 @code{parallel} form. Note that the callee will populate every
4373 location specified in the @code{parallel}, but if the first element of
4374 the @code{parallel} contains the whole return value, callers will use
4375 that element as the canonical location and ignore the others. The m68k
4376 port uses this type of @code{parallel} to return pointers in both
4377 @samp{%a0} (the canonical location) and @samp{%d0}.
4379 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4380 the same promotion rules specified in @code{PROMOTE_MODE} if
4381 @var{valtype} is a scalar type.
4383 If the precise function being called is known, @var{func} is a tree
4384 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4385 pointer. This makes it possible to use a different value-returning
4386 convention for specific functions when all their calls are
4389 Some target machines have ``register windows'' so that the register in
4390 which a function returns its value is not the same as the one in which
4391 the caller sees the value. For such machines, you should return
4392 different RTX depending on @var{outgoing}.
4394 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4395 aggregate data types, because these are returned in another way. See
4396 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4399 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4400 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4401 a new target instead.
4404 @defmac LIBCALL_VALUE (@var{mode})
4405 A C expression to create an RTX representing the place where a library
4406 function returns a value of mode @var{mode}.
4408 Note that ``library function'' in this context means a compiler
4409 support routine, used to perform arithmetic, whose name is known
4410 specially by the compiler and was not mentioned in the C code being
4414 @hook TARGET_LIBCALL_VALUE
4415 Define this hook if the back-end needs to know the name of the libcall
4416 function in order to determine where the result should be returned.
4418 The mode of the result is given by @var{mode} and the name of the called
4419 library function is given by @var{fun}. The hook should return an RTX
4420 representing the place where the library function result will be returned.
4422 If this hook is not defined, then LIBCALL_VALUE will be used.
4425 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4426 A C expression that is nonzero if @var{regno} is the number of a hard
4427 register in which the values of called function may come back.
4429 A register whose use for returning values is limited to serving as the
4430 second of a pair (for a value of type @code{double}, say) need not be
4431 recognized by this macro. So for most machines, this definition
4435 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4438 If the machine has register windows, so that the caller and the called
4439 function use different registers for the return value, this macro
4440 should recognize only the caller's register numbers.
4442 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4443 for a new target instead.
4446 @hook TARGET_FUNCTION_VALUE_REGNO_P
4447 A target hook that return @code{true} if @var{regno} is the number of a hard
4448 register in which the values of called function may come back.
4450 A register whose use for returning values is limited to serving as the
4451 second of a pair (for a value of type @code{double}, say) need not be
4452 recognized by this target hook.
4454 If the machine has register windows, so that the caller and the called
4455 function use different registers for the return value, this target hook
4456 should recognize only the caller's register numbers.
4458 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4461 @defmac APPLY_RESULT_SIZE
4462 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4463 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4464 saving and restoring an arbitrary return value.
4467 @hook TARGET_RETURN_IN_MSB
4468 This hook should return true if values of type @var{type} are returned
4469 at the most significant end of a register (in other words, if they are
4470 padded at the least significant end). You can assume that @var{type}
4471 is returned in a register; the caller is required to check this.
4473 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4474 be able to hold the complete return value. For example, if a 1-, 2-
4475 or 3-byte structure is returned at the most significant end of a
4476 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4480 @node Aggregate Return
4481 @subsection How Large Values Are Returned
4482 @cindex aggregates as return values
4483 @cindex large return values
4484 @cindex returning aggregate values
4485 @cindex structure value address
4487 When a function value's mode is @code{BLKmode} (and in some other
4488 cases), the value is not returned according to
4489 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4490 caller passes the address of a block of memory in which the value
4491 should be stored. This address is called the @dfn{structure value
4494 This section describes how to control returning structure values in
4497 @hook TARGET_RETURN_IN_MEMORY
4498 This target hook should return a nonzero value to say to return the
4499 function value in memory, just as large structures are always returned.
4500 Here @var{type} will be the data type of the value, and @var{fntype}
4501 will be the type of the function doing the returning, or @code{NULL} for
4504 Note that values of mode @code{BLKmode} must be explicitly handled
4505 by this function. Also, the option @option{-fpcc-struct-return}
4506 takes effect regardless of this macro. On most systems, it is
4507 possible to leave the hook undefined; this causes a default
4508 definition to be used, whose value is the constant 1 for @code{BLKmode}
4509 values, and 0 otherwise.
4511 Do not use this hook to indicate that structures and unions should always
4512 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4516 @defmac DEFAULT_PCC_STRUCT_RETURN
4517 Define this macro to be 1 if all structure and union return values must be
4518 in memory. Since this results in slower code, this should be defined
4519 only if needed for compatibility with other compilers or with an ABI@.
4520 If you define this macro to be 0, then the conventions used for structure
4521 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4524 If not defined, this defaults to the value 1.
4527 @hook TARGET_STRUCT_VALUE_RTX
4528 This target hook should return the location of the structure value
4529 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4530 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4531 be @code{NULL}, for libcalls. You do not need to define this target
4532 hook if the address is always passed as an ``invisible'' first
4535 On some architectures the place where the structure value address
4536 is found by the called function is not the same place that the
4537 caller put it. This can be due to register windows, or it could
4538 be because the function prologue moves it to a different place.
4539 @var{incoming} is @code{1} or @code{2} when the location is needed in
4540 the context of the called function, and @code{0} in the context of
4543 If @var{incoming} is nonzero and the address is to be found on the
4544 stack, return a @code{mem} which refers to the frame pointer. If
4545 @var{incoming} is @code{2}, the result is being used to fetch the
4546 structure value address at the beginning of a function. If you need
4547 to emit adjusting code, you should do it at this point.
4550 @defmac PCC_STATIC_STRUCT_RETURN
4551 Define this macro if the usual system convention on the target machine
4552 for returning structures and unions is for the called function to return
4553 the address of a static variable containing the value.
4555 Do not define this if the usual system convention is for the caller to
4556 pass an address to the subroutine.
4558 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4559 nothing when you use @option{-freg-struct-return} mode.
4562 @hook TARGET_GET_RAW_RESULT_MODE
4564 @hook TARGET_GET_RAW_ARG_MODE
4567 @subsection Caller-Saves Register Allocation
4569 If you enable it, GCC can save registers around function calls. This
4570 makes it possible to use call-clobbered registers to hold variables that
4571 must live across calls.
4573 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4574 A C expression to determine whether it is worthwhile to consider placing
4575 a pseudo-register in a call-clobbered hard register and saving and
4576 restoring it around each function call. The expression should be 1 when
4577 this is worth doing, and 0 otherwise.
4579 If you don't define this macro, a default is used which is good on most
4580 machines: @code{4 * @var{calls} < @var{refs}}.
4583 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4584 A C expression specifying which mode is required for saving @var{nregs}
4585 of a pseudo-register in call-clobbered hard register @var{regno}. If
4586 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4587 returned. For most machines this macro need not be defined since GCC
4588 will select the smallest suitable mode.
4591 @node Function Entry
4592 @subsection Function Entry and Exit
4593 @cindex function entry and exit
4597 This section describes the macros that output function entry
4598 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4600 @hook TARGET_ASM_FUNCTION_PROLOGUE
4601 If defined, a function that outputs the assembler code for entry to a
4602 function. The prologue is responsible for setting up the stack frame,
4603 initializing the frame pointer register, saving registers that must be
4604 saved, and allocating @var{size} additional bytes of storage for the
4605 local variables. @var{size} is an integer. @var{file} is a stdio
4606 stream to which the assembler code should be output.
4608 The label for the beginning of the function need not be output by this
4609 macro. That has already been done when the macro is run.
4611 @findex regs_ever_live
4612 To determine which registers to save, the macro can refer to the array
4613 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4614 @var{r} is used anywhere within the function. This implies the function
4615 prologue should save register @var{r}, provided it is not one of the
4616 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4617 @code{regs_ever_live}.)
4619 On machines that have ``register windows'', the function entry code does
4620 not save on the stack the registers that are in the windows, even if
4621 they are supposed to be preserved by function calls; instead it takes
4622 appropriate steps to ``push'' the register stack, if any non-call-used
4623 registers are used in the function.
4625 @findex frame_pointer_needed
4626 On machines where functions may or may not have frame-pointers, the
4627 function entry code must vary accordingly; it must set up the frame
4628 pointer if one is wanted, and not otherwise. To determine whether a
4629 frame pointer is in wanted, the macro can refer to the variable
4630 @code{frame_pointer_needed}. The variable's value will be 1 at run
4631 time in a function that needs a frame pointer. @xref{Elimination}.
4633 The function entry code is responsible for allocating any stack space
4634 required for the function. This stack space consists of the regions
4635 listed below. In most cases, these regions are allocated in the
4636 order listed, with the last listed region closest to the top of the
4637 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4638 the highest address if it is not defined). You can use a different order
4639 for a machine if doing so is more convenient or required for
4640 compatibility reasons. Except in cases where required by standard
4641 or by a debugger, there is no reason why the stack layout used by GCC
4642 need agree with that used by other compilers for a machine.
4645 @hook TARGET_ASM_FUNCTION_END_PROLOGUE
4646 If defined, a function that outputs assembler code at the end of a
4647 prologue. This should be used when the function prologue is being
4648 emitted as RTL, and you have some extra assembler that needs to be
4649 emitted. @xref{prologue instruction pattern}.
4652 @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
4653 If defined, a function that outputs assembler code at the start of an
4654 epilogue. This should be used when the function epilogue is being
4655 emitted as RTL, and you have some extra assembler that needs to be
4656 emitted. @xref{epilogue instruction pattern}.
4659 @hook TARGET_ASM_FUNCTION_EPILOGUE
4660 If defined, a function that outputs the assembler code for exit from a
4661 function. The epilogue is responsible for restoring the saved
4662 registers and stack pointer to their values when the function was
4663 called, and returning control to the caller. This macro takes the
4664 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4665 registers to restore are determined from @code{regs_ever_live} and
4666 @code{CALL_USED_REGISTERS} in the same way.
4668 On some machines, there is a single instruction that does all the work
4669 of returning from the function. On these machines, give that
4670 instruction the name @samp{return} and do not define the macro
4671 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4673 Do not define a pattern named @samp{return} if you want the
4674 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4675 switches to control whether return instructions or epilogues are used,
4676 define a @samp{return} pattern with a validity condition that tests the
4677 target switches appropriately. If the @samp{return} pattern's validity
4678 condition is false, epilogues will be used.
4680 On machines where functions may or may not have frame-pointers, the
4681 function exit code must vary accordingly. Sometimes the code for these
4682 two cases is completely different. To determine whether a frame pointer
4683 is wanted, the macro can refer to the variable
4684 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4685 a function that needs a frame pointer.
4687 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4688 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4689 The C variable @code{current_function_is_leaf} is nonzero for such a
4690 function. @xref{Leaf Functions}.
4692 On some machines, some functions pop their arguments on exit while
4693 others leave that for the caller to do. For example, the 68020 when
4694 given @option{-mrtd} pops arguments in functions that take a fixed
4695 number of arguments.
4697 @findex current_function_pops_args
4698 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4699 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4700 needs to know what was decided. The number of bytes of the current
4701 function's arguments that this function should pop is available in
4702 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4707 @findex current_function_pretend_args_size
4708 A region of @code{current_function_pretend_args_size} bytes of
4709 uninitialized space just underneath the first argument arriving on the
4710 stack. (This may not be at the very start of the allocated stack region
4711 if the calling sequence has pushed anything else since pushing the stack
4712 arguments. But usually, on such machines, nothing else has been pushed
4713 yet, because the function prologue itself does all the pushing.) This
4714 region is used on machines where an argument may be passed partly in
4715 registers and partly in memory, and, in some cases to support the
4716 features in @code{<stdarg.h>}.
4719 An area of memory used to save certain registers used by the function.
4720 The size of this area, which may also include space for such things as
4721 the return address and pointers to previous stack frames, is
4722 machine-specific and usually depends on which registers have been used
4723 in the function. Machines with register windows often do not require
4727 A region of at least @var{size} bytes, possibly rounded up to an allocation
4728 boundary, to contain the local variables of the function. On some machines,
4729 this region and the save area may occur in the opposite order, with the
4730 save area closer to the top of the stack.
4733 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4734 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4735 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4736 argument lists of the function. @xref{Stack Arguments}.
4739 @defmac EXIT_IGNORE_STACK
4740 Define this macro as a C expression that is nonzero if the return
4741 instruction or the function epilogue ignores the value of the stack
4742 pointer; in other words, if it is safe to delete an instruction to
4743 adjust the stack pointer before a return from the function. The
4746 Note that this macro's value is relevant only for functions for which
4747 frame pointers are maintained. It is never safe to delete a final
4748 stack adjustment in a function that has no frame pointer, and the
4749 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4752 @defmac EPILOGUE_USES (@var{regno})
4753 Define this macro as a C expression that is nonzero for registers that are
4754 used by the epilogue or the @samp{return} pattern. The stack and frame
4755 pointer registers are already assumed to be used as needed.
4758 @defmac EH_USES (@var{regno})
4759 Define this macro as a C expression that is nonzero for registers that are
4760 used by the exception handling mechanism, and so should be considered live
4761 on entry to an exception edge.
4764 @defmac DELAY_SLOTS_FOR_EPILOGUE
4765 Define this macro if the function epilogue contains delay slots to which
4766 instructions from the rest of the function can be ``moved''. The
4767 definition should be a C expression whose value is an integer
4768 representing the number of delay slots there.
4771 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4772 A C expression that returns 1 if @var{insn} can be placed in delay
4773 slot number @var{n} of the epilogue.
4775 The argument @var{n} is an integer which identifies the delay slot now
4776 being considered (since different slots may have different rules of
4777 eligibility). It is never negative and is always less than the number
4778 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4779 If you reject a particular insn for a given delay slot, in principle, it
4780 may be reconsidered for a subsequent delay slot. Also, other insns may
4781 (at least in principle) be considered for the so far unfilled delay
4784 @findex current_function_epilogue_delay_list
4785 @findex final_scan_insn
4786 The insns accepted to fill the epilogue delay slots are put in an RTL
4787 list made with @code{insn_list} objects, stored in the variable
4788 @code{current_function_epilogue_delay_list}. The insn for the first
4789 delay slot comes first in the list. Your definition of the macro
4790 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4791 outputting the insns in this list, usually by calling
4792 @code{final_scan_insn}.
4794 You need not define this macro if you did not define
4795 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4798 @hook TARGET_ASM_OUTPUT_MI_THUNK
4799 A function that outputs the assembler code for a thunk
4800 function, used to implement C++ virtual function calls with multiple
4801 inheritance. The thunk acts as a wrapper around a virtual function,
4802 adjusting the implicit object parameter before handing control off to
4805 First, emit code to add the integer @var{delta} to the location that
4806 contains the incoming first argument. Assume that this argument
4807 contains a pointer, and is the one used to pass the @code{this} pointer
4808 in C++. This is the incoming argument @emph{before} the function prologue,
4809 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4810 all other incoming arguments.
4812 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4813 made after adding @code{delta}. In particular, if @var{p} is the
4814 adjusted pointer, the following adjustment should be made:
4817 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4820 After the additions, emit code to jump to @var{function}, which is a
4821 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4822 not touch the return address. Hence returning from @var{FUNCTION} will
4823 return to whoever called the current @samp{thunk}.
4825 The effect must be as if @var{function} had been called directly with
4826 the adjusted first argument. This macro is responsible for emitting all
4827 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4828 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4830 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4831 have already been extracted from it.) It might possibly be useful on
4832 some targets, but probably not.
4834 If you do not define this macro, the target-independent code in the C++
4835 front end will generate a less efficient heavyweight thunk that calls
4836 @var{function} instead of jumping to it. The generic approach does
4837 not support varargs.
4840 @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
4841 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4842 to output the assembler code for the thunk function specified by the
4843 arguments it is passed, and false otherwise. In the latter case, the
4844 generic approach will be used by the C++ front end, with the limitations
4849 @subsection Generating Code for Profiling
4850 @cindex profiling, code generation
4852 These macros will help you generate code for profiling.
4854 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4855 A C statement or compound statement to output to @var{file} some
4856 assembler code to call the profiling subroutine @code{mcount}.
4859 The details of how @code{mcount} expects to be called are determined by
4860 your operating system environment, not by GCC@. To figure them out,
4861 compile a small program for profiling using the system's installed C
4862 compiler and look at the assembler code that results.
4864 Older implementations of @code{mcount} expect the address of a counter
4865 variable to be loaded into some register. The name of this variable is
4866 @samp{LP} followed by the number @var{labelno}, so you would generate
4867 the name using @samp{LP%d} in a @code{fprintf}.
4870 @defmac PROFILE_HOOK
4871 A C statement or compound statement to output to @var{file} some assembly
4872 code to call the profiling subroutine @code{mcount} even the target does
4873 not support profiling.
4876 @defmac NO_PROFILE_COUNTERS
4877 Define this macro to be an expression with a nonzero value if the
4878 @code{mcount} subroutine on your system does not need a counter variable
4879 allocated for each function. This is true for almost all modern
4880 implementations. If you define this macro, you must not use the
4881 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4884 @defmac PROFILE_BEFORE_PROLOGUE
4885 Define this macro if the code for function profiling should come before
4886 the function prologue. Normally, the profiling code comes after.
4890 @subsection Permitting tail calls
4893 @hook TARGET_FUNCTION_OK_FOR_SIBCALL
4894 True if it is ok to do sibling call optimization for the specified
4895 call expression @var{exp}. @var{decl} will be the called function,
4896 or @code{NULL} if this is an indirect call.
4898 It is not uncommon for limitations of calling conventions to prevent
4899 tail calls to functions outside the current unit of translation, or
4900 during PIC compilation. The hook is used to enforce these restrictions,
4901 as the @code{sibcall} md pattern can not fail, or fall over to a
4902 ``normal'' call. The criteria for successful sibling call optimization
4903 may vary greatly between different architectures.
4906 @hook TARGET_EXTRA_LIVE_ON_ENTRY
4907 Add any hard registers to @var{regs} that are live on entry to the
4908 function. This hook only needs to be defined to provide registers that
4909 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4910 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4911 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4912 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4915 @hook TARGET_SET_UP_BY_PROLOGUE
4917 @node Stack Smashing Protection
4918 @subsection Stack smashing protection
4919 @cindex stack smashing protection
4921 @hook TARGET_STACK_PROTECT_GUARD
4922 This hook returns a @code{DECL} node for the external variable to use
4923 for the stack protection guard. This variable is initialized by the
4924 runtime to some random value and is used to initialize the guard value
4925 that is placed at the top of the local stack frame. The type of this
4926 variable must be @code{ptr_type_node}.
4928 The default version of this hook creates a variable called
4929 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4932 @hook TARGET_STACK_PROTECT_FAIL
4933 This hook returns a tree expression that alerts the runtime that the
4934 stack protect guard variable has been modified. This expression should
4935 involve a call to a @code{noreturn} function.
4937 The default version of this hook invokes a function called
4938 @samp{__stack_chk_fail}, taking no arguments. This function is
4939 normally defined in @file{libgcc2.c}.
4942 @hook TARGET_SUPPORTS_SPLIT_STACK
4945 @section Implementing the Varargs Macros
4946 @cindex varargs implementation
4948 GCC comes with an implementation of @code{<varargs.h>} and
4949 @code{<stdarg.h>} that work without change on machines that pass arguments
4950 on the stack. Other machines require their own implementations of
4951 varargs, and the two machine independent header files must have
4952 conditionals to include it.
4954 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4955 the calling convention for @code{va_start}. The traditional
4956 implementation takes just one argument, which is the variable in which
4957 to store the argument pointer. The ISO implementation of
4958 @code{va_start} takes an additional second argument. The user is
4959 supposed to write the last named argument of the function here.
4961 However, @code{va_start} should not use this argument. The way to find
4962 the end of the named arguments is with the built-in functions described
4965 @defmac __builtin_saveregs ()
4966 Use this built-in function to save the argument registers in memory so
4967 that the varargs mechanism can access them. Both ISO and traditional
4968 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4969 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4971 On some machines, @code{__builtin_saveregs} is open-coded under the
4972 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4973 other machines, it calls a routine written in assembler language,
4974 found in @file{libgcc2.c}.
4976 Code generated for the call to @code{__builtin_saveregs} appears at the
4977 beginning of the function, as opposed to where the call to
4978 @code{__builtin_saveregs} is written, regardless of what the code is.
4979 This is because the registers must be saved before the function starts
4980 to use them for its own purposes.
4981 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4985 @defmac __builtin_next_arg (@var{lastarg})
4986 This builtin returns the address of the first anonymous stack
4987 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4988 returns the address of the location above the first anonymous stack
4989 argument. Use it in @code{va_start} to initialize the pointer for
4990 fetching arguments from the stack. Also use it in @code{va_start} to
4991 verify that the second parameter @var{lastarg} is the last named argument
4992 of the current function.
4995 @defmac __builtin_classify_type (@var{object})
4996 Since each machine has its own conventions for which data types are
4997 passed in which kind of register, your implementation of @code{va_arg}
4998 has to embody these conventions. The easiest way to categorize the
4999 specified data type is to use @code{__builtin_classify_type} together
5000 with @code{sizeof} and @code{__alignof__}.
5002 @code{__builtin_classify_type} ignores the value of @var{object},
5003 considering only its data type. It returns an integer describing what
5004 kind of type that is---integer, floating, pointer, structure, and so on.
5006 The file @file{typeclass.h} defines an enumeration that you can use to
5007 interpret the values of @code{__builtin_classify_type}.
5010 These machine description macros help implement varargs:
5012 @hook TARGET_EXPAND_BUILTIN_SAVEREGS
5013 If defined, this hook produces the machine-specific code for a call to
5014 @code{__builtin_saveregs}. This code will be moved to the very
5015 beginning of the function, before any parameter access are made. The
5016 return value of this function should be an RTX that contains the value
5017 to use as the return of @code{__builtin_saveregs}.
5020 @hook TARGET_SETUP_INCOMING_VARARGS
5021 This target hook offers an alternative to using
5022 @code{__builtin_saveregs} and defining the hook
5023 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5024 register arguments into the stack so that all the arguments appear to
5025 have been passed consecutively on the stack. Once this is done, you can
5026 use the standard implementation of varargs that works for machines that
5027 pass all their arguments on the stack.
5029 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5030 structure, containing the values that are obtained after processing the
5031 named arguments. The arguments @var{mode} and @var{type} describe the
5032 last named argument---its machine mode and its data type as a tree node.
5034 The target hook should do two things: first, push onto the stack all the
5035 argument registers @emph{not} used for the named arguments, and second,
5036 store the size of the data thus pushed into the @code{int}-valued
5037 variable pointed to by @var{pretend_args_size}. The value that you
5038 store here will serve as additional offset for setting up the stack
5041 Because you must generate code to push the anonymous arguments at
5042 compile time without knowing their data types,
5043 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5044 have just a single category of argument register and use it uniformly
5047 If the argument @var{second_time} is nonzero, it means that the
5048 arguments of the function are being analyzed for the second time. This
5049 happens for an inline function, which is not actually compiled until the
5050 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5051 not generate any instructions in this case.
5054 @hook TARGET_STRICT_ARGUMENT_NAMING
5055 Define this hook to return @code{true} if the location where a function
5056 argument is passed depends on whether or not it is a named argument.
5058 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5059 is set for varargs and stdarg functions. If this hook returns
5060 @code{true}, the @var{named} argument is always true for named
5061 arguments, and false for unnamed arguments. If it returns @code{false},
5062 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5063 then all arguments are treated as named. Otherwise, all named arguments
5064 except the last are treated as named.
5066 You need not define this hook if it always returns @code{false}.
5069 @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
5070 If you need to conditionally change ABIs so that one works with
5071 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5072 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5073 defined, then define this hook to return @code{true} if
5074 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5075 Otherwise, you should not define this hook.
5079 @section Trampolines for Nested Functions
5080 @cindex trampolines for nested functions
5081 @cindex nested functions, trampolines for
5083 A @dfn{trampoline} is a small piece of code that is created at run time
5084 when the address of a nested function is taken. It normally resides on
5085 the stack, in the stack frame of the containing function. These macros
5086 tell GCC how to generate code to allocate and initialize a
5089 The instructions in the trampoline must do two things: load a constant
5090 address into the static chain register, and jump to the real address of
5091 the nested function. On CISC machines such as the m68k, this requires
5092 two instructions, a move immediate and a jump. Then the two addresses
5093 exist in the trampoline as word-long immediate operands. On RISC
5094 machines, it is often necessary to load each address into a register in
5095 two parts. Then pieces of each address form separate immediate
5098 The code generated to initialize the trampoline must store the variable
5099 parts---the static chain value and the function address---into the
5100 immediate operands of the instructions. On a CISC machine, this is
5101 simply a matter of copying each address to a memory reference at the
5102 proper offset from the start of the trampoline. On a RISC machine, it
5103 may be necessary to take out pieces of the address and store them
5106 @hook TARGET_ASM_TRAMPOLINE_TEMPLATE
5107 This hook is called by @code{assemble_trampoline_template} to output,
5108 on the stream @var{f}, assembler code for a block of data that contains
5109 the constant parts of a trampoline. This code should not include a
5110 label---the label is taken care of automatically.
5112 If you do not define this hook, it means no template is needed
5113 for the target. Do not define this hook on systems where the block move
5114 code to copy the trampoline into place would be larger than the code
5115 to generate it on the spot.
5118 @defmac TRAMPOLINE_SECTION
5119 Return the section into which the trampoline template is to be placed
5120 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5123 @defmac TRAMPOLINE_SIZE
5124 A C expression for the size in bytes of the trampoline, as an integer.
5127 @defmac TRAMPOLINE_ALIGNMENT
5128 Alignment required for trampolines, in bits.
5130 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5131 is used for aligning trampolines.
5134 @hook TARGET_TRAMPOLINE_INIT
5135 This hook is called to initialize a trampoline.
5136 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5137 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5138 RTX for the static chain value that should be passed to the function
5141 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5142 first thing this hook should do is emit a block move into @var{m_tramp}
5143 from the memory block returned by @code{assemble_trampoline_template}.
5144 Note that the block move need only cover the constant parts of the
5145 trampoline. If the target isolates the variable parts of the trampoline
5146 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5148 If the target requires any other actions, such as flushing caches or
5149 enabling stack execution, these actions should be performed after
5150 initializing the trampoline proper.
5153 @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
5154 This hook should perform any machine-specific adjustment in
5155 the address of the trampoline. Its argument contains the address of the
5156 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5157 the address to be used for a function call should be different from the
5158 address at which the template was stored, the different address should
5159 be returned; otherwise @var{addr} should be returned unchanged.
5160 If this hook is not defined, @var{addr} will be used for function calls.
5163 Implementing trampolines is difficult on many machines because they have
5164 separate instruction and data caches. Writing into a stack location
5165 fails to clear the memory in the instruction cache, so when the program
5166 jumps to that location, it executes the old contents.
5168 Here are two possible solutions. One is to clear the relevant parts of
5169 the instruction cache whenever a trampoline is set up. The other is to
5170 make all trampolines identical, by having them jump to a standard
5171 subroutine. The former technique makes trampoline execution faster; the
5172 latter makes initialization faster.
5174 To clear the instruction cache when a trampoline is initialized, define
5175 the following macro.
5177 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5178 If defined, expands to a C expression clearing the @emph{instruction
5179 cache} in the specified interval. The definition of this macro would
5180 typically be a series of @code{asm} statements. Both @var{beg} and
5181 @var{end} are both pointer expressions.
5184 To use a standard subroutine, define the following macro. In addition,
5185 you must make sure that the instructions in a trampoline fill an entire
5186 cache line with identical instructions, or else ensure that the
5187 beginning of the trampoline code is always aligned at the same point in
5188 its cache line. Look in @file{m68k.h} as a guide.
5190 @defmac TRANSFER_FROM_TRAMPOLINE
5191 Define this macro if trampolines need a special subroutine to do their
5192 work. The macro should expand to a series of @code{asm} statements
5193 which will be compiled with GCC@. They go in a library function named
5194 @code{__transfer_from_trampoline}.
5196 If you need to avoid executing the ordinary prologue code of a compiled
5197 C function when you jump to the subroutine, you can do so by placing a
5198 special label of your own in the assembler code. Use one @code{asm}
5199 statement to generate an assembler label, and another to make the label
5200 global. Then trampolines can use that label to jump directly to your
5201 special assembler code.
5205 @section Implicit Calls to Library Routines
5206 @cindex library subroutine names
5207 @cindex @file{libgcc.a}
5209 @c prevent bad page break with this line
5210 Here is an explanation of implicit calls to library routines.
5212 @defmac DECLARE_LIBRARY_RENAMES
5213 This macro, if defined, should expand to a piece of C code that will get
5214 expanded when compiling functions for libgcc.a. It can be used to
5215 provide alternate names for GCC's internal library functions if there
5216 are ABI-mandated names that the compiler should provide.
5219 @findex set_optab_libfunc
5220 @findex init_one_libfunc
5221 @hook TARGET_INIT_LIBFUNCS
5222 This hook should declare additional library routines or rename
5223 existing ones, using the functions @code{set_optab_libfunc} and
5224 @code{init_one_libfunc} defined in @file{optabs.c}.
5225 @code{init_optabs} calls this macro after initializing all the normal
5228 The default is to do nothing. Most ports don't need to define this hook.
5231 @hook TARGET_LIBFUNC_GNU_PREFIX
5233 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5234 This macro should return @code{true} if the library routine that
5235 implements the floating point comparison operator @var{comparison} in
5236 mode @var{mode} will return a boolean, and @var{false} if it will
5239 GCC's own floating point libraries return tristates from the
5240 comparison operators, so the default returns false always. Most ports
5241 don't need to define this macro.
5244 @defmac TARGET_LIB_INT_CMP_BIASED
5245 This macro should evaluate to @code{true} if the integer comparison
5246 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5247 operand is smaller than the second, 1 to indicate that they are equal,
5248 and 2 to indicate that the first operand is greater than the second.
5249 If this macro evaluates to @code{false} the comparison functions return
5250 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5251 in @file{libgcc.a}, you do not need to define this macro.
5254 @cindex @code{EDOM}, implicit usage
5257 The value of @code{EDOM} on the target machine, as a C integer constant
5258 expression. If you don't define this macro, GCC does not attempt to
5259 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5260 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5263 If you do not define @code{TARGET_EDOM}, then compiled code reports
5264 domain errors by calling the library function and letting it report the
5265 error. If mathematical functions on your system use @code{matherr} when
5266 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5267 that @code{matherr} is used normally.
5270 @cindex @code{errno}, implicit usage
5271 @defmac GEN_ERRNO_RTX
5272 Define this macro as a C expression to create an rtl expression that
5273 refers to the global ``variable'' @code{errno}. (On certain systems,
5274 @code{errno} may not actually be a variable.) If you don't define this
5275 macro, a reasonable default is used.
5278 @cindex C99 math functions, implicit usage
5279 @defmac TARGET_C99_FUNCTIONS
5280 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5281 @code{sinf} and similarly for other functions defined by C99 standard. The
5282 default is zero because a number of existing systems lack support for these
5283 functions in their runtime so this macro needs to be redefined to one on
5284 systems that do support the C99 runtime.
5287 @cindex sincos math function, implicit usage
5288 @defmac TARGET_HAS_SINCOS
5289 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5290 and @code{cos} with the same argument to a call to @code{sincos}. The
5291 default is zero. The target has to provide the following functions:
5293 void sincos(double x, double *sin, double *cos);
5294 void sincosf(float x, float *sin, float *cos);
5295 void sincosl(long double x, long double *sin, long double *cos);
5299 @defmac NEXT_OBJC_RUNTIME
5300 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
5301 by default. This calling convention involves passing the object, the selector
5302 and the method arguments all at once to the method-lookup library function.
5303 This is the usual setting when targeting Darwin/Mac OS X systems, which have
5304 the NeXT runtime installed.
5306 If the macro is set to 0, the "GNU" Objective-C message sending convention
5307 will be used by default. This convention passes just the object and the
5308 selector to the method-lookup function, which returns a pointer to the method.
5310 In either case, it remains possible to select code-generation for the alternate
5311 scheme, by means of compiler command line switches.
5314 @node Addressing Modes
5315 @section Addressing Modes
5316 @cindex addressing modes
5318 @c prevent bad page break with this line
5319 This is about addressing modes.
5321 @defmac HAVE_PRE_INCREMENT
5322 @defmacx HAVE_PRE_DECREMENT
5323 @defmacx HAVE_POST_INCREMENT
5324 @defmacx HAVE_POST_DECREMENT
5325 A C expression that is nonzero if the machine supports pre-increment,
5326 pre-decrement, post-increment, or post-decrement addressing respectively.
5329 @defmac HAVE_PRE_MODIFY_DISP
5330 @defmacx HAVE_POST_MODIFY_DISP
5331 A C expression that is nonzero if the machine supports pre- or
5332 post-address side-effect generation involving constants other than
5333 the size of the memory operand.
5336 @defmac HAVE_PRE_MODIFY_REG
5337 @defmacx HAVE_POST_MODIFY_REG
5338 A C expression that is nonzero if the machine supports pre- or
5339 post-address side-effect generation involving a register displacement.
5342 @defmac CONSTANT_ADDRESS_P (@var{x})
5343 A C expression that is 1 if the RTX @var{x} is a constant which
5344 is a valid address. On most machines the default definition of
5345 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5346 is acceptable, but a few machines are more restrictive as to which
5347 constant addresses are supported.
5350 @defmac CONSTANT_P (@var{x})
5351 @code{CONSTANT_P}, which is defined by target-independent code,
5352 accepts integer-values expressions whose values are not explicitly
5353 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5354 expressions and @code{const} arithmetic expressions, in addition to
5355 @code{const_int} and @code{const_double} expressions.
5358 @defmac MAX_REGS_PER_ADDRESS
5359 A number, the maximum number of registers that can appear in a valid
5360 memory address. Note that it is up to you to specify a value equal to
5361 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5365 @hook TARGET_LEGITIMATE_ADDRESS_P
5366 A function that returns whether @var{x} (an RTX) is a legitimate memory
5367 address on the target machine for a memory operand of mode @var{mode}.
5369 Legitimate addresses are defined in two variants: a strict variant and a
5370 non-strict one. The @var{strict} parameter chooses which variant is
5371 desired by the caller.
5373 The strict variant is used in the reload pass. It must be defined so
5374 that any pseudo-register that has not been allocated a hard register is
5375 considered a memory reference. This is because in contexts where some
5376 kind of register is required, a pseudo-register with no hard register
5377 must be rejected. For non-hard registers, the strict variant should look
5378 up the @code{reg_renumber} array; it should then proceed using the hard
5379 register number in the array, or treat the pseudo as a memory reference
5380 if the array holds @code{-1}.
5382 The non-strict variant is used in other passes. It must be defined to
5383 accept all pseudo-registers in every context where some kind of
5384 register is required.
5386 Normally, constant addresses which are the sum of a @code{symbol_ref}
5387 and an integer are stored inside a @code{const} RTX to mark them as
5388 constant. Therefore, there is no need to recognize such sums
5389 specifically as legitimate addresses. Normally you would simply
5390 recognize any @code{const} as legitimate.
5392 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5393 sums that are not marked with @code{const}. It assumes that a naked
5394 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5395 naked constant sums as illegitimate addresses, so that none of them will
5396 be given to @code{PRINT_OPERAND_ADDRESS}.
5398 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5399 On some machines, whether a symbolic address is legitimate depends on
5400 the section that the address refers to. On these machines, define the
5401 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5402 into the @code{symbol_ref}, and then check for it here. When you see a
5403 @code{const}, you will have to look inside it to find the
5404 @code{symbol_ref} in order to determine the section. @xref{Assembler
5407 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5408 Some ports are still using a deprecated legacy substitute for
5409 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5413 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5417 and should @code{goto @var{label}} if the address @var{x} is a valid
5418 address on the target machine for a memory operand of mode @var{mode}.
5420 @findex REG_OK_STRICT
5421 Compiler source files that want to use the strict variant of this
5422 macro define the macro @code{REG_OK_STRICT}. You should use an
5423 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5424 that case and the non-strict variant otherwise.
5426 Using the hook is usually simpler because it limits the number of
5427 files that are recompiled when changes are made.
5430 @defmac TARGET_MEM_CONSTRAINT
5431 A single character to be used instead of the default @code{'m'}
5432 character for general memory addresses. This defines the constraint
5433 letter which matches the memory addresses accepted by
5434 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5435 support new address formats in your back end without changing the
5436 semantics of the @code{'m'} constraint. This is necessary in order to
5437 preserve functionality of inline assembly constructs using the
5438 @code{'m'} constraint.
5441 @defmac FIND_BASE_TERM (@var{x})
5442 A C expression to determine the base term of address @var{x},
5443 or to provide a simplified version of @var{x} from which @file{alias.c}
5444 can easily find the base term. This macro is used in only two places:
5445 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5447 It is always safe for this macro to not be defined. It exists so
5448 that alias analysis can understand machine-dependent addresses.
5450 The typical use of this macro is to handle addresses containing
5451 a label_ref or symbol_ref within an UNSPEC@.
5454 @hook TARGET_LEGITIMIZE_ADDRESS
5455 This hook is given an invalid memory address @var{x} for an
5456 operand of mode @var{mode} and should try to return a valid memory
5459 @findex break_out_memory_refs
5460 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5461 and @var{oldx} will be the operand that was given to that function to produce
5464 The code of the hook should not alter the substructure of
5465 @var{x}. If it transforms @var{x} into a more legitimate form, it
5466 should return the new @var{x}.
5468 It is not necessary for this hook to come up with a legitimate address.
5469 The compiler has standard ways of doing so in all cases. In fact, it
5470 is safe to omit this hook or make it return @var{x} if it cannot find
5471 a valid way to legitimize the address. But often a machine-dependent
5472 strategy can generate better code.
5475 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5476 A C compound statement that attempts to replace @var{x}, which is an address
5477 that needs reloading, with a valid memory address for an operand of mode
5478 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5479 It is not necessary to define this macro, but it might be useful for
5480 performance reasons.
5482 For example, on the i386, it is sometimes possible to use a single
5483 reload register instead of two by reloading a sum of two pseudo
5484 registers into a register. On the other hand, for number of RISC
5485 processors offsets are limited so that often an intermediate address
5486 needs to be generated in order to address a stack slot. By defining
5487 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5488 generated for adjacent some stack slots can be made identical, and thus
5491 @emph{Note}: This macro should be used with caution. It is necessary
5492 to know something of how reload works in order to effectively use this,
5493 and it is quite easy to produce macros that build in too much knowledge
5494 of reload internals.
5496 @emph{Note}: This macro must be able to reload an address created by a
5497 previous invocation of this macro. If it fails to handle such addresses
5498 then the compiler may generate incorrect code or abort.
5501 The macro definition should use @code{push_reload} to indicate parts that
5502 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5503 suitable to be passed unaltered to @code{push_reload}.
5505 The code generated by this macro must not alter the substructure of
5506 @var{x}. If it transforms @var{x} into a more legitimate form, it
5507 should assign @var{x} (which will always be a C variable) a new value.
5508 This also applies to parts that you change indirectly by calling
5511 @findex strict_memory_address_p
5512 The macro definition may use @code{strict_memory_address_p} to test if
5513 the address has become legitimate.
5516 If you want to change only a part of @var{x}, one standard way of doing
5517 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5518 single level of rtl. Thus, if the part to be changed is not at the
5519 top level, you'll need to replace first the top level.
5520 It is not necessary for this macro to come up with a legitimate
5521 address; but often a machine-dependent strategy can generate better code.
5524 @hook TARGET_MODE_DEPENDENT_ADDRESS_P
5525 This hook returns @code{true} if memory address @var{addr} can have
5526 different meanings depending on the machine mode of the memory
5527 reference it is used for or if the address is valid for some modes
5530 Autoincrement and autodecrement addresses typically have mode-dependent
5531 effects because the amount of the increment or decrement is the size
5532 of the operand being addressed. Some machines have other mode-dependent
5533 addresses. Many RISC machines have no mode-dependent addresses.
5535 You may assume that @var{addr} is a valid address for the machine.
5537 The default version of this hook returns @code{false}.
5540 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5541 A C statement or compound statement with a conditional @code{goto
5542 @var{label};} executed if memory address @var{x} (an RTX) can have
5543 different meanings depending on the machine mode of the memory
5544 reference it is used for or if the address is valid for some modes
5547 Autoincrement and autodecrement addresses typically have mode-dependent
5548 effects because the amount of the increment or decrement is the size
5549 of the operand being addressed. Some machines have other mode-dependent
5550 addresses. Many RISC machines have no mode-dependent addresses.
5552 You may assume that @var{addr} is a valid address for the machine.
5554 These are obsolete macros, replaced by the
5555 @code{TARGET_MODE_DEPENDENT_ADDRESS_P} target hook.
5558 @hook TARGET_LEGITIMATE_CONSTANT_P
5559 This hook returns true if @var{x} is a legitimate constant for a
5560 @var{mode}-mode immediate operand on the target machine. You can assume that
5561 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5563 The default definition returns true.
5566 @hook TARGET_DELEGITIMIZE_ADDRESS
5567 This hook is used to undo the possibly obfuscating effects of the
5568 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5569 macros. Some backend implementations of these macros wrap symbol
5570 references inside an @code{UNSPEC} rtx to represent PIC or similar
5571 addressing modes. This target hook allows GCC's optimizers to understand
5572 the semantics of these opaque @code{UNSPEC}s by converting them back
5573 into their original form.
5576 @hook TARGET_CONST_NOT_OK_FOR_DEBUG_P
5577 This hook should return true if @var{x} should not be emitted into
5581 @hook TARGET_CANNOT_FORCE_CONST_MEM
5582 This hook should return true if @var{x} is of a form that cannot (or
5583 should not) be spilled to the constant pool. @var{mode} is the mode
5586 The default version of this hook returns false.
5588 The primary reason to define this hook is to prevent reload from
5589 deciding that a non-legitimate constant would be better reloaded
5590 from the constant pool instead of spilling and reloading a register
5591 holding the constant. This restriction is often true of addresses
5592 of TLS symbols for various targets.
5595 @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
5596 This hook should return true if pool entries for constant @var{x} can
5597 be placed in an @code{object_block} structure. @var{mode} is the mode
5600 The default version returns false for all constants.
5603 @hook TARGET_BUILTIN_RECIPROCAL
5604 This hook should return the DECL of a function that implements reciprocal of
5605 the builtin function with builtin function code @var{fn}, or
5606 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5607 when @var{fn} is a code of a machine-dependent builtin function. When
5608 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5609 of a square root function are performed, and only reciprocals of @code{sqrt}
5613 @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
5614 This hook should return the DECL of a function @var{f} that given an
5615 address @var{addr} as an argument returns a mask @var{m} that can be
5616 used to extract from two vectors the relevant data that resides in
5617 @var{addr} in case @var{addr} is not properly aligned.
5619 The autovectorizer, when vectorizing a load operation from an address
5620 @var{addr} that may be unaligned, will generate two vector loads from
5621 the two aligned addresses around @var{addr}. It then generates a
5622 @code{REALIGN_LOAD} operation to extract the relevant data from the
5623 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5624 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5625 the third argument, @var{OFF}, defines how the data will be extracted
5626 from these two vectors: if @var{OFF} is 0, then the returned vector is
5627 @var{v2}; otherwise, the returned vector is composed from the last
5628 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5629 @var{OFF} elements of @var{v2}.
5631 If this hook is defined, the autovectorizer will generate a call
5632 to @var{f} (using the DECL tree that this hook returns) and will
5633 use the return value of @var{f} as the argument @var{OFF} to
5634 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5635 should comply with the semantics expected by @code{REALIGN_LOAD}
5637 If this hook is not defined, then @var{addr} will be used as
5638 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5639 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5642 @hook TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN
5643 This hook should return the DECL of a function @var{f} that implements
5644 widening multiplication of the even elements of two input vectors of type @var{x}.
5646 If this hook is defined, the autovectorizer will use it along with the
5647 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5648 widening multiplication in cases that the order of the results does not have to be
5649 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5650 @code{widen_mult_hi/lo} idioms will be used.
5653 @hook TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD
5654 This hook should return the DECL of a function @var{f} that implements
5655 widening multiplication of the odd elements of two input vectors of type @var{x}.
5657 If this hook is defined, the autovectorizer will use it along with the
5658 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5659 widening multiplication in cases that the order of the results does not have to be
5660 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5661 @code{widen_mult_hi/lo} idioms will be used.
5664 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
5665 Returns cost of different scalar or vector statements for vectorization cost model.
5666 For vector memory operations the cost may depend on type (@var{vectype}) and
5667 misalignment value (@var{misalign}).
5670 @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
5671 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5674 @hook TARGET_VECTORIZE_VEC_PERM_CONST_OK
5675 Return true if a vector created for @code{vec_perm_const} is valid.
5678 @hook TARGET_VECTORIZE_BUILTIN_CONVERSION
5679 This hook should return the DECL of a function that implements conversion of the
5680 input vector of type @var{src_type} to type @var{dest_type}.
5681 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5682 specifies how the conversion is to be applied
5683 (truncation, rounding, etc.).
5685 If this hook is defined, the autovectorizer will use the
5686 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5687 conversion. Otherwise, it will return @code{NULL_TREE}.
5690 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
5691 This hook should return the decl of a function that implements the
5692 vectorized variant of the builtin function with builtin function code
5693 @var{code} or @code{NULL_TREE} if such a function is not available.
5694 The value of @var{fndecl} is the builtin function declaration. The
5695 return type of the vectorized function shall be of vector type
5696 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5699 @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
5700 This hook should return true if the target supports misaligned vector
5701 store/load of a specific factor denoted in the @var{misalignment}
5702 parameter. The vector store/load should be of machine mode @var{mode} and
5703 the elements in the vectors should be of type @var{type}. @var{is_packed}
5704 parameter is true if the memory access is defined in a packed struct.
5707 @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
5708 This hook should return the preferred mode for vectorizing scalar
5709 mode @var{mode}. The default is
5710 equal to @code{word_mode}, because the vectorizer can do some
5711 transformations even in absence of specialized @acronym{SIMD} hardware.
5714 @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES
5715 This hook should return a mask of sizes that should be iterated over
5716 after trying to autovectorize using the vector size derived from the
5717 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5718 The default is zero which means to not iterate over other vector sizes.
5721 @hook TARGET_VECTORIZE_BUILTIN_TM_LOAD
5723 @hook TARGET_VECTORIZE_BUILTIN_TM_STORE
5725 @hook TARGET_VECTORIZE_BUILTIN_GATHER
5726 Target builtin that implements vector gather operation. @var{mem_vectype}
5727 is the vector type of the load and @var{index_type} is scalar type of
5728 the index, scaled by @var{scale}.
5729 The default is @code{NULL_TREE} which means to not vectorize gather
5733 @node Anchored Addresses
5734 @section Anchored Addresses
5735 @cindex anchored addresses
5736 @cindex @option{-fsection-anchors}
5738 GCC usually addresses every static object as a separate entity.
5739 For example, if we have:
5743 int foo (void) @{ return a + b + c; @}
5746 the code for @code{foo} will usually calculate three separate symbolic
5747 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5748 it would be better to calculate just one symbolic address and access
5749 the three variables relative to it. The equivalent pseudocode would
5755 register int *xr = &x;
5756 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5760 (which isn't valid C). We refer to shared addresses like @code{x} as
5761 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5763 The hooks below describe the target properties that GCC needs to know
5764 in order to make effective use of section anchors. It won't use
5765 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5766 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5768 @hook TARGET_MIN_ANCHOR_OFFSET
5769 The minimum offset that should be applied to a section anchor.
5770 On most targets, it should be the smallest offset that can be
5771 applied to a base register while still giving a legitimate address
5772 for every mode. The default value is 0.
5775 @hook TARGET_MAX_ANCHOR_OFFSET
5776 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5777 offset that should be applied to section anchors. The default
5781 @hook TARGET_ASM_OUTPUT_ANCHOR
5782 Write the assembly code to define section anchor @var{x}, which is a
5783 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5784 The hook is called with the assembly output position set to the beginning
5785 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5787 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5788 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5789 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5790 is @code{NULL}, which disables the use of section anchors altogether.
5793 @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
5794 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5795 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5796 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5798 The default version is correct for most targets, but you might need to
5799 intercept this hook to handle things like target-specific attributes
5800 or target-specific sections.
5803 @node Condition Code
5804 @section Condition Code Status
5805 @cindex condition code status
5807 The macros in this section can be split in two families, according to the
5808 two ways of representing condition codes in GCC.
5810 The first representation is the so called @code{(cc0)} representation
5811 (@pxref{Jump Patterns}), where all instructions can have an implicit
5812 clobber of the condition codes. The second is the condition code
5813 register representation, which provides better schedulability for
5814 architectures that do have a condition code register, but on which
5815 most instructions do not affect it. The latter category includes
5818 The implicit clobbering poses a strong restriction on the placement of
5819 the definition and use of the condition code, which need to be in adjacent
5820 insns for machines using @code{(cc0)}. This can prevent important
5821 optimizations on some machines. For example, on the IBM RS/6000, there
5822 is a delay for taken branches unless the condition code register is set
5823 three instructions earlier than the conditional branch. The instruction
5824 scheduler cannot perform this optimization if it is not permitted to
5825 separate the definition and use of the condition code register.
5827 For this reason, it is possible and suggested to use a register to
5828 represent the condition code for new ports. If there is a specific
5829 condition code register in the machine, use a hard register. If the
5830 condition code or comparison result can be placed in any general register,
5831 or if there are multiple condition registers, use a pseudo register.
5832 Registers used to store the condition code value will usually have a mode
5833 that is in class @code{MODE_CC}.
5835 Alternatively, you can use @code{BImode} if the comparison operator is
5836 specified already in the compare instruction. In this case, you are not
5837 interested in most macros in this section.
5840 * CC0 Condition Codes:: Old style representation of condition codes.
5841 * MODE_CC Condition Codes:: Modern representation of condition codes.
5842 * Cond Exec Macros:: Macros to control conditional execution.
5845 @node CC0 Condition Codes
5846 @subsection Representation of condition codes using @code{(cc0)}
5850 The file @file{conditions.h} defines a variable @code{cc_status} to
5851 describe how the condition code was computed (in case the interpretation of
5852 the condition code depends on the instruction that it was set by). This
5853 variable contains the RTL expressions on which the condition code is
5854 currently based, and several standard flags.
5856 Sometimes additional machine-specific flags must be defined in the machine
5857 description header file. It can also add additional machine-specific
5858 information by defining @code{CC_STATUS_MDEP}.
5860 @defmac CC_STATUS_MDEP
5861 C code for a data type which is used for declaring the @code{mdep}
5862 component of @code{cc_status}. It defaults to @code{int}.
5864 This macro is not used on machines that do not use @code{cc0}.
5867 @defmac CC_STATUS_MDEP_INIT
5868 A C expression to initialize the @code{mdep} field to ``empty''.
5869 The default definition does nothing, since most machines don't use
5870 the field anyway. If you want to use the field, you should probably
5871 define this macro to initialize it.
5873 This macro is not used on machines that do not use @code{cc0}.
5876 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5877 A C compound statement to set the components of @code{cc_status}
5878 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5879 this macro's responsibility to recognize insns that set the condition
5880 code as a byproduct of other activity as well as those that explicitly
5883 This macro is not used on machines that do not use @code{cc0}.
5885 If there are insns that do not set the condition code but do alter
5886 other machine registers, this macro must check to see whether they
5887 invalidate the expressions that the condition code is recorded as
5888 reflecting. For example, on the 68000, insns that store in address
5889 registers do not set the condition code, which means that usually
5890 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5891 insns. But suppose that the previous insn set the condition code
5892 based on location @samp{a4@@(102)} and the current insn stores a new
5893 value in @samp{a4}. Although the condition code is not changed by
5894 this, it will no longer be true that it reflects the contents of
5895 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5896 @code{cc_status} in this case to say that nothing is known about the
5897 condition code value.
5899 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5900 with the results of peephole optimization: insns whose patterns are
5901 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5902 constants which are just the operands. The RTL structure of these
5903 insns is not sufficient to indicate what the insns actually do. What
5904 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5905 @code{CC_STATUS_INIT}.
5907 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5908 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5909 @samp{cc}. This avoids having detailed information about patterns in
5910 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5913 @node MODE_CC Condition Codes
5914 @subsection Representation of condition codes using registers
5918 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5919 On many machines, the condition code may be produced by other instructions
5920 than compares, for example the branch can use directly the condition
5921 code set by a subtract instruction. However, on some machines
5922 when the condition code is set this way some bits (such as the overflow
5923 bit) are not set in the same way as a test instruction, so that a different
5924 branch instruction must be used for some conditional branches. When
5925 this happens, use the machine mode of the condition code register to
5926 record different formats of the condition code register. Modes can
5927 also be used to record which compare instruction (e.g. a signed or an
5928 unsigned comparison) produced the condition codes.
5930 If other modes than @code{CCmode} are required, add them to
5931 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
5932 a mode given an operand of a compare. This is needed because the modes
5933 have to be chosen not only during RTL generation but also, for example,
5934 by instruction combination. The result of @code{SELECT_CC_MODE} should
5935 be consistent with the mode used in the patterns; for example to support
5936 the case of the add on the SPARC discussed above, we have the pattern
5940 [(set (reg:CC_NOOV 0)
5942 (plus:SI (match_operand:SI 0 "register_operand" "%r")
5943 (match_operand:SI 1 "arith_operand" "rI"))
5950 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
5951 for comparisons whose argument is a @code{plus}:
5954 #define SELECT_CC_MODE(OP,X,Y) \
5955 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5956 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5957 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5958 || GET_CODE (X) == NEG) \
5959 ? CC_NOOVmode : CCmode))
5962 Another reason to use modes is to retain information on which operands
5963 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
5966 You should define this macro if and only if you define extra CC modes
5967 in @file{@var{machine}-modes.def}.
5970 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5971 On some machines not all possible comparisons are defined, but you can
5972 convert an invalid comparison into a valid one. For example, the Alpha
5973 does not have a @code{GT} comparison, but you can use an @code{LT}
5974 comparison instead and swap the order of the operands.
5976 On such machines, define this macro to be a C statement to do any
5977 required conversions. @var{code} is the initial comparison code
5978 and @var{op0} and @var{op1} are the left and right operands of the
5979 comparison, respectively. You should modify @var{code}, @var{op0}, and
5980 @var{op1} as required.
5982 GCC will not assume that the comparison resulting from this macro is
5983 valid but will see if the resulting insn matches a pattern in the
5986 You need not define this macro if it would never change the comparison
5990 @defmac REVERSIBLE_CC_MODE (@var{mode})
5991 A C expression whose value is one if it is always safe to reverse a
5992 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5993 can ever return @var{mode} for a floating-point inequality comparison,
5994 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5996 You need not define this macro if it would always returns zero or if the
5997 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5998 For example, here is the definition used on the SPARC, where floating-point
5999 inequality comparisons are always given @code{CCFPEmode}:
6002 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
6006 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6007 A C expression whose value is reversed condition code of the @var{code} for
6008 comparison done in CC_MODE @var{mode}. The macro is used only in case
6009 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6010 machine has some non-standard way how to reverse certain conditionals. For
6011 instance in case all floating point conditions are non-trapping, compiler may
6012 freely convert unordered compares to ordered one. Then definition may look
6016 #define REVERSE_CONDITION(CODE, MODE) \
6017 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6018 : reverse_condition_maybe_unordered (CODE))
6022 @hook TARGET_FIXED_CONDITION_CODE_REGS
6023 On targets which do not use @code{(cc0)}, and which use a hard
6024 register rather than a pseudo-register to hold condition codes, the
6025 regular CSE passes are often not able to identify cases in which the
6026 hard register is set to a common value. Use this hook to enable a
6027 small pass which optimizes such cases. This hook should return true
6028 to enable this pass, and it should set the integers to which its
6029 arguments point to the hard register numbers used for condition codes.
6030 When there is only one such register, as is true on most systems, the
6031 integer pointed to by @var{p2} should be set to
6032 @code{INVALID_REGNUM}.
6034 The default version of this hook returns false.
6037 @hook TARGET_CC_MODES_COMPATIBLE
6038 On targets which use multiple condition code modes in class
6039 @code{MODE_CC}, it is sometimes the case that a comparison can be
6040 validly done in more than one mode. On such a system, define this
6041 target hook to take two mode arguments and to return a mode in which
6042 both comparisons may be validly done. If there is no such mode,
6043 return @code{VOIDmode}.
6045 The default version of this hook checks whether the modes are the
6046 same. If they are, it returns that mode. If they are different, it
6047 returns @code{VOIDmode}.
6050 @node Cond Exec Macros
6051 @subsection Macros to control conditional execution
6052 @findex conditional execution
6055 There is one macro that may need to be defined for targets
6056 supporting conditional execution, independent of how they
6057 represent conditional branches.
6059 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6060 A C expression that returns true if the conditional execution predicate
6061 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6062 versa. Define this to return 0 if the target has conditional execution
6063 predicates that cannot be reversed safely. There is no need to validate
6064 that the arguments of op1 and op2 are the same, this is done separately.
6065 If no expansion is specified, this macro is defined as follows:
6068 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6069 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6074 @section Describing Relative Costs of Operations
6075 @cindex costs of instructions
6076 @cindex relative costs
6077 @cindex speed of instructions
6079 These macros let you describe the relative speed of various operations
6080 on the target machine.
6082 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6083 A C expression for the cost of moving data of mode @var{mode} from a
6084 register in class @var{from} to one in class @var{to}. The classes are
6085 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6086 value of 2 is the default; other values are interpreted relative to
6089 It is not required that the cost always equal 2 when @var{from} is the
6090 same as @var{to}; on some machines it is expensive to move between
6091 registers if they are not general registers.
6093 If reload sees an insn consisting of a single @code{set} between two
6094 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6095 classes returns a value of 2, reload does not check to ensure that the
6096 constraints of the insn are met. Setting a cost of other than 2 will
6097 allow reload to verify that the constraints are met. You should do this
6098 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6100 These macros are obsolete, new ports should use the target hook
6101 @code{TARGET_REGISTER_MOVE_COST} instead.
6104 @hook TARGET_REGISTER_MOVE_COST
6105 This target hook should return the cost of moving data of mode @var{mode}
6106 from a register in class @var{from} to one in class @var{to}. The classes
6107 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6108 A value of 2 is the default; other values are interpreted relative to
6111 It is not required that the cost always equal 2 when @var{from} is the
6112 same as @var{to}; on some machines it is expensive to move between
6113 registers if they are not general registers.
6115 If reload sees an insn consisting of a single @code{set} between two
6116 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6117 classes returns a value of 2, reload does not check to ensure that the
6118 constraints of the insn are met. Setting a cost of other than 2 will
6119 allow reload to verify that the constraints are met. You should do this
6120 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6122 The default version of this function returns 2.
6125 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6126 A C expression for the cost of moving data of mode @var{mode} between a
6127 register of class @var{class} and memory; @var{in} is zero if the value
6128 is to be written to memory, nonzero if it is to be read in. This cost
6129 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6130 registers and memory is more expensive than between two registers, you
6131 should define this macro to express the relative cost.
6133 If you do not define this macro, GCC uses a default cost of 4 plus
6134 the cost of copying via a secondary reload register, if one is
6135 needed. If your machine requires a secondary reload register to copy
6136 between memory and a register of @var{class} but the reload mechanism is
6137 more complex than copying via an intermediate, define this macro to
6138 reflect the actual cost of the move.
6140 GCC defines the function @code{memory_move_secondary_cost} if
6141 secondary reloads are needed. It computes the costs due to copying via
6142 a secondary register. If your machine copies from memory using a
6143 secondary register in the conventional way but the default base value of
6144 4 is not correct for your machine, define this macro to add some other
6145 value to the result of that function. The arguments to that function
6146 are the same as to this macro.
6148 These macros are obsolete, new ports should use the target hook
6149 @code{TARGET_MEMORY_MOVE_COST} instead.
6152 @hook TARGET_MEMORY_MOVE_COST
6153 This target hook should return the cost of moving data of mode @var{mode}
6154 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6155 if the value is to be written to memory, @code{true} if it is to be read in.
6156 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6157 If moving between registers and memory is more expensive than between two
6158 registers, you should add this target hook to express the relative cost.
6160 If you do not add this target hook, GCC uses a default cost of 4 plus
6161 the cost of copying via a secondary reload register, if one is
6162 needed. If your machine requires a secondary reload register to copy
6163 between memory and a register of @var{rclass} but the reload mechanism is
6164 more complex than copying via an intermediate, use this target hook to
6165 reflect the actual cost of the move.
6167 GCC defines the function @code{memory_move_secondary_cost} if
6168 secondary reloads are needed. It computes the costs due to copying via
6169 a secondary register. If your machine copies from memory using a
6170 secondary register in the conventional way but the default base value of
6171 4 is not correct for your machine, use this target hook to add some other
6172 value to the result of that function. The arguments to that function
6173 are the same as to this target hook.
6176 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6177 A C expression for the cost of a branch instruction. A value of 1 is
6178 the default; other values are interpreted relative to that. Parameter
6179 @var{speed_p} is true when the branch in question should be optimized
6180 for speed. When it is false, @code{BRANCH_COST} should return a value
6181 optimal for code size rather than performance. @var{predictable_p} is
6182 true for well-predicted branches. On many architectures the
6183 @code{BRANCH_COST} can be reduced then.
6186 Here are additional macros which do not specify precise relative costs,
6187 but only that certain actions are more expensive than GCC would
6190 @defmac SLOW_BYTE_ACCESS
6191 Define this macro as a C expression which is nonzero if accessing less
6192 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6193 faster than accessing a word of memory, i.e., if such access
6194 require more than one instruction or if there is no difference in cost
6195 between byte and (aligned) word loads.
6197 When this macro is not defined, the compiler will access a field by
6198 finding the smallest containing object; when it is defined, a fullword
6199 load will be used if alignment permits. Unless bytes accesses are
6200 faster than word accesses, using word accesses is preferable since it
6201 may eliminate subsequent memory access if subsequent accesses occur to
6202 other fields in the same word of the structure, but to different bytes.
6205 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6206 Define this macro to be the value 1 if memory accesses described by the
6207 @var{mode} and @var{alignment} parameters have a cost many times greater
6208 than aligned accesses, for example if they are emulated in a trap
6211 When this macro is nonzero, the compiler will act as if
6212 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6213 moves. This can cause significantly more instructions to be produced.
6214 Therefore, do not set this macro nonzero if unaligned accesses only add a
6215 cycle or two to the time for a memory access.
6217 If the value of this macro is always zero, it need not be defined. If
6218 this macro is defined, it should produce a nonzero value when
6219 @code{STRICT_ALIGNMENT} is nonzero.
6222 @defmac MOVE_RATIO (@var{speed})
6223 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6224 which a sequence of insns should be generated instead of a
6225 string move insn or a library call. Increasing the value will always
6226 make code faster, but eventually incurs high cost in increased code size.
6228 Note that on machines where the corresponding move insn is a
6229 @code{define_expand} that emits a sequence of insns, this macro counts
6230 the number of such sequences.
6232 The parameter @var{speed} is true if the code is currently being
6233 optimized for speed rather than size.
6235 If you don't define this, a reasonable default is used.
6238 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6239 A C expression used to determine whether @code{move_by_pieces} will be used to
6240 copy a chunk of memory, or whether some other block move mechanism
6241 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6242 than @code{MOVE_RATIO}.
6245 @defmac MOVE_MAX_PIECES
6246 A C expression used by @code{move_by_pieces} to determine the largest unit
6247 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6250 @defmac CLEAR_RATIO (@var{speed})
6251 The threshold of number of scalar move insns, @emph{below} which a sequence
6252 of insns should be generated to clear memory instead of a string clear insn
6253 or a library call. Increasing the value will always make code faster, but
6254 eventually incurs high cost in increased code size.
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 CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6263 A C expression used to determine whether @code{clear_by_pieces} will be used
6264 to clear a chunk of memory, or whether some other block clear mechanism
6265 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6266 than @code{CLEAR_RATIO}.
6269 @defmac SET_RATIO (@var{speed})
6270 The threshold of number of scalar move insns, @emph{below} which a sequence
6271 of insns should be generated to set memory to a constant value, instead of
6272 a block set insn or a library call.
6273 Increasing the value will always make code faster, but
6274 eventually incurs high cost in increased code size.
6276 The parameter @var{speed} is true if the code is currently being
6277 optimized for speed rather than size.
6279 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6282 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6283 A C expression used to determine whether @code{store_by_pieces} will be
6284 used to set a chunk of memory to a constant value, or whether some
6285 other mechanism will be used. Used by @code{__builtin_memset} when
6286 storing values other than constant zero.
6287 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6288 than @code{SET_RATIO}.
6291 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6292 A C expression used to determine whether @code{store_by_pieces} will be
6293 used to set a chunk of memory to a constant string value, or whether some
6294 other mechanism will be used. Used by @code{__builtin_strcpy} when
6295 called with a constant source string.
6296 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6297 than @code{MOVE_RATIO}.
6300 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6301 A C expression used to determine whether a load postincrement is a good
6302 thing to use for a given mode. Defaults to the value of
6303 @code{HAVE_POST_INCREMENT}.
6306 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6307 A C expression used to determine whether a load postdecrement is a good
6308 thing to use for a given mode. Defaults to the value of
6309 @code{HAVE_POST_DECREMENT}.
6312 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6313 A C expression used to determine whether a load preincrement is a good
6314 thing to use for a given mode. Defaults to the value of
6315 @code{HAVE_PRE_INCREMENT}.
6318 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6319 A C expression used to determine whether a load predecrement is a good
6320 thing to use for a given mode. Defaults to the value of
6321 @code{HAVE_PRE_DECREMENT}.
6324 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6325 A C expression used to determine whether a store 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_STORE_POST_DECREMENT (@var{mode})
6331 A C expression used to determine whether a store 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_STORE_PRE_INCREMENT (@var{mode})
6337 This macro is used to determine whether a store 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_STORE_PRE_DECREMENT (@var{mode})
6343 This macro is used to determine whether a store predecrement is a good
6344 thing to use for a given mode. Defaults to the value of
6345 @code{HAVE_PRE_DECREMENT}.
6348 @defmac NO_FUNCTION_CSE
6349 Define this macro if it is as good or better to call a constant
6350 function address than to call an address kept in a register.
6353 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6354 Define this macro if a non-short-circuit operation produced by
6355 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6356 @code{BRANCH_COST} is greater than or equal to the value 2.
6359 @hook TARGET_RTX_COSTS
6360 This target hook describes the relative costs of RTL expressions.
6362 The cost may depend on the precise form of the expression, which is
6363 available for examination in @var{x}, and the fact that @var{x} appears
6364 as operand @var{opno} of an expression with rtx code @var{outer_code}.
6365 That is, the hook can assume that there is some rtx @var{y} such
6366 that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that
6367 either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or
6368 (b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}.
6370 @var{code} is @var{x}'s expression code---redundant, since it can be
6371 obtained with @code{GET_CODE (@var{x})}.
6373 In implementing this hook, you can use the construct
6374 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6377 On entry to the hook, @code{*@var{total}} contains a default estimate
6378 for the cost of the expression. The hook should modify this value as
6379 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6380 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6381 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6383 When optimizing for code size, i.e.@: when @code{speed} is
6384 false, this target hook should be used to estimate the relative
6385 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6387 The hook returns true when all subexpressions of @var{x} have been
6388 processed, and false when @code{rtx_cost} should recurse.
6391 @hook TARGET_ADDRESS_COST
6392 This hook computes the cost of an addressing mode that contains
6393 @var{address}. If not defined, the cost is computed from
6394 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6396 For most CISC machines, the default cost is a good approximation of the
6397 true cost of the addressing mode. However, on RISC machines, all
6398 instructions normally have the same length and execution time. Hence
6399 all addresses will have equal costs.
6401 In cases where more than one form of an address is known, the form with
6402 the lowest cost will be used. If multiple forms have the same, lowest,
6403 cost, the one that is the most complex will be used.
6405 For example, suppose an address that is equal to the sum of a register
6406 and a constant is used twice in the same basic block. When this macro
6407 is not defined, the address will be computed in a register and memory
6408 references will be indirect through that register. On machines where
6409 the cost of the addressing mode containing the sum is no higher than
6410 that of a simple indirect reference, this will produce an additional
6411 instruction and possibly require an additional register. Proper
6412 specification of this macro eliminates this overhead for such machines.
6414 This hook is never called with an invalid address.
6416 On machines where an address involving more than one register is as
6417 cheap as an address computation involving only one register, defining
6418 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6419 be live over a region of code where only one would have been if
6420 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6421 should be considered in the definition of this macro. Equivalent costs
6422 should probably only be given to addresses with different numbers of
6423 registers on machines with lots of registers.
6427 @section Adjusting the Instruction Scheduler
6429 The instruction scheduler may need a fair amount of machine-specific
6430 adjustment in order to produce good code. GCC provides several target
6431 hooks for this purpose. It is usually enough to define just a few of
6432 them: try the first ones in this list first.
6434 @hook TARGET_SCHED_ISSUE_RATE
6435 This hook returns the maximum number of instructions that can ever
6436 issue at the same time on the target machine. The default is one.
6437 Although the insn scheduler can define itself the possibility of issue
6438 an insn on the same cycle, the value can serve as an additional
6439 constraint to issue insns on the same simulated processor cycle (see
6440 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6441 This value must be constant over the entire compilation. If you need
6442 it to vary depending on what the instructions are, you must use
6443 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6446 @hook TARGET_SCHED_VARIABLE_ISSUE
6447 This hook is executed by the scheduler after it has scheduled an insn
6448 from the ready list. It should return the number of insns which can
6449 still be issued in the current cycle. The default is
6450 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6451 @code{USE}, which normally are not counted against the issue rate.
6452 You should define this hook if some insns take more machine resources
6453 than others, so that fewer insns can follow them in the same cycle.
6454 @var{file} is either a null pointer, or a stdio stream to write any
6455 debug output to. @var{verbose} is the verbose level provided by
6456 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6460 @hook TARGET_SCHED_ADJUST_COST
6461 This function corrects the value of @var{cost} based on the
6462 relationship between @var{insn} and @var{dep_insn} through the
6463 dependence @var{link}. It should return the new value. The default
6464 is to make no adjustment to @var{cost}. This can be used for example
6465 to specify to the scheduler using the traditional pipeline description
6466 that an output- or anti-dependence does not incur the same cost as a
6467 data-dependence. If the scheduler using the automaton based pipeline
6468 description, the cost of anti-dependence is zero and the cost of
6469 output-dependence is maximum of one and the difference of latency
6470 times of the first and the second insns. If these values are not
6471 acceptable, you could use the hook to modify them too. See also
6472 @pxref{Processor pipeline description}.
6475 @hook TARGET_SCHED_ADJUST_PRIORITY
6476 This hook adjusts the integer scheduling priority @var{priority} of
6477 @var{insn}. It should return the new priority. Increase the priority to
6478 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6479 later. Do not define this hook if you do not need to adjust the
6480 scheduling priorities of insns.
6483 @hook TARGET_SCHED_REORDER
6484 This hook is executed by the scheduler after it has scheduled the ready
6485 list, to allow the machine description to reorder it (for example to
6486 combine two small instructions together on @samp{VLIW} machines).
6487 @var{file} is either a null pointer, or a stdio stream to write any
6488 debug output to. @var{verbose} is the verbose level provided by
6489 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6490 list of instructions that are ready to be scheduled. @var{n_readyp} is
6491 a pointer to the number of elements in the ready list. The scheduler
6492 reads the ready list in reverse order, starting with
6493 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6494 is the timer tick of the scheduler. You may modify the ready list and
6495 the number of ready insns. The return value is the number of insns that
6496 can issue this cycle; normally this is just @code{issue_rate}. See also
6497 @samp{TARGET_SCHED_REORDER2}.
6500 @hook TARGET_SCHED_REORDER2
6501 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6502 function is called whenever the scheduler starts a new cycle. This one
6503 is called once per iteration over a cycle, immediately after
6504 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6505 return the number of insns to be scheduled in the same cycle. Defining
6506 this hook can be useful if there are frequent situations where
6507 scheduling one insn causes other insns to become ready in the same
6508 cycle. These other insns can then be taken into account properly.
6511 @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
6512 This hook is called after evaluation forward dependencies of insns in
6513 chain given by two parameter values (@var{head} and @var{tail}
6514 correspondingly) but before insns scheduling of the insn chain. For
6515 example, it can be used for better insn classification if it requires
6516 analysis of dependencies. This hook can use backward and forward
6517 dependencies of the insn scheduler because they are already
6521 @hook TARGET_SCHED_INIT
6522 This hook is executed by the scheduler at the beginning of each block of
6523 instructions that are to be scheduled. @var{file} is either a null
6524 pointer, or a stdio stream to write any debug output to. @var{verbose}
6525 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6526 @var{max_ready} is the maximum number of insns in the current scheduling
6527 region that can be live at the same time. This can be used to allocate
6528 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6531 @hook TARGET_SCHED_FINISH
6532 This hook is executed by the scheduler at the end of each block of
6533 instructions that are to be scheduled. It can be used to perform
6534 cleanup of any actions done by the other scheduling hooks. @var{file}
6535 is either a null pointer, or a stdio stream to write any debug output
6536 to. @var{verbose} is the verbose level provided by
6537 @option{-fsched-verbose-@var{n}}.
6540 @hook TARGET_SCHED_INIT_GLOBAL
6541 This hook is executed by the scheduler after function level initializations.
6542 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6543 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6544 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6547 @hook TARGET_SCHED_FINISH_GLOBAL
6548 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6549 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6550 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6553 @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
6554 The hook returns an RTL insn. The automaton state used in the
6555 pipeline hazard recognizer is changed as if the insn were scheduled
6556 when the new simulated processor cycle starts. Usage of the hook may
6557 simplify the automaton pipeline description for some @acronym{VLIW}
6558 processors. If the hook is defined, it is used only for the automaton
6559 based pipeline description. The default is not to change the state
6560 when the new simulated processor cycle starts.
6563 @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
6564 The hook can be used to initialize data used by the previous hook.
6567 @hook TARGET_SCHED_DFA_POST_CYCLE_INSN
6568 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6569 to changed the state as if the insn were scheduled when the new
6570 simulated processor cycle finishes.
6573 @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
6574 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6575 used to initialize data used by the previous hook.
6578 @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
6579 The hook to notify target that the current simulated cycle is about to finish.
6580 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6581 to change the state in more complicated situations - e.g., when advancing
6582 state on a single insn is not enough.
6585 @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
6586 The hook to notify target that new simulated cycle has just started.
6587 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6588 to change the state in more complicated situations - e.g., when advancing
6589 state on a single insn is not enough.
6592 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
6593 This hook controls better choosing an insn from the ready insn queue
6594 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6595 chooses the first insn from the queue. If the hook returns a positive
6596 value, an additional scheduler code tries all permutations of
6597 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6598 subsequent ready insns to choose an insn whose issue will result in
6599 maximal number of issued insns on the same cycle. For the
6600 @acronym{VLIW} processor, the code could actually solve the problem of
6601 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6602 rules of @acronym{VLIW} packing are described in the automaton.
6604 This code also could be used for superscalar @acronym{RISC}
6605 processors. Let us consider a superscalar @acronym{RISC} processor
6606 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6607 @var{B}, some insns can be executed only in pipelines @var{B} or
6608 @var{C}, and one insn can be executed in pipeline @var{B}. The
6609 processor may issue the 1st insn into @var{A} and the 2nd one into
6610 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6611 until the next cycle. If the scheduler issues the 3rd insn the first,
6612 the processor could issue all 3 insns per cycle.
6614 Actually this code demonstrates advantages of the automaton based
6615 pipeline hazard recognizer. We try quickly and easy many insn
6616 schedules to choose the best one.
6618 The default is no multipass scheduling.
6621 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
6623 This hook controls what insns from the ready insn queue will be
6624 considered for the multipass insn scheduling. If the hook returns
6625 zero for @var{insn}, the insn will be not chosen to
6628 The default is that any ready insns can be chosen to be issued.
6631 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
6632 This hook prepares the target backend for a new round of multipass
6636 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
6637 This hook is called when multipass scheduling evaluates instruction INSN.
6640 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
6641 This is called when multipass scheduling backtracks from evaluation of
6645 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
6646 This hook notifies the target about the result of the concluded current
6647 round of multipass scheduling.
6650 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
6651 This hook initializes target-specific data used in multipass scheduling.
6654 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
6655 This hook finalizes target-specific data used in multipass scheduling.
6658 @hook TARGET_SCHED_DFA_NEW_CYCLE
6659 This hook is called by the insn scheduler before issuing @var{insn}
6660 on cycle @var{clock}. If the hook returns nonzero,
6661 @var{insn} is not issued on this processor cycle. Instead,
6662 the processor cycle is advanced. If *@var{sort_p}
6663 is zero, the insn ready queue is not sorted on the new cycle
6664 start as usually. @var{dump} and @var{verbose} specify the file and
6665 verbosity level to use for debugging output.
6666 @var{last_clock} and @var{clock} are, respectively, the
6667 processor cycle on which the previous insn has been issued,
6668 and the current processor cycle.
6671 @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
6672 This hook is used to define which dependences are considered costly by
6673 the target, so costly that it is not advisable to schedule the insns that
6674 are involved in the dependence too close to one another. The parameters
6675 to this hook are as follows: The first parameter @var{_dep} is the dependence
6676 being evaluated. The second parameter @var{cost} is the cost of the
6677 dependence as estimated by the scheduler, and the third
6678 parameter @var{distance} is the distance in cycles between the two insns.
6679 The hook returns @code{true} if considering the distance between the two
6680 insns the dependence between them is considered costly by the target,
6681 and @code{false} otherwise.
6683 Defining this hook can be useful in multiple-issue out-of-order machines,
6684 where (a) it's practically hopeless to predict the actual data/resource
6685 delays, however: (b) there's a better chance to predict the actual grouping
6686 that will be formed, and (c) correctly emulating the grouping can be very
6687 important. In such targets one may want to allow issuing dependent insns
6688 closer to one another---i.e., closer than the dependence distance; however,
6689 not in cases of ``costly dependences'', which this hooks allows to define.
6692 @hook TARGET_SCHED_H_I_D_EXTENDED
6693 This hook is called by the insn scheduler after emitting a new instruction to
6694 the instruction stream. The hook notifies a target backend to extend its
6695 per instruction data structures.
6698 @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
6699 Return a pointer to a store large enough to hold target scheduling context.
6702 @hook TARGET_SCHED_INIT_SCHED_CONTEXT
6703 Initialize store pointed to by @var{tc} to hold target scheduling context.
6704 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6705 beginning of the block. Otherwise, copy the current context into @var{tc}.
6708 @hook TARGET_SCHED_SET_SCHED_CONTEXT
6709 Copy target scheduling context pointed to by @var{tc} to the current context.
6712 @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
6713 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6716 @hook TARGET_SCHED_FREE_SCHED_CONTEXT
6717 Deallocate a store for target scheduling context pointed to by @var{tc}.
6720 @hook TARGET_SCHED_SPECULATE_INSN
6721 This hook is called by the insn scheduler when @var{insn} has only
6722 speculative dependencies and therefore can be scheduled speculatively.
6723 The hook is used to check if the pattern of @var{insn} has a speculative
6724 version and, in case of successful check, to generate that speculative
6725 pattern. The hook should return 1, if the instruction has a speculative form,
6726 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6727 speculation. If the return value equals 1 then @var{new_pat} is assigned
6728 the generated speculative pattern.
6731 @hook TARGET_SCHED_NEEDS_BLOCK_P
6732 This hook is called by the insn scheduler during generation of recovery code
6733 for @var{insn}. It should return @code{true}, if the corresponding check
6734 instruction should branch to recovery code, or @code{false} otherwise.
6737 @hook TARGET_SCHED_GEN_SPEC_CHECK
6738 This hook is called by the insn scheduler to generate a pattern for recovery
6739 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6740 speculative instruction for which the check should be generated.
6741 @var{label} is either a label of a basic block, where recovery code should
6742 be emitted, or a null pointer, when requested check doesn't branch to
6743 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6744 a pattern for a branchy check corresponding to a simple check denoted by
6745 @var{insn} should be generated. In this case @var{label} can't be null.
6748 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC
6749 This hook is used as a workaround for
6750 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6751 called on the first instruction of the ready list. The hook is used to
6752 discard speculative instructions that stand first in the ready list from
6753 being scheduled on the current cycle. If the hook returns @code{false},
6754 @var{insn} will not be chosen to be issued.
6755 For non-speculative instructions,
6756 the hook should always return @code{true}. For example, in the ia64 backend
6757 the hook is used to cancel data speculative insns when the ALAT table
6761 @hook TARGET_SCHED_SET_SCHED_FLAGS
6762 This hook is used by the insn scheduler to find out what features should be
6764 The structure *@var{spec_info} should be filled in by the target.
6765 The structure describes speculation types that can be used in the scheduler.
6768 @hook TARGET_SCHED_SMS_RES_MII
6769 This hook is called by the swing modulo scheduler to calculate a
6770 resource-based lower bound which is based on the resources available in
6771 the machine and the resources required by each instruction. The target
6772 backend can use @var{g} to calculate such bound. A very simple lower
6773 bound will be used in case this hook is not implemented: the total number
6774 of instructions divided by the issue rate.
6777 @hook TARGET_SCHED_DISPATCH
6778 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6779 is supported in hardware and the condition specified in the parameter is true.
6782 @hook TARGET_SCHED_DISPATCH_DO
6783 This hook is called by Haifa Scheduler. It performs the operation specified
6784 in its second parameter.
6787 @hook TARGET_SCHED_EXPOSED_PIPELINE
6789 @hook TARGET_SCHED_REASSOCIATION_WIDTH
6792 @section Dividing the Output into Sections (Texts, Data, @dots{})
6793 @c the above section title is WAY too long. maybe cut the part between
6794 @c the (...)? --mew 10feb93
6796 An object file is divided into sections containing different types of
6797 data. In the most common case, there are three sections: the @dfn{text
6798 section}, which holds instructions and read-only data; the @dfn{data
6799 section}, which holds initialized writable data; and the @dfn{bss
6800 section}, which holds uninitialized data. Some systems have other kinds
6803 @file{varasm.c} provides several well-known sections, such as
6804 @code{text_section}, @code{data_section} and @code{bss_section}.
6805 The normal way of controlling a @code{@var{foo}_section} variable
6806 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6807 as described below. The macros are only read once, when @file{varasm.c}
6808 initializes itself, so their values must be run-time constants.
6809 They may however depend on command-line flags.
6811 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6812 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6813 to be string literals.
6815 Some assemblers require a different string to be written every time a
6816 section is selected. If your assembler falls into this category, you
6817 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6818 @code{get_unnamed_section} to set up the sections.
6820 You must always create a @code{text_section}, either by defining
6821 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6822 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6823 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6824 create a distinct @code{readonly_data_section}, the default is to
6825 reuse @code{text_section}.
6827 All the other @file{varasm.c} sections are optional, and are null
6828 if the target does not provide them.
6830 @defmac TEXT_SECTION_ASM_OP
6831 A C expression whose value is a string, including spacing, containing the
6832 assembler operation that should precede instructions and read-only data.
6833 Normally @code{"\t.text"} is right.
6836 @defmac HOT_TEXT_SECTION_NAME
6837 If defined, a C string constant for the name of the section containing most
6838 frequently executed functions of the program. If not defined, GCC will provide
6839 a default definition if the target supports named sections.
6842 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6843 If defined, a C string constant for the name of the section containing unlikely
6844 executed functions in the program.
6847 @defmac DATA_SECTION_ASM_OP
6848 A C expression whose value is a string, including spacing, containing the
6849 assembler operation to identify the following data as writable initialized
6850 data. Normally @code{"\t.data"} is right.
6853 @defmac SDATA_SECTION_ASM_OP
6854 If defined, a C expression whose value is a string, including spacing,
6855 containing the assembler operation to identify the following data as
6856 initialized, writable small data.
6859 @defmac READONLY_DATA_SECTION_ASM_OP
6860 A C expression whose value is a string, including spacing, containing the
6861 assembler operation to identify the following data as read-only initialized
6865 @defmac BSS_SECTION_ASM_OP
6866 If defined, a C expression whose value is a string, including spacing,
6867 containing the assembler operation to identify the following data as
6868 uninitialized global data. If not defined, and
6869 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
6870 uninitialized global data will be output in the data section if
6871 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6875 @defmac SBSS_SECTION_ASM_OP
6876 If defined, a C expression whose value is a string, including spacing,
6877 containing the assembler operation to identify the following data as
6878 uninitialized, writable small data.
6881 @defmac TLS_COMMON_ASM_OP
6882 If defined, a C expression whose value is a string containing the
6883 assembler operation to identify the following data as thread-local
6884 common data. The default is @code{".tls_common"}.
6887 @defmac TLS_SECTION_ASM_FLAG
6888 If defined, a C expression whose value is a character constant
6889 containing the flag used to mark a section as a TLS section. The
6890 default is @code{'T'}.
6893 @defmac INIT_SECTION_ASM_OP
6894 If defined, a C expression whose value is a string, including spacing,
6895 containing the assembler operation to identify the following data as
6896 initialization code. If not defined, GCC will assume such a section does
6897 not exist. This section has no corresponding @code{init_section}
6898 variable; it is used entirely in runtime code.
6901 @defmac FINI_SECTION_ASM_OP
6902 If defined, a C expression whose value is a string, including spacing,
6903 containing the assembler operation to identify the following data as
6904 finalization code. If not defined, GCC will assume such a section does
6905 not exist. This section has no corresponding @code{fini_section}
6906 variable; it is used entirely in runtime code.
6909 @defmac INIT_ARRAY_SECTION_ASM_OP
6910 If defined, a C expression whose value is a string, including spacing,
6911 containing the assembler operation to identify the following data as
6912 part of the @code{.init_array} (or equivalent) section. If not
6913 defined, GCC will assume such a section does not exist. Do not define
6914 both this macro and @code{INIT_SECTION_ASM_OP}.
6917 @defmac FINI_ARRAY_SECTION_ASM_OP
6918 If defined, a C expression whose value is a string, including spacing,
6919 containing the assembler operation to identify the following data as
6920 part of the @code{.fini_array} (or equivalent) section. If not
6921 defined, GCC will assume such a section does not exist. Do not define
6922 both this macro and @code{FINI_SECTION_ASM_OP}.
6925 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6926 If defined, an ASM statement that switches to a different section
6927 via @var{section_op}, calls @var{function}, and switches back to
6928 the text section. This is used in @file{crtstuff.c} if
6929 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6930 to initialization and finalization functions from the init and fini
6931 sections. By default, this macro uses a simple function call. Some
6932 ports need hand-crafted assembly code to avoid dependencies on
6933 registers initialized in the function prologue or to ensure that
6934 constant pools don't end up too far way in the text section.
6937 @defmac TARGET_LIBGCC_SDATA_SECTION
6938 If defined, a string which names the section into which small
6939 variables defined in crtstuff and libgcc should go. This is useful
6940 when the target has options for optimizing access to small data, and
6941 you want the crtstuff and libgcc routines to be conservative in what
6942 they expect of your application yet liberal in what your application
6943 expects. For example, for targets with a @code{.sdata} section (like
6944 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6945 require small data support from your application, but use this macro
6946 to put small data into @code{.sdata} so that your application can
6947 access these variables whether it uses small data or not.
6950 @defmac FORCE_CODE_SECTION_ALIGN
6951 If defined, an ASM statement that aligns a code section to some
6952 arbitrary boundary. This is used to force all fragments of the
6953 @code{.init} and @code{.fini} sections to have to same alignment
6954 and thus prevent the linker from having to add any padding.
6957 @defmac JUMP_TABLES_IN_TEXT_SECTION
6958 Define this macro to be an expression with a nonzero value if jump
6959 tables (for @code{tablejump} insns) should be output in the text
6960 section, along with the assembler instructions. Otherwise, the
6961 readonly data section is used.
6963 This macro is irrelevant if there is no separate readonly data section.
6966 @hook TARGET_ASM_INIT_SECTIONS
6967 Define this hook if you need to do something special to set up the
6968 @file{varasm.c} sections, or if your target has some special sections
6969 of its own that you need to create.
6971 GCC calls this hook after processing the command line, but before writing
6972 any assembly code, and before calling any of the section-returning hooks
6976 @hook TARGET_ASM_RELOC_RW_MASK
6977 Return a mask describing how relocations should be treated when
6978 selecting sections. Bit 1 should be set if global relocations
6979 should be placed in a read-write section; bit 0 should be set if
6980 local relocations should be placed in a read-write section.
6982 The default version of this function returns 3 when @option{-fpic}
6983 is in effect, and 0 otherwise. The hook is typically redefined
6984 when the target cannot support (some kinds of) dynamic relocations
6985 in read-only sections even in executables.
6988 @hook TARGET_ASM_SELECT_SECTION
6989 Return the section into which @var{exp} should be placed. You can
6990 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6991 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6992 requires link-time relocations. Bit 0 is set when variable contains
6993 local relocations only, while bit 1 is set for global relocations.
6994 @var{align} is the constant alignment in bits.
6996 The default version of this function takes care of putting read-only
6997 variables in @code{readonly_data_section}.
6999 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7002 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7003 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7004 for @code{FUNCTION_DECL}s as well as for variables and constants.
7006 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7007 function has been determined to be likely to be called, and nonzero if
7008 it is unlikely to be called.
7011 @hook TARGET_ASM_UNIQUE_SECTION
7012 Build up a unique section name, expressed as a @code{STRING_CST} node,
7013 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7014 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7015 the initial value of @var{exp} requires link-time relocations.
7017 The default version of this function appends the symbol name to the
7018 ELF section name that would normally be used for the symbol. For
7019 example, the function @code{foo} would be placed in @code{.text.foo}.
7020 Whatever the actual target object format, this is often good enough.
7023 @hook TARGET_ASM_FUNCTION_RODATA_SECTION
7024 Return the readonly data section associated with
7025 @samp{DECL_SECTION_NAME (@var{decl})}.
7026 The default version of this function selects @code{.gnu.linkonce.r.name} if
7027 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7028 if function is in @code{.text.name}, and the normal readonly-data section
7032 @hook TARGET_ASM_MERGEABLE_RODATA_PREFIX
7034 @hook TARGET_ASM_TM_CLONE_TABLE_SECTION
7036 @hook TARGET_ASM_SELECT_RTX_SECTION
7037 Return the section into which a constant @var{x}, of mode @var{mode},
7038 should be placed. You can assume that @var{x} is some kind of
7039 constant in RTL@. The argument @var{mode} is redundant except in the
7040 case of a @code{const_int} rtx. @var{align} is the constant alignment
7043 The default version of this function takes care of putting symbolic
7044 constants in @code{flag_pic} mode in @code{data_section} and everything
7045 else in @code{readonly_data_section}.
7048 @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
7049 Define this hook if you need to postprocess the assembler name generated
7050 by target-independent code. The @var{id} provided to this hook will be
7051 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7052 or the mangled name of the @var{decl} in C++). The return value of the
7053 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7054 your target system. The default implementation of this hook just
7055 returns the @var{id} provided.
7058 @hook TARGET_ENCODE_SECTION_INFO
7059 Define this hook if references to a symbol or a constant must be
7060 treated differently depending on something about the variable or
7061 function named by the symbol (such as what section it is in).
7063 The hook is executed immediately after rtl has been created for
7064 @var{decl}, which may be a variable or function declaration or
7065 an entry in the constant pool. In either case, @var{rtl} is the
7066 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7067 in this hook; that field may not have been initialized yet.
7069 In the case of a constant, it is safe to assume that the rtl is
7070 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7071 will also have this form, but that is not guaranteed. Global
7072 register variables, for instance, will have a @code{reg} for their
7073 rtl. (Normally the right thing to do with such unusual rtl is
7076 The @var{new_decl_p} argument will be true if this is the first time
7077 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7078 be false for subsequent invocations, which will happen for duplicate
7079 declarations. Whether or not anything must be done for the duplicate
7080 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7081 @var{new_decl_p} is always true when the hook is called for a constant.
7083 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7084 The usual thing for this hook to do is to record flags in the
7085 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7086 Historically, the name string was modified if it was necessary to
7087 encode more than one bit of information, but this practice is now
7088 discouraged; use @code{SYMBOL_REF_FLAGS}.
7090 The default definition of this hook, @code{default_encode_section_info}
7091 in @file{varasm.c}, sets a number of commonly-useful bits in
7092 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7093 before overriding it.
7096 @hook TARGET_STRIP_NAME_ENCODING
7097 Decode @var{name} and return the real name part, sans
7098 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7102 @hook TARGET_IN_SMALL_DATA_P
7103 Returns true if @var{exp} should be placed into a ``small data'' section.
7104 The default version of this hook always returns false.
7107 @hook TARGET_HAVE_SRODATA_SECTION
7108 Contains the value true if the target places read-only
7109 ``small data'' into a separate section. The default value is false.
7112 @hook TARGET_PROFILE_BEFORE_PROLOGUE
7114 @hook TARGET_BINDS_LOCAL_P
7115 Returns true if @var{exp} names an object for which name resolution
7116 rules must resolve to the current ``module'' (dynamic shared library
7117 or executable image).
7119 The default version of this hook implements the name resolution rules
7120 for ELF, which has a looser model of global name binding than other
7121 currently supported object file formats.
7124 @hook TARGET_HAVE_TLS
7125 Contains the value true if the target supports thread-local storage.
7126 The default value is false.
7131 @section Position Independent Code
7132 @cindex position independent code
7135 This section describes macros that help implement generation of position
7136 independent code. Simply defining these macros is not enough to
7137 generate valid PIC; you must also add support to the hook
7138 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7139 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7140 must modify the definition of @samp{movsi} to do something appropriate
7141 when the source operand contains a symbolic address. You may also
7142 need to alter the handling of switch statements so that they use
7144 @c i rearranged the order of the macros above to try to force one of
7145 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7147 @defmac PIC_OFFSET_TABLE_REGNUM
7148 The register number of the register used to address a table of static
7149 data addresses in memory. In some cases this register is defined by a
7150 processor's ``application binary interface'' (ABI)@. When this macro
7151 is defined, RTL is generated for this register once, as with the stack
7152 pointer and frame pointer registers. If this macro is not defined, it
7153 is up to the machine-dependent files to allocate such a register (if
7154 necessary). Note that this register must be fixed when in use (e.g.@:
7155 when @code{flag_pic} is true).
7158 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7159 A C expression that is nonzero if the register defined by
7160 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7161 the default is zero. Do not define
7162 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7165 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7166 A C expression that is nonzero if @var{x} is a legitimate immediate
7167 operand on the target machine when generating position independent code.
7168 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7169 check this. You can also assume @var{flag_pic} is true, so you need not
7170 check it either. You need not define this macro if all constants
7171 (including @code{SYMBOL_REF}) can be immediate operands when generating
7172 position independent code.
7175 @node Assembler Format
7176 @section Defining the Output Assembler Language
7178 This section describes macros whose principal purpose is to describe how
7179 to write instructions in assembler language---rather than what the
7183 * File Framework:: Structural information for the assembler file.
7184 * Data Output:: Output of constants (numbers, strings, addresses).
7185 * Uninitialized Data:: Output of uninitialized variables.
7186 * Label Output:: Output and generation of labels.
7187 * Initialization:: General principles of initialization
7188 and termination routines.
7189 * Macros for Initialization::
7190 Specific macros that control the handling of
7191 initialization and termination routines.
7192 * Instruction Output:: Output of actual instructions.
7193 * Dispatch Tables:: Output of jump tables.
7194 * Exception Region Output:: Output of exception region code.
7195 * Alignment Output:: Pseudo ops for alignment and skipping data.
7198 @node File Framework
7199 @subsection The Overall Framework of an Assembler File
7200 @cindex assembler format
7201 @cindex output of assembler code
7203 @c prevent bad page break with this line
7204 This describes the overall framework of an assembly file.
7206 @findex default_file_start
7207 @hook TARGET_ASM_FILE_START
7208 Output to @code{asm_out_file} any text which the assembler expects to
7209 find at the beginning of a file. The default behavior is controlled
7210 by two flags, documented below. Unless your target's assembler is
7211 quite unusual, if you override the default, you should call
7212 @code{default_file_start} at some point in your target hook. This
7213 lets other target files rely on these variables.
7216 @hook TARGET_ASM_FILE_START_APP_OFF
7217 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7218 printed as the very first line in the assembly file, unless
7219 @option{-fverbose-asm} is in effect. (If that macro has been defined
7220 to the empty string, this variable has no effect.) With the normal
7221 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7222 assembler that it need not bother stripping comments or extra
7223 whitespace from its input. This allows it to work a bit faster.
7225 The default is false. You should not set it to true unless you have
7226 verified that your port does not generate any extra whitespace or
7227 comments that will cause GAS to issue errors in NO_APP mode.
7230 @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
7231 If this flag is true, @code{output_file_directive} will be called
7232 for the primary source file, immediately after printing
7233 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7234 this to be done. The default is false.
7237 @hook TARGET_ASM_FILE_END
7238 Output to @code{asm_out_file} any text which the assembler expects
7239 to find at the end of a file. The default is to output nothing.
7242 @deftypefun void file_end_indicate_exec_stack ()
7243 Some systems use a common convention, the @samp{.note.GNU-stack}
7244 special section, to indicate whether or not an object file relies on
7245 the stack being executable. If your system uses this convention, you
7246 should define @code{TARGET_ASM_FILE_END} to this function. If you
7247 need to do other things in that hook, have your hook function call
7251 @hook TARGET_ASM_LTO_START
7252 Output to @code{asm_out_file} any text which the assembler expects
7253 to find at the start of an LTO section. The default is to output
7257 @hook TARGET_ASM_LTO_END
7258 Output to @code{asm_out_file} any text which the assembler expects
7259 to find at the end of an LTO section. The default is to output
7263 @hook TARGET_ASM_CODE_END
7264 Output to @code{asm_out_file} any text which is needed before emitting
7265 unwind info and debug info at the end of a file. Some targets emit
7266 here PIC setup thunks that cannot be emitted at the end of file,
7267 because they couldn't have unwind info then. The default is to output
7271 @defmac ASM_COMMENT_START
7272 A C string constant describing how to begin a comment in the target
7273 assembler language. The compiler assumes that the comment will end at
7274 the end of the line.
7278 A C string constant for text to be output before each @code{asm}
7279 statement or group of consecutive ones. Normally this is
7280 @code{"#APP"}, which is a comment that has no effect on most
7281 assemblers but tells the GNU assembler that it must check the lines
7282 that follow for all valid assembler constructs.
7286 A C string constant for text to be output after each @code{asm}
7287 statement or group of consecutive ones. Normally this is
7288 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7289 time-saving assumptions that are valid for ordinary compiler output.
7292 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7293 A C statement to output COFF information or DWARF debugging information
7294 which indicates that filename @var{name} is the current source file to
7295 the stdio stream @var{stream}.
7297 This macro need not be defined if the standard form of output
7298 for the file format in use is appropriate.
7301 @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
7303 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7304 A C statement to output the string @var{string} to the stdio stream
7305 @var{stream}. If you do not call the function @code{output_quoted_string}
7306 in your config files, GCC will only call it to output filenames to
7307 the assembler source. So you can use it to canonicalize the format
7308 of the filename using this macro.
7311 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7312 A C statement to output something to the assembler file to handle a
7313 @samp{#ident} directive containing the text @var{string}. If this
7314 macro is not defined, nothing is output for a @samp{#ident} directive.
7317 @hook TARGET_ASM_NAMED_SECTION
7318 Output assembly directives to switch to section @var{name}. The section
7319 should have attributes as specified by @var{flags}, which is a bit mask
7320 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7321 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7322 this section is associated.
7325 @hook TARGET_ASM_FUNCTION_SECTION
7326 Return preferred text (sub)section for function @var{decl}.
7327 Main purpose of this function is to separate cold, normal and hot
7328 functions. @var{startup} is true when function is known to be used only
7329 at startup (from static constructors or it is @code{main()}).
7330 @var{exit} is true when function is known to be used only at exit
7331 (from static destructors).
7332 Return NULL if function should go to default text section.
7335 @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
7337 @hook TARGET_HAVE_NAMED_SECTIONS
7338 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7339 It must not be modified by command-line option processing.
7342 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7343 @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7344 This flag is true if we can create zeroed data by switching to a BSS
7345 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7346 This is true on most ELF targets.
7349 @hook TARGET_SECTION_TYPE_FLAGS
7350 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7351 based on a variable or function decl, a section name, and whether or not the
7352 declaration's initializer may contain runtime relocations. @var{decl} may be
7353 null, in which case read-write data should be assumed.
7355 The default version of this function handles choosing code vs data,
7356 read-only vs read-write data, and @code{flag_pic}. You should only
7357 need to override this if your target has special flags that might be
7358 set via @code{__attribute__}.
7361 @hook TARGET_ASM_RECORD_GCC_SWITCHES
7362 Provides the target with the ability to record the gcc command line
7363 switches that have been passed to the compiler, and options that are
7364 enabled. The @var{type} argument specifies what is being recorded.
7365 It can take the following values:
7368 @item SWITCH_TYPE_PASSED
7369 @var{text} is a command line switch that has been set by the user.
7371 @item SWITCH_TYPE_ENABLED
7372 @var{text} is an option which has been enabled. This might be as a
7373 direct result of a command line switch, or because it is enabled by
7374 default or because it has been enabled as a side effect of a different
7375 command line switch. For example, the @option{-O2} switch enables
7376 various different individual optimization passes.
7378 @item SWITCH_TYPE_DESCRIPTIVE
7379 @var{text} is either NULL or some descriptive text which should be
7380 ignored. If @var{text} is NULL then it is being used to warn the
7381 target hook that either recording is starting or ending. The first
7382 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7383 warning is for start up and the second time the warning is for
7384 wind down. This feature is to allow the target hook to make any
7385 necessary preparations before it starts to record switches and to
7386 perform any necessary tidying up after it has finished recording
7389 @item SWITCH_TYPE_LINE_START
7390 This option can be ignored by this target hook.
7392 @item SWITCH_TYPE_LINE_END
7393 This option can be ignored by this target hook.
7396 The hook's return value must be zero. Other return values may be
7397 supported in the future.
7399 By default this hook is set to NULL, but an example implementation is
7400 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7401 it records the switches as ASCII text inside a new, string mergeable
7402 section in the assembler output file. The name of the new section is
7403 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7407 @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7408 This is the name of the section that will be created by the example
7409 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7415 @subsection Output of Data
7418 @hook TARGET_ASM_BYTE_OP
7419 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7420 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7421 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7422 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7423 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7424 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7425 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7426 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7427 These hooks specify assembly directives for creating certain kinds
7428 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7429 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7430 aligned two-byte object, and so on. Any of the hooks may be
7431 @code{NULL}, indicating that no suitable directive is available.
7433 The compiler will print these strings at the start of a new line,
7434 followed immediately by the object's initial value. In most cases,
7435 the string should contain a tab, a pseudo-op, and then another tab.
7438 @hook TARGET_ASM_INTEGER
7439 The @code{assemble_integer} function uses this hook to output an
7440 integer object. @var{x} is the object's value, @var{size} is its size
7441 in bytes and @var{aligned_p} indicates whether it is aligned. The
7442 function should return @code{true} if it was able to output the
7443 object. If it returns false, @code{assemble_integer} will try to
7444 split the object into smaller parts.
7446 The default implementation of this hook will use the
7447 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7448 when the relevant string is @code{NULL}.
7451 @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
7452 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7453 can't deal with, and output assembly code to @var{file} corresponding to
7454 the pattern @var{x}. This may be used to allow machine-dependent
7455 @code{UNSPEC}s to appear within constants.
7457 If target hook fails to recognize a pattern, it must return @code{false},
7458 so that a standard error message is printed. If it prints an error message
7459 itself, by calling, for example, @code{output_operand_lossage}, it may just
7463 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7464 A C statement to output to the stdio stream @var{stream} an assembler
7465 instruction to assemble a string constant containing the @var{len}
7466 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7467 @code{char *} and @var{len} a C expression of type @code{int}.
7469 If the assembler has a @code{.ascii} pseudo-op as found in the
7470 Berkeley Unix assembler, do not define the macro
7471 @code{ASM_OUTPUT_ASCII}.
7474 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7475 A C statement to output word @var{n} of a function descriptor for
7476 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7477 is defined, and is otherwise unused.
7480 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7481 You may define this macro as a C expression. You should define the
7482 expression to have a nonzero value if GCC should output the constant
7483 pool for a function before the code for the function, or a zero value if
7484 GCC should output the constant pool after the function. If you do
7485 not define this macro, the usual case, GCC will output the constant
7486 pool before the function.
7489 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7490 A C statement to output assembler commands to define the start of the
7491 constant pool for a function. @var{funname} is a string giving
7492 the name of the function. Should the return type of the function
7493 be required, it can be obtained via @var{fundecl}. @var{size}
7494 is the size, in bytes, of the constant pool that will be written
7495 immediately after this call.
7497 If no constant-pool prefix is required, the usual case, this macro need
7501 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7502 A C statement (with or without semicolon) to output a constant in the
7503 constant pool, if it needs special treatment. (This macro need not do
7504 anything for RTL expressions that can be output normally.)
7506 The argument @var{file} is the standard I/O stream to output the
7507 assembler code on. @var{x} is the RTL expression for the constant to
7508 output, and @var{mode} is the machine mode (in case @var{x} is a
7509 @samp{const_int}). @var{align} is the required alignment for the value
7510 @var{x}; you should output an assembler directive to force this much
7513 The argument @var{labelno} is a number to use in an internal label for
7514 the address of this pool entry. The definition of this macro is
7515 responsible for outputting the label definition at the proper place.
7516 Here is how to do this:
7519 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7522 When you output a pool entry specially, you should end with a
7523 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7524 entry from being output a second time in the usual manner.
7526 You need not define this macro if it would do nothing.
7529 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7530 A C statement to output assembler commands to at the end of the constant
7531 pool for a function. @var{funname} is a string giving the name of the
7532 function. Should the return type of the function be required, you can
7533 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7534 constant pool that GCC wrote immediately before this call.
7536 If no constant-pool epilogue is required, the usual case, you need not
7540 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7541 Define this macro as a C expression which is nonzero if @var{C} is
7542 used as a logical line separator by the assembler. @var{STR} points
7543 to the position in the string where @var{C} was found; this can be used if
7544 a line separator uses multiple characters.
7546 If you do not define this macro, the default is that only
7547 the character @samp{;} is treated as a logical line separator.
7550 @hook TARGET_ASM_OPEN_PAREN
7551 These target hooks are C string constants, describing the syntax in the
7552 assembler for grouping arithmetic expressions. If not overridden, they
7553 default to normal parentheses, which is correct for most assemblers.
7556 These macros are provided by @file{real.h} for writing the definitions
7557 of @code{ASM_OUTPUT_DOUBLE} and the like:
7559 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7560 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7561 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7562 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7563 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7564 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7565 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7566 target's floating point representation, and store its bit pattern in
7567 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7568 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7569 simple @code{long int}. For the others, it should be an array of
7570 @code{long int}. The number of elements in this array is determined
7571 by the size of the desired target floating point data type: 32 bits of
7572 it go in each @code{long int} array element. Each array element holds
7573 32 bits of the result, even if @code{long int} is wider than 32 bits
7574 on the host machine.
7576 The array element values are designed so that you can print them out
7577 using @code{fprintf} in the order they should appear in the target
7581 @node Uninitialized Data
7582 @subsection Output of Uninitialized Variables
7584 Each of the macros in this section is used to do the whole job of
7585 outputting a single uninitialized variable.
7587 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7588 A C statement (sans semicolon) to output to the stdio stream
7589 @var{stream} the assembler definition of a common-label named
7590 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7591 is the size rounded up to whatever alignment the caller wants. It is
7592 possible that @var{size} may be zero, for instance if a struct with no
7593 other member than a zero-length array is defined. In this case, the
7594 backend must output a symbol definition that allocates at least one
7595 byte, both so that the address of the resulting object does not compare
7596 equal to any other, and because some object formats cannot even express
7597 the concept of a zero-sized common symbol, as that is how they represent
7598 an ordinary undefined external.
7600 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7601 output the name itself; before and after that, output the additional
7602 assembler syntax for defining the name, and a newline.
7604 This macro controls how the assembler definitions of uninitialized
7605 common global variables are output.
7608 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7609 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7610 separate, explicit argument. If you define this macro, it is used in
7611 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7612 handling the required alignment of the variable. The alignment is specified
7613 as the number of bits.
7616 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7617 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7618 variable to be output, if there is one, or @code{NULL_TREE} if there
7619 is no corresponding variable. If you define this macro, GCC will use it
7620 in place of both @code{ASM_OUTPUT_COMMON} and
7621 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7622 the variable's decl in order to chose what to output.
7625 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7626 A C statement (sans semicolon) to output to the stdio stream
7627 @var{stream} the assembler definition of uninitialized global @var{decl} named
7628 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
7629 is the alignment specified as the number of bits.
7631 Try to use function @code{asm_output_aligned_bss} defined in file
7632 @file{varasm.c} when defining this macro. If unable, use the expression
7633 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7634 before and after that, output the additional assembler syntax for defining
7635 the name, and a newline.
7637 There are two ways of handling global BSS@. One is to define this macro.
7638 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7639 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7640 You do not need to do both.
7642 Some languages do not have @code{common} data, and require a
7643 non-common form of global BSS in order to handle uninitialized globals
7644 efficiently. C++ is one example of this. However, if the target does
7645 not support global BSS, the front end may choose to make globals
7646 common in order to save space in the object file.
7649 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7650 A C statement (sans semicolon) to output to the stdio stream
7651 @var{stream} the assembler definition of a local-common-label named
7652 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7653 is the size rounded up to whatever alignment the caller wants.
7655 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7656 output the name itself; before and after that, output the additional
7657 assembler syntax for defining the name, and a newline.
7659 This macro controls how the assembler definitions of uninitialized
7660 static variables are output.
7663 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7664 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7665 separate, explicit argument. If you define this macro, it is used in
7666 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7667 handling the required alignment of the variable. The alignment is specified
7668 as the number of bits.
7671 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7672 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7673 variable to be output, if there is one, or @code{NULL_TREE} if there
7674 is no corresponding variable. If you define this macro, GCC will use it
7675 in place of both @code{ASM_OUTPUT_DECL} and
7676 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7677 the variable's decl in order to chose what to output.
7681 @subsection Output and Generation of Labels
7683 @c prevent bad page break with this line
7684 This is about outputting labels.
7686 @findex assemble_name
7687 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7688 A C statement (sans semicolon) to output to the stdio stream
7689 @var{stream} the assembler definition of a label named @var{name}.
7690 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7691 output the name itself; before and after that, output the additional
7692 assembler syntax for defining the name, and a newline. A default
7693 definition of this macro is provided which is correct for most systems.
7696 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7697 A C statement (sans semicolon) to output to the stdio stream
7698 @var{stream} the assembler definition of a label named @var{name} of
7700 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7701 output the name itself; before and after that, output the additional
7702 assembler syntax for defining the name, and a newline. A default
7703 definition of this macro is provided which is correct for most systems.
7705 If this macro is not defined, then the function name is defined in the
7706 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7709 @findex assemble_name_raw
7710 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7711 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7712 to refer to a compiler-generated label. The default definition uses
7713 @code{assemble_name_raw}, which is like @code{assemble_name} except
7714 that it is more efficient.
7718 A C string containing the appropriate assembler directive to specify the
7719 size of a symbol, without any arguments. On systems that use ELF, the
7720 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7721 systems, the default is not to define this macro.
7723 Define this macro only if it is correct to use the default definitions
7724 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7725 for your system. If you need your own custom definitions of those
7726 macros, or if you do not need explicit symbol sizes at all, do not
7730 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7731 A C statement (sans semicolon) to output to the stdio stream
7732 @var{stream} a directive telling the assembler that the size of the
7733 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7734 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7738 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7739 A C statement (sans semicolon) to output to the stdio stream
7740 @var{stream} a directive telling the assembler to calculate the size of
7741 the symbol @var{name} by subtracting its address from the current
7744 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7745 provided. The default assumes that the assembler recognizes a special
7746 @samp{.} symbol as referring to the current address, and can calculate
7747 the difference between this and another symbol. If your assembler does
7748 not recognize @samp{.} or cannot do calculations with it, you will need
7749 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7753 A C string containing the appropriate assembler directive to specify the
7754 type of a symbol, without any arguments. On systems that use ELF, the
7755 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7756 systems, the default is not to define this macro.
7758 Define this macro only if it is correct to use the default definition of
7759 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7760 custom definition of this macro, or if you do not need explicit symbol
7761 types at all, do not define this macro.
7764 @defmac TYPE_OPERAND_FMT
7765 A C string which specifies (using @code{printf} syntax) the format of
7766 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7767 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7768 the default is not to define this macro.
7770 Define this macro only if it is correct to use the default definition of
7771 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7772 custom definition of this macro, or if you do not need explicit symbol
7773 types at all, do not define this macro.
7776 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7777 A C statement (sans semicolon) to output to the stdio stream
7778 @var{stream} a directive telling the assembler that the type of the
7779 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7780 that string is always either @samp{"function"} or @samp{"object"}, but
7781 you should not count on this.
7783 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7784 definition of this macro is provided.
7787 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7788 A C statement (sans semicolon) to output to the stdio stream
7789 @var{stream} any text necessary for declaring the name @var{name} of a
7790 function which is being defined. This macro is responsible for
7791 outputting the label definition (perhaps using
7792 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7793 @code{FUNCTION_DECL} tree node representing the function.
7795 If this macro is not defined, then the function name is defined in the
7796 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7798 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7802 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7803 A C statement (sans semicolon) to output to the stdio stream
7804 @var{stream} any text necessary for declaring the size of a function
7805 which is being defined. The argument @var{name} is the name of the
7806 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7807 representing the function.
7809 If this macro is not defined, then the function size is not defined.
7811 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7815 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7816 A C statement (sans semicolon) to output to the stdio stream
7817 @var{stream} any text necessary for declaring the name @var{name} of an
7818 initialized variable which is being defined. This macro must output the
7819 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7820 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7822 If this macro is not defined, then the variable name is defined in the
7823 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7825 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7826 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7829 @hook TARGET_ASM_DECLARE_CONSTANT_NAME
7830 A target hook to output to the stdio stream @var{file} any text necessary
7831 for declaring the name @var{name} of a constant which is being defined. This
7832 target hook is responsible for outputting the label definition (perhaps using
7833 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7834 and @var{size} is the size of the constant in bytes. The @var{name}
7835 will be an internal label.
7837 The default version of this target hook, define the @var{name} in the
7838 usual manner as a label (by means of @code{assemble_label}).
7840 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7843 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7844 A C statement (sans semicolon) to output to the stdio stream
7845 @var{stream} any text necessary for claiming a register @var{regno}
7846 for a global variable @var{decl} with name @var{name}.
7848 If you don't define this macro, that is equivalent to defining it to do
7852 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7853 A C statement (sans semicolon) to finish up declaring a variable name
7854 once the compiler has processed its initializer fully and thus has had a
7855 chance to determine the size of an array when controlled by an
7856 initializer. This is used on systems where it's necessary to declare
7857 something about the size of the object.
7859 If you don't define this macro, that is equivalent to defining it to do
7862 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7863 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7866 @hook TARGET_ASM_GLOBALIZE_LABEL
7867 This target hook is a function to output to the stdio stream
7868 @var{stream} some commands that will make the label @var{name} global;
7869 that is, available for reference from other files.
7871 The default implementation relies on a proper definition of
7872 @code{GLOBAL_ASM_OP}.
7875 @hook TARGET_ASM_GLOBALIZE_DECL_NAME
7876 This target hook is a function to output to the stdio stream
7877 @var{stream} some commands that will make the name associated with @var{decl}
7878 global; that is, available for reference from other files.
7880 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7883 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7884 A C statement (sans semicolon) to output to the stdio stream
7885 @var{stream} some commands that will make the label @var{name} weak;
7886 that is, available for reference from other files but only used if
7887 no other definition is available. Use the expression
7888 @code{assemble_name (@var{stream}, @var{name})} to output the name
7889 itself; before and after that, output the additional assembler syntax
7890 for making that name weak, and a newline.
7892 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7893 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7897 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7898 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7899 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7900 or variable decl. If @var{value} is not @code{NULL}, this C statement
7901 should output to the stdio stream @var{stream} assembler code which
7902 defines (equates) the weak symbol @var{name} to have the value
7903 @var{value}. If @var{value} is @code{NULL}, it should output commands
7904 to make @var{name} weak.
7907 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7908 Outputs a directive that enables @var{name} to be used to refer to
7909 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7910 declaration of @code{name}.
7913 @defmac SUPPORTS_WEAK
7914 A preprocessor constant expression which evaluates to true if the target
7915 supports weak symbols.
7917 If you don't define this macro, @file{defaults.h} provides a default
7918 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7919 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
7922 @defmac TARGET_SUPPORTS_WEAK
7923 A C expression which evaluates to true if the target supports weak symbols.
7925 If you don't define this macro, @file{defaults.h} provides a default
7926 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
7927 this macro if you want to control weak symbol support with a compiler
7928 flag such as @option{-melf}.
7931 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7932 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7933 public symbol such that extra copies in multiple translation units will
7934 be discarded by the linker. Define this macro if your object file
7935 format provides support for this concept, such as the @samp{COMDAT}
7936 section flags in the Microsoft Windows PE/COFF format, and this support
7937 requires changes to @var{decl}, such as putting it in a separate section.
7940 @defmac SUPPORTS_ONE_ONLY
7941 A C expression which evaluates to true if the target supports one-only
7944 If you don't define this macro, @file{varasm.c} provides a default
7945 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7946 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7947 you want to control one-only symbol support with a compiler flag, or if
7948 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7949 be emitted as one-only.
7952 @hook TARGET_ASM_ASSEMBLE_VISIBILITY
7953 This target hook is a function to output to @var{asm_out_file} some
7954 commands that will make the symbol(s) associated with @var{decl} have
7955 hidden, protected or internal visibility as specified by @var{visibility}.
7958 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7959 A C expression that evaluates to true if the target's linker expects
7960 that weak symbols do not appear in a static archive's table of contents.
7961 The default is @code{0}.
7963 Leaving weak symbols out of an archive's table of contents means that,
7964 if a symbol will only have a definition in one translation unit and
7965 will have undefined references from other translation units, that
7966 symbol should not be weak. Defining this macro to be nonzero will
7967 thus have the effect that certain symbols that would normally be weak
7968 (explicit template instantiations, and vtables for polymorphic classes
7969 with noninline key methods) will instead be nonweak.
7971 The C++ ABI requires this macro to be zero. Define this macro for
7972 targets where full C++ ABI compliance is impossible and where linker
7973 restrictions require weak symbols to be left out of a static archive's
7977 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7978 A C statement (sans semicolon) to output to the stdio stream
7979 @var{stream} any text necessary for declaring the name of an external
7980 symbol named @var{name} which is referenced in this compilation but
7981 not defined. The value of @var{decl} is the tree node for the
7984 This macro need not be defined if it does not need to output anything.
7985 The GNU assembler and most Unix assemblers don't require anything.
7988 @hook TARGET_ASM_EXTERNAL_LIBCALL
7989 This target hook is a function to output to @var{asm_out_file} an assembler
7990 pseudo-op to declare a library function name external. The name of the
7991 library function is given by @var{symref}, which is a @code{symbol_ref}.
7994 @hook TARGET_ASM_MARK_DECL_PRESERVED
7995 This target hook is a function to output to @var{asm_out_file} an assembler
7996 directive to annotate @var{symbol} as used. The Darwin target uses the
7997 .no_dead_code_strip directive.
8000 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8001 A C statement (sans semicolon) to output to the stdio stream
8002 @var{stream} a reference in assembler syntax to a label named
8003 @var{name}. This should add @samp{_} to the front of the name, if that
8004 is customary on your operating system, as it is in most Berkeley Unix
8005 systems. This macro is used in @code{assemble_name}.
8008 @hook TARGET_MANGLE_ASSEMBLER_NAME
8010 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8011 A C statement (sans semicolon) to output a reference to
8012 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8013 will be used to output the name of the symbol. This macro may be used
8014 to modify the way a symbol is referenced depending on information
8015 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8018 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8019 A C statement (sans semicolon) to output a reference to @var{buf}, the
8020 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8021 @code{assemble_name} will be used to output the name of the symbol.
8022 This macro is not used by @code{output_asm_label}, or the @code{%l}
8023 specifier that calls it; the intention is that this macro should be set
8024 when it is necessary to output a label differently when its address is
8028 @hook TARGET_ASM_INTERNAL_LABEL
8029 A function to output to the stdio stream @var{stream} a label whose
8030 name is made from the string @var{prefix} and the number @var{labelno}.
8032 It is absolutely essential that these labels be distinct from the labels
8033 used for user-level functions and variables. Otherwise, certain programs
8034 will have name conflicts with internal labels.
8036 It is desirable to exclude internal labels from the symbol table of the
8037 object file. Most assemblers have a naming convention for labels that
8038 should be excluded; on many systems, the letter @samp{L} at the
8039 beginning of a label has this effect. You should find out what
8040 convention your system uses, and follow it.
8042 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8045 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8046 A C statement to output to the stdio stream @var{stream} a debug info
8047 label whose name is made from the string @var{prefix} and the number
8048 @var{num}. This is useful for VLIW targets, where debug info labels
8049 may need to be treated differently than branch target labels. On some
8050 systems, branch target labels must be at the beginning of instruction
8051 bundles, but debug info labels can occur in the middle of instruction
8054 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8058 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8059 A C statement to store into the string @var{string} a label whose name
8060 is made from the string @var{prefix} and the number @var{num}.
8062 This string, when output subsequently by @code{assemble_name}, should
8063 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8064 with the same @var{prefix} and @var{num}.
8066 If the string begins with @samp{*}, then @code{assemble_name} will
8067 output the rest of the string unchanged. It is often convenient for
8068 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8069 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8070 to output the string, and may change it. (Of course,
8071 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8072 you should know what it does on your machine.)
8075 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8076 A C expression to assign to @var{outvar} (which is a variable of type
8077 @code{char *}) a newly allocated string made from the string
8078 @var{name} and the number @var{number}, with some suitable punctuation
8079 added. Use @code{alloca} to get space for the string.
8081 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8082 produce an assembler label for an internal static variable whose name is
8083 @var{name}. Therefore, the string must be such as to result in valid
8084 assembler code. The argument @var{number} is different each time this
8085 macro is executed; it prevents conflicts between similarly-named
8086 internal static variables in different scopes.
8088 Ideally this string should not be a valid C identifier, to prevent any
8089 conflict with the user's own symbols. Most assemblers allow periods
8090 or percent signs in assembler symbols; putting at least one of these
8091 between the name and the number will suffice.
8093 If this macro is not defined, a default definition will be provided
8094 which is correct for most systems.
8097 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8098 A C statement to output to the stdio stream @var{stream} assembler code
8099 which defines (equates) the symbol @var{name} to have the value @var{value}.
8102 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8103 correct for most systems.
8106 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8107 A C statement to output to the stdio stream @var{stream} assembler code
8108 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8109 to have the value of the tree node @var{decl_of_value}. This macro will
8110 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8111 the tree nodes are available.
8114 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8115 correct for most systems.
8118 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8119 A C statement that evaluates to true if the assembler code which defines
8120 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8121 of the tree node @var{decl_of_value} should be emitted near the end of the
8122 current compilation unit. The default is to not defer output of defines.
8123 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8124 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8127 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8128 A C statement to output to the stdio stream @var{stream} assembler code
8129 which defines (equates) the weak symbol @var{name} to have the value
8130 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8131 an undefined weak symbol.
8133 Define this macro if the target only supports weak aliases; define
8134 @code{ASM_OUTPUT_DEF} instead if possible.
8137 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8138 Define this macro to override the default assembler names used for
8139 Objective-C methods.
8141 The default name is a unique method number followed by the name of the
8142 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8143 the category is also included in the assembler name (e.g.@:
8146 These names are safe on most systems, but make debugging difficult since
8147 the method's selector is not present in the name. Therefore, particular
8148 systems define other ways of computing names.
8150 @var{buf} is an expression of type @code{char *} which gives you a
8151 buffer in which to store the name; its length is as long as
8152 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8153 50 characters extra.
8155 The argument @var{is_inst} specifies whether the method is an instance
8156 method or a class method; @var{class_name} is the name of the class;
8157 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8158 in a category); and @var{sel_name} is the name of the selector.
8160 On systems where the assembler can handle quoted names, you can use this
8161 macro to provide more human-readable names.
8164 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
8165 A C statement (sans semicolon) to output to the stdio stream
8166 @var{stream} commands to declare that the label @var{name} is an
8167 Objective-C class reference. This is only needed for targets whose
8168 linkers have special support for NeXT-style runtimes.
8171 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
8172 A C statement (sans semicolon) to output to the stdio stream
8173 @var{stream} commands to declare that the label @var{name} is an
8174 unresolved Objective-C class reference. This is only needed for targets
8175 whose linkers have special support for NeXT-style runtimes.
8178 @node Initialization
8179 @subsection How Initialization Functions Are Handled
8180 @cindex initialization routines
8181 @cindex termination routines
8182 @cindex constructors, output of
8183 @cindex destructors, output of
8185 The compiled code for certain languages includes @dfn{constructors}
8186 (also called @dfn{initialization routines})---functions to initialize
8187 data in the program when the program is started. These functions need
8188 to be called before the program is ``started''---that is to say, before
8189 @code{main} is called.
8191 Compiling some languages generates @dfn{destructors} (also called
8192 @dfn{termination routines}) that should be called when the program
8195 To make the initialization and termination functions work, the compiler
8196 must output something in the assembler code to cause those functions to
8197 be called at the appropriate time. When you port the compiler to a new
8198 system, you need to specify how to do this.
8200 There are two major ways that GCC currently supports the execution of
8201 initialization and termination functions. Each way has two variants.
8202 Much of the structure is common to all four variations.
8204 @findex __CTOR_LIST__
8205 @findex __DTOR_LIST__
8206 The linker must build two lists of these functions---a list of
8207 initialization functions, called @code{__CTOR_LIST__}, and a list of
8208 termination functions, called @code{__DTOR_LIST__}.
8210 Each list always begins with an ignored function pointer (which may hold
8211 0, @minus{}1, or a count of the function pointers after it, depending on
8212 the environment). This is followed by a series of zero or more function
8213 pointers to constructors (or destructors), followed by a function
8214 pointer containing zero.
8216 Depending on the operating system and its executable file format, either
8217 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8218 time and exit time. Constructors are called in reverse order of the
8219 list; destructors in forward order.
8221 The best way to handle static constructors works only for object file
8222 formats which provide arbitrarily-named sections. A section is set
8223 aside for a list of constructors, and another for a list of destructors.
8224 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8225 object file that defines an initialization function also puts a word in
8226 the constructor section to point to that function. The linker
8227 accumulates all these words into one contiguous @samp{.ctors} section.
8228 Termination functions are handled similarly.
8230 This method will be chosen as the default by @file{target-def.h} if
8231 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8232 support arbitrary sections, but does support special designated
8233 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8234 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8236 When arbitrary sections are available, there are two variants, depending
8237 upon how the code in @file{crtstuff.c} is called. On systems that
8238 support a @dfn{.init} section which is executed at program startup,
8239 parts of @file{crtstuff.c} are compiled into that section. The
8240 program is linked by the @command{gcc} driver like this:
8243 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8246 The prologue of a function (@code{__init}) appears in the @code{.init}
8247 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8248 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8249 files are provided by the operating system or by the GNU C library, but
8250 are provided by GCC for a few targets.
8252 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8253 compiled from @file{crtstuff.c}. They contain, among other things, code
8254 fragments within the @code{.init} and @code{.fini} sections that branch
8255 to routines in the @code{.text} section. The linker will pull all parts
8256 of a section together, which results in a complete @code{__init} function
8257 that invokes the routines we need at startup.
8259 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8262 If no init section is available, when GCC compiles any function called
8263 @code{main} (or more accurately, any function designated as a program
8264 entry point by the language front end calling @code{expand_main_function}),
8265 it inserts a procedure call to @code{__main} as the first executable code
8266 after the function prologue. The @code{__main} function is defined
8267 in @file{libgcc2.c} and runs the global constructors.
8269 In file formats that don't support arbitrary sections, there are again
8270 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8271 and an `a.out' format must be used. In this case,
8272 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8273 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8274 and with the address of the void function containing the initialization
8275 code as its value. The GNU linker recognizes this as a request to add
8276 the value to a @dfn{set}; the values are accumulated, and are eventually
8277 placed in the executable as a vector in the format described above, with
8278 a leading (ignored) count and a trailing zero element.
8279 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8280 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8281 the compilation of @code{main} to call @code{__main} as above, starting
8282 the initialization process.
8284 The last variant uses neither arbitrary sections nor the GNU linker.
8285 This is preferable when you want to do dynamic linking and when using
8286 file formats which the GNU linker does not support, such as `ECOFF'@. In
8287 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8288 termination functions are recognized simply by their names. This requires
8289 an extra program in the linkage step, called @command{collect2}. This program
8290 pretends to be the linker, for use with GCC; it does its job by running
8291 the ordinary linker, but also arranges to include the vectors of
8292 initialization and termination functions. These functions are called
8293 via @code{__main} as described above. In order to use this method,
8294 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8297 The following section describes the specific macros that control and
8298 customize the handling of initialization and termination functions.
8301 @node Macros for Initialization
8302 @subsection Macros Controlling Initialization Routines
8304 Here are the macros that control how the compiler handles initialization
8305 and termination functions:
8307 @defmac INIT_SECTION_ASM_OP
8308 If defined, a C string constant, including spacing, for the assembler
8309 operation to identify the following data as initialization code. If not
8310 defined, GCC will assume such a section does not exist. When you are
8311 using special sections for initialization and termination functions, this
8312 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8313 run the initialization functions.
8316 @defmac HAS_INIT_SECTION
8317 If defined, @code{main} will not call @code{__main} as described above.
8318 This macro should be defined for systems that control start-up code
8319 on a symbol-by-symbol basis, such as OSF/1, and should not
8320 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8323 @defmac LD_INIT_SWITCH
8324 If defined, a C string constant for a switch that tells the linker that
8325 the following symbol is an initialization routine.
8328 @defmac LD_FINI_SWITCH
8329 If defined, a C string constant for a switch that tells the linker that
8330 the following symbol is a finalization routine.
8333 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8334 If defined, a C statement that will write a function that can be
8335 automatically called when a shared library is loaded. The function
8336 should call @var{func}, which takes no arguments. If not defined, and
8337 the object format requires an explicit initialization function, then a
8338 function called @code{_GLOBAL__DI} will be generated.
8340 This function and the following one are used by collect2 when linking a
8341 shared library that needs constructors or destructors, or has DWARF2
8342 exception tables embedded in the code.
8345 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8346 If defined, a C statement that will write a function that can be
8347 automatically called when a shared library is unloaded. The function
8348 should call @var{func}, which takes no arguments. If not defined, and
8349 the object format requires an explicit finalization function, then a
8350 function called @code{_GLOBAL__DD} will be generated.
8353 @defmac INVOKE__main
8354 If defined, @code{main} will call @code{__main} despite the presence of
8355 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8356 where the init section is not actually run automatically, but is still
8357 useful for collecting the lists of constructors and destructors.
8360 @defmac SUPPORTS_INIT_PRIORITY
8361 If nonzero, the C++ @code{init_priority} attribute is supported and the
8362 compiler should emit instructions to control the order of initialization
8363 of objects. If zero, the compiler will issue an error message upon
8364 encountering an @code{init_priority} attribute.
8367 @hook TARGET_HAVE_CTORS_DTORS
8368 This value is true if the target supports some ``native'' method of
8369 collecting constructors and destructors to be run at startup and exit.
8370 It is false if we must use @command{collect2}.
8373 @hook TARGET_ASM_CONSTRUCTOR
8374 If defined, a function that outputs assembler code to arrange to call
8375 the function referenced by @var{symbol} at initialization time.
8377 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8378 no arguments and with no return value. If the target supports initialization
8379 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8380 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8382 If this macro is not defined by the target, a suitable default will
8383 be chosen if (1) the target supports arbitrary section names, (2) the
8384 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8388 @hook TARGET_ASM_DESTRUCTOR
8389 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8390 functions rather than initialization functions.
8393 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8394 generated for the generated object file will have static linkage.
8396 If your system uses @command{collect2} as the means of processing
8397 constructors, then that program normally uses @command{nm} to scan
8398 an object file for constructor functions to be called.
8400 On certain kinds of systems, you can define this macro to make
8401 @command{collect2} work faster (and, in some cases, make it work at all):
8403 @defmac OBJECT_FORMAT_COFF
8404 Define this macro if the system uses COFF (Common Object File Format)
8405 object files, so that @command{collect2} can assume this format and scan
8406 object files directly for dynamic constructor/destructor functions.
8408 This macro is effective only in a native compiler; @command{collect2} as
8409 part of a cross compiler always uses @command{nm} for the target machine.
8412 @defmac REAL_NM_FILE_NAME
8413 Define this macro as a C string constant containing the file name to use
8414 to execute @command{nm}. The default is to search the path normally for
8419 @command{collect2} calls @command{nm} to scan object files for static
8420 constructors and destructors and LTO info. By default, @option{-n} is
8421 passed. Define @code{NM_FLAGS} to a C string constant if other options
8422 are needed to get the same output format as GNU @command{nm -n}
8426 If your system supports shared libraries and has a program to list the
8427 dynamic dependencies of a given library or executable, you can define
8428 these macros to enable support for running initialization and
8429 termination functions in shared libraries:
8432 Define this macro to a C string constant containing the name of the program
8433 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8436 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8437 Define this macro to be C code that extracts filenames from the output
8438 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8439 of type @code{char *} that points to the beginning of a line of output
8440 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8441 code must advance @var{ptr} to the beginning of the filename on that
8442 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8445 @defmac SHLIB_SUFFIX
8446 Define this macro to a C string constant containing the default shared
8447 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8448 strips version information after this suffix when generating global
8449 constructor and destructor names. This define is only needed on targets
8450 that use @command{collect2} to process constructors and destructors.
8453 @node Instruction Output
8454 @subsection Output of Assembler Instructions
8456 @c prevent bad page break with this line
8457 This describes assembler instruction output.
8459 @defmac REGISTER_NAMES
8460 A C initializer containing the assembler's names for the machine
8461 registers, each one as a C string constant. This is what translates
8462 register numbers in the compiler into assembler language.
8465 @defmac ADDITIONAL_REGISTER_NAMES
8466 If defined, a C initializer for an array of structures containing a name
8467 and a register number. This macro defines additional names for hard
8468 registers, thus allowing the @code{asm} option in declarations to refer
8469 to registers using alternate names.
8472 @defmac OVERLAPPING_REGISTER_NAMES
8473 If defined, a C initializer for an array of structures containing a
8474 name, a register number and a count of the number of consecutive
8475 machine registers the name overlaps. This macro defines additional
8476 names for hard registers, thus allowing the @code{asm} option in
8477 declarations to refer to registers using alternate names. Unlike
8478 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8479 register name implies multiple underlying registers.
8481 This macro should be used when it is important that a clobber in an
8482 @code{asm} statement clobbers all the underlying values implied by the
8483 register name. For example, on ARM, clobbering the double-precision
8484 VFP register ``d0'' implies clobbering both single-precision registers
8488 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8489 Define this macro if you are using an unusual assembler that
8490 requires different names for the machine instructions.
8492 The definition is a C statement or statements which output an
8493 assembler instruction opcode to the stdio stream @var{stream}. The
8494 macro-operand @var{ptr} is a variable of type @code{char *} which
8495 points to the opcode name in its ``internal'' form---the form that is
8496 written in the machine description. The definition should output the
8497 opcode name to @var{stream}, performing any translation you desire, and
8498 increment the variable @var{ptr} to point at the end of the opcode
8499 so that it will not be output twice.
8501 In fact, your macro definition may process less than the entire opcode
8502 name, or more than the opcode name; but if you want to process text
8503 that includes @samp{%}-sequences to substitute operands, you must take
8504 care of the substitution yourself. Just be sure to increment
8505 @var{ptr} over whatever text should not be output normally.
8507 @findex recog_data.operand
8508 If you need to look at the operand values, they can be found as the
8509 elements of @code{recog_data.operand}.
8511 If the macro definition does nothing, the instruction is output
8515 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8516 If defined, a C statement to be executed just prior to the output of
8517 assembler code for @var{insn}, to modify the extracted operands so
8518 they will be output differently.
8520 Here the argument @var{opvec} is the vector containing the operands
8521 extracted from @var{insn}, and @var{noperands} is the number of
8522 elements of the vector which contain meaningful data for this insn.
8523 The contents of this vector are what will be used to convert the insn
8524 template into assembler code, so you can change the assembler output
8525 by changing the contents of the vector.
8527 This macro is useful when various assembler syntaxes share a single
8528 file of instruction patterns; by defining this macro differently, you
8529 can cause a large class of instructions to be output differently (such
8530 as with rearranged operands). Naturally, variations in assembler
8531 syntax affecting individual insn patterns ought to be handled by
8532 writing conditional output routines in those patterns.
8534 If this macro is not defined, it is equivalent to a null statement.
8537 @hook TARGET_ASM_FINAL_POSTSCAN_INSN
8538 If defined, this target hook is a function which is executed just after the
8539 output of assembler code for @var{insn}, to change the mode of the assembler
8542 Here the argument @var{opvec} is the vector containing the operands
8543 extracted from @var{insn}, and @var{noperands} is the number of
8544 elements of the vector which contain meaningful data for this insn.
8545 The contents of this vector are what was used to convert the insn
8546 template into assembler code, so you can change the assembler mode
8547 by checking the contents of the vector.
8550 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8551 A C compound statement to output to stdio stream @var{stream} the
8552 assembler syntax for an instruction operand @var{x}. @var{x} is an
8555 @var{code} is a value that can be used to specify one of several ways
8556 of printing the operand. It is used when identical operands must be
8557 printed differently depending on the context. @var{code} comes from
8558 the @samp{%} specification that was used to request printing of the
8559 operand. If the specification was just @samp{%@var{digit}} then
8560 @var{code} is 0; if the specification was @samp{%@var{ltr}
8561 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8564 If @var{x} is a register, this macro should print the register's name.
8565 The names can be found in an array @code{reg_names} whose type is
8566 @code{char *[]}. @code{reg_names} is initialized from
8567 @code{REGISTER_NAMES}.
8569 When the machine description has a specification @samp{%@var{punct}}
8570 (a @samp{%} followed by a punctuation character), this macro is called
8571 with a null pointer for @var{x} and the punctuation character for
8575 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8576 A C expression which evaluates to true if @var{code} is a valid
8577 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8578 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8579 punctuation characters (except for the standard one, @samp{%}) are used
8583 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8584 A C compound statement to output to stdio stream @var{stream} the
8585 assembler syntax for an instruction operand that is a memory reference
8586 whose address is @var{x}. @var{x} is an RTL expression.
8588 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8589 On some machines, the syntax for a symbolic address depends on the
8590 section that the address refers to. On these machines, define the hook
8591 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8592 @code{symbol_ref}, and then check for it here. @xref{Assembler
8596 @findex dbr_sequence_length
8597 @defmac DBR_OUTPUT_SEQEND (@var{file})
8598 A C statement, to be executed after all slot-filler instructions have
8599 been output. If necessary, call @code{dbr_sequence_length} to
8600 determine the number of slots filled in a sequence (zero if not
8601 currently outputting a sequence), to decide how many no-ops to output,
8604 Don't define this macro if it has nothing to do, but it is helpful in
8605 reading assembly output if the extent of the delay sequence is made
8606 explicit (e.g.@: with white space).
8609 @findex final_sequence
8610 Note that output routines for instructions with delay slots must be
8611 prepared to deal with not being output as part of a sequence
8612 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8613 found.) The variable @code{final_sequence} is null when not
8614 processing a sequence, otherwise it contains the @code{sequence} rtx
8618 @defmac REGISTER_PREFIX
8619 @defmacx LOCAL_LABEL_PREFIX
8620 @defmacx USER_LABEL_PREFIX
8621 @defmacx IMMEDIATE_PREFIX
8622 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8623 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8624 @file{final.c}). These are useful when a single @file{md} file must
8625 support multiple assembler formats. In that case, the various @file{tm.h}
8626 files can define these macros differently.
8629 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8630 If defined this macro should expand to a series of @code{case}
8631 statements which will be parsed inside the @code{switch} statement of
8632 the @code{asm_fprintf} function. This allows targets to define extra
8633 printf formats which may useful when generating their assembler
8634 statements. Note that uppercase letters are reserved for future
8635 generic extensions to asm_fprintf, and so are not available to target
8636 specific code. The output file is given by the parameter @var{file}.
8637 The varargs input pointer is @var{argptr} and the rest of the format
8638 string, starting the character after the one that is being switched
8639 upon, is pointed to by @var{format}.
8642 @defmac ASSEMBLER_DIALECT
8643 If your target supports multiple dialects of assembler language (such as
8644 different opcodes), define this macro as a C expression that gives the
8645 numeric index of the assembler language dialect to use, with zero as the
8648 If this macro is defined, you may use constructs of the form
8650 @samp{@{option0|option1|option2@dots{}@}}
8653 in the output templates of patterns (@pxref{Output Template}) or in the
8654 first argument of @code{asm_fprintf}. This construct outputs
8655 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8656 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8657 within these strings retain their usual meaning. If there are fewer
8658 alternatives within the braces than the value of
8659 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8661 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8662 @samp{@}} do not have any special meaning when used in templates or
8663 operands to @code{asm_fprintf}.
8665 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8666 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8667 the variations in assembler language syntax with that mechanism. Define
8668 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8669 if the syntax variant are larger and involve such things as different
8670 opcodes or operand order.
8673 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8674 A C expression to output to @var{stream} some assembler code
8675 which will push hard register number @var{regno} onto the stack.
8676 The code need not be optimal, since this macro is used only when
8680 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8681 A C expression to output to @var{stream} some assembler code
8682 which will pop hard register number @var{regno} off of the stack.
8683 The code need not be optimal, since this macro is used only when
8687 @node Dispatch Tables
8688 @subsection Output of Dispatch Tables
8690 @c prevent bad page break with this line
8691 This concerns dispatch tables.
8693 @cindex dispatch table
8694 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8695 A C statement to output to the stdio stream @var{stream} an assembler
8696 pseudo-instruction to generate a difference between two labels.
8697 @var{value} and @var{rel} are the numbers of two internal labels. The
8698 definitions of these labels are output using
8699 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8700 way here. For example,
8703 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8704 @var{value}, @var{rel})
8707 You must provide this macro on machines where the addresses in a
8708 dispatch table are relative to the table's own address. If defined, GCC
8709 will also use this macro on all machines when producing PIC@.
8710 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8711 mode and flags can be read.
8714 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8715 This macro should be provided on machines where the addresses
8716 in a dispatch table are absolute.
8718 The definition should be a C statement to output to the stdio stream
8719 @var{stream} an assembler pseudo-instruction to generate a reference to
8720 a label. @var{value} is the number of an internal label whose
8721 definition is output using @code{(*targetm.asm_out.internal_label)}.
8725 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8729 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8730 Define this if the label before a jump-table needs to be output
8731 specially. The first three arguments are the same as for
8732 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8733 jump-table which follows (a @code{jump_insn} containing an
8734 @code{addr_vec} or @code{addr_diff_vec}).
8736 This feature is used on system V to output a @code{swbeg} statement
8739 If this macro is not defined, these labels are output with
8740 @code{(*targetm.asm_out.internal_label)}.
8743 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8744 Define this if something special must be output at the end of a
8745 jump-table. The definition should be a C statement to be executed
8746 after the assembler code for the table is written. It should write
8747 the appropriate code to stdio stream @var{stream}. The argument
8748 @var{table} is the jump-table insn, and @var{num} is the label-number
8749 of the preceding label.
8751 If this macro is not defined, nothing special is output at the end of
8755 @hook TARGET_ASM_EMIT_UNWIND_LABEL
8756 This target hook emits a label at the beginning of each FDE@. It
8757 should be defined on targets where FDEs need special labels, and it
8758 should write the appropriate label, for the FDE associated with the
8759 function declaration @var{decl}, to the stdio stream @var{stream}.
8760 The third argument, @var{for_eh}, is a boolean: true if this is for an
8761 exception table. The fourth argument, @var{empty}, is a boolean:
8762 true if this is a placeholder label for an omitted FDE@.
8764 The default is that FDEs are not given nonlocal labels.
8767 @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
8768 This target hook emits a label at the beginning of the exception table.
8769 It should be defined on targets where it is desirable for the table
8770 to be broken up according to function.
8772 The default is that no label is emitted.
8775 @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
8777 @hook TARGET_ASM_UNWIND_EMIT
8778 This target hook emits assembly directives required to unwind the
8779 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8780 returns @code{UI_TARGET}.
8783 @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8785 @node Exception Region Output
8786 @subsection Assembler Commands for Exception Regions
8788 @c prevent bad page break with this line
8790 This describes commands marking the start and the end of an exception
8793 @defmac EH_FRAME_SECTION_NAME
8794 If defined, a C string constant for the name of the section containing
8795 exception handling frame unwind information. If not defined, GCC will
8796 provide a default definition if the target supports named sections.
8797 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8799 You should define this symbol if your target supports DWARF 2 frame
8800 unwind information and the default definition does not work.
8803 @defmac EH_FRAME_IN_DATA_SECTION
8804 If defined, DWARF 2 frame unwind information will be placed in the
8805 data section even though the target supports named sections. This
8806 might be necessary, for instance, if the system linker does garbage
8807 collection and sections cannot be marked as not to be collected.
8809 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8813 @defmac EH_TABLES_CAN_BE_READ_ONLY
8814 Define this macro to 1 if your target is such that no frame unwind
8815 information encoding used with non-PIC code will ever require a
8816 runtime relocation, but the linker may not support merging read-only
8817 and read-write sections into a single read-write section.
8820 @defmac MASK_RETURN_ADDR
8821 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8822 that it does not contain any extraneous set bits in it.
8825 @defmac DWARF2_UNWIND_INFO
8826 Define this macro to 0 if your target supports DWARF 2 frame unwind
8827 information, but it does not yet work with exception handling.
8828 Otherwise, if your target supports this information (if it defines
8829 @code{INCOMING_RETURN_ADDR_RTX} and either @code{UNALIGNED_INT_ASM_OP}
8830 or @code{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8833 @hook TARGET_EXCEPT_UNWIND_INFO
8834 This hook defines the mechanism that will be used for exception handling
8835 by the target. If the target has ABI specified unwind tables, the hook
8836 should return @code{UI_TARGET}. If the target is to use the
8837 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8838 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8839 information, the hook should return @code{UI_DWARF2}.
8841 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8842 This may end up simplifying other parts of target-specific code. The
8843 default implementation of this hook never returns @code{UI_NONE}.
8845 Note that the value returned by this hook should be constant. It should
8846 not depend on anything except the command-line switches described by
8847 @var{opts}. In particular, the
8848 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8849 macros and builtin functions related to exception handling are set up
8850 depending on this setting.
8852 The default implementation of the hook first honors the
8853 @option{--enable-sjlj-exceptions} configure option, then
8854 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
8855 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
8856 must define this hook so that @var{opts} is used correctly.
8859 @hook TARGET_UNWIND_TABLES_DEFAULT
8860 This variable should be set to @code{true} if the target ABI requires unwinding
8861 tables even when exceptions are not used. It must not be modified by
8862 command-line option processing.
8865 @defmac DONT_USE_BUILTIN_SETJMP
8866 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8867 should use the @code{setjmp}/@code{longjmp} functions from the C library
8868 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8871 @defmac DWARF_CIE_DATA_ALIGNMENT
8872 This macro need only be defined if the target might save registers in the
8873 function prologue at an offset to the stack pointer that is not aligned to
8874 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8875 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8876 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8877 the target supports DWARF 2 frame unwind information.
8880 @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
8881 Contains the value true if the target should add a zero word onto the
8882 end of a Dwarf-2 frame info section when used for exception handling.
8883 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8887 @hook TARGET_DWARF_REGISTER_SPAN
8888 Given a register, this hook should return a parallel of registers to
8889 represent where to find the register pieces. Define this hook if the
8890 register and its mode are represented in Dwarf in non-contiguous
8891 locations, or if the register should be represented in more than one
8892 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8893 If not defined, the default is to return @code{NULL_RTX}.
8896 @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
8897 If some registers are represented in Dwarf-2 unwind information in
8898 multiple pieces, define this hook to fill in information about the
8899 sizes of those pieces in the table used by the unwinder at runtime.
8900 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8901 filling in a single size corresponding to each hard register;
8902 @var{address} is the address of the table.
8905 @hook TARGET_ASM_TTYPE
8906 This hook is used to output a reference from a frame unwinding table to
8907 the type_info object identified by @var{sym}. It should return @code{true}
8908 if the reference was output. Returning @code{false} will cause the
8909 reference to be output using the normal Dwarf2 routines.
8912 @hook TARGET_ARM_EABI_UNWINDER
8913 This flag should be set to @code{true} on targets that use an ARM EABI
8914 based unwinding library, and @code{false} on other targets. This effects
8915 the format of unwinding tables, and how the unwinder in entered after
8916 running a cleanup. The default is @code{false}.
8919 @node Alignment Output
8920 @subsection Assembler Commands for Alignment
8922 @c prevent bad page break with this line
8923 This describes commands for alignment.
8925 @defmac JUMP_ALIGN (@var{label})
8926 The alignment (log base 2) to put in front of @var{label}, which is
8927 a common destination of jumps and has no fallthru incoming edge.
8929 This macro need not be defined if you don't want any special alignment
8930 to be done at such a time. Most machine descriptions do not currently
8933 Unless it's necessary to inspect the @var{label} parameter, it is better
8934 to set the variable @var{align_jumps} in the target's
8935 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8936 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8939 @hook TARGET_ASM_JUMP_ALIGN_MAX_SKIP
8940 The maximum number of bytes to skip before @var{label} when applying
8941 @code{JUMP_ALIGN}. This works only if
8942 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8945 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8946 The alignment (log base 2) to put in front of @var{label}, which follows
8949 This macro need not be defined if you don't want any special alignment
8950 to be done at such a time. Most machine descriptions do not currently
8954 @hook TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8955 The maximum number of bytes to skip before @var{label} when applying
8956 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8957 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8960 @defmac LOOP_ALIGN (@var{label})
8961 The alignment (log base 2) to put in front of @var{label}, which follows
8962 a @code{NOTE_INSN_LOOP_BEG} note.
8964 This macro need not be defined if you don't want any special alignment
8965 to be done at such a time. Most machine descriptions do not currently
8968 Unless it's necessary to inspect the @var{label} parameter, it is better
8969 to set the variable @code{align_loops} in the target's
8970 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8971 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8974 @hook TARGET_ASM_LOOP_ALIGN_MAX_SKIP
8975 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
8976 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
8980 @defmac LABEL_ALIGN (@var{label})
8981 The alignment (log base 2) to put in front of @var{label}.
8982 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8983 the maximum of the specified values is used.
8985 Unless it's necessary to inspect the @var{label} parameter, it is better
8986 to set the variable @code{align_labels} in the target's
8987 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8988 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8991 @hook TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8992 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
8993 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
8997 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8998 A C statement to output to the stdio stream @var{stream} an assembler
8999 instruction to advance the location counter by @var{nbytes} bytes.
9000 Those bytes should be zero when loaded. @var{nbytes} will be a C
9001 expression of type @code{unsigned HOST_WIDE_INT}.
9004 @defmac ASM_NO_SKIP_IN_TEXT
9005 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9006 text section because it fails to put zeros in the bytes that are skipped.
9007 This is true on many Unix systems, where the pseudo--op to skip bytes
9008 produces no-op instructions rather than zeros when used in the text
9012 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9013 A C statement to output to the stdio stream @var{stream} an assembler
9014 command to advance the location counter to a multiple of 2 to the
9015 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9018 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9019 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9020 for padding, if necessary.
9023 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9024 A C statement to output to the stdio stream @var{stream} an assembler
9025 command to advance the location counter to a multiple of 2 to the
9026 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9027 satisfy the alignment request. @var{power} and @var{max_skip} will be
9028 a C expression of type @code{int}.
9032 @node Debugging Info
9033 @section Controlling Debugging Information Format
9035 @c prevent bad page break with this line
9036 This describes how to specify debugging information.
9039 * All Debuggers:: Macros that affect all debugging formats uniformly.
9040 * DBX Options:: Macros enabling specific options in DBX format.
9041 * DBX Hooks:: Hook macros for varying DBX format.
9042 * File Names and DBX:: Macros controlling output of file names in DBX format.
9043 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9044 * VMS Debug:: Macros for VMS debug format.
9048 @subsection Macros Affecting All Debugging Formats
9050 @c prevent bad page break with this line
9051 These macros affect all debugging formats.
9053 @defmac DBX_REGISTER_NUMBER (@var{regno})
9054 A C expression that returns the DBX register number for the compiler
9055 register number @var{regno}. In the default macro provided, the value
9056 of this expression will be @var{regno} itself. But sometimes there are
9057 some registers that the compiler knows about and DBX does not, or vice
9058 versa. In such cases, some register may need to have one number in the
9059 compiler and another for DBX@.
9061 If two registers have consecutive numbers inside GCC, and they can be
9062 used as a pair to hold a multiword value, then they @emph{must} have
9063 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9064 Otherwise, debuggers will be unable to access such a pair, because they
9065 expect register pairs to be consecutive in their own numbering scheme.
9067 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9068 does not preserve register pairs, then what you must do instead is
9069 redefine the actual register numbering scheme.
9072 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9073 A C expression that returns the integer offset value for an automatic
9074 variable having address @var{x} (an RTL expression). The default
9075 computation assumes that @var{x} is based on the frame-pointer and
9076 gives the offset from the frame-pointer. This is required for targets
9077 that produce debugging output for DBX or COFF-style debugging output
9078 for SDB and allow the frame-pointer to be eliminated when the
9079 @option{-g} options is used.
9082 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9083 A C expression that returns the integer offset value for an argument
9084 having address @var{x} (an RTL expression). The nominal offset is
9088 @defmac PREFERRED_DEBUGGING_TYPE
9089 A C expression that returns the type of debugging output GCC should
9090 produce when the user specifies just @option{-g}. Define
9091 this if you have arranged for GCC to support more than one format of
9092 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9093 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9094 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9096 When the user specifies @option{-ggdb}, GCC normally also uses the
9097 value of this macro to select the debugging output format, but with two
9098 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9099 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9100 defined, GCC uses @code{DBX_DEBUG}.
9102 The value of this macro only affects the default debugging output; the
9103 user can always get a specific type of output by using @option{-gstabs},
9104 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9108 @subsection Specific Options for DBX Output
9110 @c prevent bad page break with this line
9111 These are specific options for DBX output.
9113 @defmac DBX_DEBUGGING_INFO
9114 Define this macro if GCC should produce debugging output for DBX
9115 in response to the @option{-g} option.
9118 @defmac XCOFF_DEBUGGING_INFO
9119 Define this macro if GCC should produce XCOFF format debugging output
9120 in response to the @option{-g} option. This is a variant of DBX format.
9123 @defmac DEFAULT_GDB_EXTENSIONS
9124 Define this macro to control whether GCC should by default generate
9125 GDB's extended version of DBX debugging information (assuming DBX-format
9126 debugging information is enabled at all). If you don't define the
9127 macro, the default is 1: always generate the extended information
9128 if there is any occasion to.
9131 @defmac DEBUG_SYMS_TEXT
9132 Define this macro if all @code{.stabs} commands should be output while
9133 in the text section.
9136 @defmac ASM_STABS_OP
9137 A C string constant, including spacing, naming the assembler pseudo op to
9138 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9139 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9140 applies only to DBX debugging information format.
9143 @defmac ASM_STABD_OP
9144 A C string constant, including spacing, naming the assembler pseudo op to
9145 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9146 value is the current location. If you don't define this macro,
9147 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9151 @defmac ASM_STABN_OP
9152 A C string constant, including spacing, naming the assembler pseudo op to
9153 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9154 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9155 macro applies only to DBX debugging information format.
9158 @defmac DBX_NO_XREFS
9159 Define this macro if DBX on your system does not support the construct
9160 @samp{xs@var{tagname}}. On some systems, this construct is used to
9161 describe a forward reference to a structure named @var{tagname}.
9162 On other systems, this construct is not supported at all.
9165 @defmac DBX_CONTIN_LENGTH
9166 A symbol name in DBX-format debugging information is normally
9167 continued (split into two separate @code{.stabs} directives) when it
9168 exceeds a certain length (by default, 80 characters). On some
9169 operating systems, DBX requires this splitting; on others, splitting
9170 must not be done. You can inhibit splitting by defining this macro
9171 with the value zero. You can override the default splitting-length by
9172 defining this macro as an expression for the length you desire.
9175 @defmac DBX_CONTIN_CHAR
9176 Normally continuation is indicated by adding a @samp{\} character to
9177 the end of a @code{.stabs} string when a continuation follows. To use
9178 a different character instead, define this macro as a character
9179 constant for the character you want to use. Do not define this macro
9180 if backslash is correct for your system.
9183 @defmac DBX_STATIC_STAB_DATA_SECTION
9184 Define this macro if it is necessary to go to the data section before
9185 outputting the @samp{.stabs} pseudo-op for a non-global static
9189 @defmac DBX_TYPE_DECL_STABS_CODE
9190 The value to use in the ``code'' field of the @code{.stabs} directive
9191 for a typedef. The default is @code{N_LSYM}.
9194 @defmac DBX_STATIC_CONST_VAR_CODE
9195 The value to use in the ``code'' field of the @code{.stabs} directive
9196 for a static variable located in the text section. DBX format does not
9197 provide any ``right'' way to do this. The default is @code{N_FUN}.
9200 @defmac DBX_REGPARM_STABS_CODE
9201 The value to use in the ``code'' field of the @code{.stabs} directive
9202 for a parameter passed in registers. DBX format does not provide any
9203 ``right'' way to do this. The default is @code{N_RSYM}.
9206 @defmac DBX_REGPARM_STABS_LETTER
9207 The letter to use in DBX symbol data to identify a symbol as a parameter
9208 passed in registers. DBX format does not customarily provide any way to
9209 do this. The default is @code{'P'}.
9212 @defmac DBX_FUNCTION_FIRST
9213 Define this macro if the DBX information for a function and its
9214 arguments should precede the assembler code for the function. Normally,
9215 in DBX format, the debugging information entirely follows the assembler
9219 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9220 Define this macro, with value 1, if the value of a symbol describing
9221 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9222 relative to the start of the enclosing function. Normally, GCC uses
9223 an absolute address.
9226 @defmac DBX_LINES_FUNCTION_RELATIVE
9227 Define this macro, with value 1, if the value of a symbol indicating
9228 the current line number (@code{N_SLINE}) should be relative to the
9229 start of the enclosing function. Normally, GCC uses an absolute address.
9232 @defmac DBX_USE_BINCL
9233 Define this macro if GCC should generate @code{N_BINCL} and
9234 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9235 macro also directs GCC to output a type number as a pair of a file
9236 number and a type number within the file. Normally, GCC does not
9237 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9238 number for a type number.
9242 @subsection Open-Ended Hooks for DBX Format
9244 @c prevent bad page break with this line
9245 These are hooks for DBX format.
9247 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9248 Define this macro to say how to output to @var{stream} the debugging
9249 information for the start of a scope level for variable names. The
9250 argument @var{name} is the name of an assembler symbol (for use with
9251 @code{assemble_name}) whose value is the address where the scope begins.
9254 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9255 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9258 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9259 Define this macro if the target machine requires special handling to
9260 output an @code{N_FUN} entry for the function @var{decl}.
9263 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9264 A C statement to output DBX debugging information before code for line
9265 number @var{line} of the current source file to the stdio stream
9266 @var{stream}. @var{counter} is the number of time the macro was
9267 invoked, including the current invocation; it is intended to generate
9268 unique labels in the assembly output.
9270 This macro should not be defined if the default output is correct, or
9271 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9274 @defmac NO_DBX_FUNCTION_END
9275 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9276 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9277 On those machines, define this macro to turn this feature off without
9278 disturbing the rest of the gdb extensions.
9281 @defmac NO_DBX_BNSYM_ENSYM
9282 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9283 extension construct. On those machines, define this macro to turn this
9284 feature off without disturbing the rest of the gdb extensions.
9287 @node File Names and DBX
9288 @subsection File Names in DBX Format
9290 @c prevent bad page break with this line
9291 This describes file names in DBX format.
9293 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9294 A C statement to output DBX debugging information to the stdio stream
9295 @var{stream}, which indicates that file @var{name} is the main source
9296 file---the file specified as the input file for compilation.
9297 This macro is called only once, at the beginning of compilation.
9299 This macro need not be defined if the standard form of output
9300 for DBX debugging information is appropriate.
9302 It may be necessary to refer to a label equal to the beginning of the
9303 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9304 to do so. If you do this, you must also set the variable
9305 @var{used_ltext_label_name} to @code{true}.
9308 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9309 Define this macro, with value 1, if GCC should not emit an indication
9310 of the current directory for compilation and current source language at
9311 the beginning of the file.
9314 @defmac NO_DBX_GCC_MARKER
9315 Define this macro, with value 1, if GCC should not emit an indication
9316 that this object file was compiled by GCC@. The default is to emit
9317 an @code{N_OPT} stab at the beginning of every source file, with
9318 @samp{gcc2_compiled.} for the string and value 0.
9321 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9322 A C statement to output DBX debugging information at the end of
9323 compilation of the main source file @var{name}. Output should be
9324 written to the stdio stream @var{stream}.
9326 If you don't define this macro, nothing special is output at the end
9327 of compilation, which is correct for most machines.
9330 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9331 Define this macro @emph{instead of} defining
9332 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9333 the end of compilation is an @code{N_SO} stab with an empty string,
9334 whose value is the highest absolute text address in the file.
9339 @subsection Macros for SDB and DWARF Output
9341 @c prevent bad page break with this line
9342 Here are macros for SDB and DWARF output.
9344 @defmac SDB_DEBUGGING_INFO
9345 Define this macro if GCC should produce COFF-style debugging output
9346 for SDB in response to the @option{-g} option.
9349 @defmac DWARF2_DEBUGGING_INFO
9350 Define this macro if GCC should produce dwarf version 2 format
9351 debugging output in response to the @option{-g} option.
9353 @hook TARGET_DWARF_CALLING_CONVENTION
9354 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9355 be emitted for each function. Instead of an integer return the enum
9356 value for the @code{DW_CC_} tag.
9359 To support optional call frame debugging information, you must also
9360 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9361 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9362 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9363 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9366 @defmac DWARF2_FRAME_INFO
9367 Define this macro to a nonzero value if GCC should always output
9368 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9369 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9370 exceptions are enabled, GCC will output this information not matter
9371 how you define @code{DWARF2_FRAME_INFO}.
9374 @hook TARGET_DEBUG_UNWIND_INFO
9375 This hook defines the mechanism that will be used for describing frame
9376 unwind information to the debugger. Normally the hook will return
9377 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9378 return @code{UI_NONE} otherwise.
9380 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9381 is disabled in order to always output DWARF 2 frame information.
9383 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9384 This will suppress generation of the normal debug frame unwind information.
9387 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9388 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9389 line debug info sections. This will result in much more compact line number
9390 tables, and hence is desirable if it works.
9393 @hook TARGET_WANT_DEBUG_PUB_SECTIONS
9395 @hook TARGET_FORCE_AT_COMP_DIR
9397 @hook TARGET_DELAY_SCHED2
9399 @hook TARGET_DELAY_VARTRACK
9401 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9402 A C statement to issue assembly directives that create a difference
9403 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9406 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9407 A C statement to issue assembly directives that create a difference
9408 between the two given labels in system defined units, e.g. instruction
9409 slots on IA64 VMS, using an integer of the given size.
9412 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9413 A C statement to issue assembly directives that create a
9414 section-relative reference to the given @var{label}, using an integer of the
9415 given @var{size}. The label is known to be defined in the given @var{section}.
9418 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9419 A C statement to issue assembly directives that create a self-relative
9420 reference to the given @var{label}, using an integer of the given @var{size}.
9423 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9424 A C statement to issue assembly directives that create a reference to
9425 the DWARF table identifier @var{label} from the current section. This
9426 is used on some systems to avoid garbage collecting a DWARF table which
9427 is referenced by a function.
9430 @hook TARGET_ASM_OUTPUT_DWARF_DTPREL
9431 If defined, this target hook is a function which outputs a DTP-relative
9432 reference to the given TLS symbol of the specified size.
9435 @defmac PUT_SDB_@dots{}
9436 Define these macros to override the assembler syntax for the special
9437 SDB assembler directives. See @file{sdbout.c} for a list of these
9438 macros and their arguments. If the standard syntax is used, you need
9439 not define them yourself.
9443 Some assemblers do not support a semicolon as a delimiter, even between
9444 SDB assembler directives. In that case, define this macro to be the
9445 delimiter to use (usually @samp{\n}). It is not necessary to define
9446 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9450 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9451 Define this macro to allow references to unknown structure,
9452 union, or enumeration tags to be emitted. Standard COFF does not
9453 allow handling of unknown references, MIPS ECOFF has support for
9457 @defmac SDB_ALLOW_FORWARD_REFERENCES
9458 Define this macro to allow references to structure, union, or
9459 enumeration tags that have not yet been seen to be handled. Some
9460 assemblers choke if forward tags are used, while some require it.
9463 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9464 A C statement to output SDB debugging information before code for line
9465 number @var{line} of the current source file to the stdio stream
9466 @var{stream}. The default is to emit an @code{.ln} directive.
9471 @subsection Macros for VMS Debug Format
9473 @c prevent bad page break with this line
9474 Here are macros for VMS debug format.
9476 @defmac VMS_DEBUGGING_INFO
9477 Define this macro if GCC should produce debugging output for VMS
9478 in response to the @option{-g} option. The default behavior for VMS
9479 is to generate minimal debug info for a traceback in the absence of
9480 @option{-g} unless explicitly overridden with @option{-g0}. This
9481 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9482 @code{TARGET_OPTION_OVERRIDE}.
9485 @node Floating Point
9486 @section Cross Compilation and Floating Point
9487 @cindex cross compilation and floating point
9488 @cindex floating point and cross compilation
9490 While all modern machines use twos-complement representation for integers,
9491 there are a variety of representations for floating point numbers. This
9492 means that in a cross-compiler the representation of floating point numbers
9493 in the compiled program may be different from that used in the machine
9494 doing the compilation.
9496 Because different representation systems may offer different amounts of
9497 range and precision, all floating point constants must be represented in
9498 the target machine's format. Therefore, the cross compiler cannot
9499 safely use the host machine's floating point arithmetic; it must emulate
9500 the target's arithmetic. To ensure consistency, GCC always uses
9501 emulation to work with floating point values, even when the host and
9502 target floating point formats are identical.
9504 The following macros are provided by @file{real.h} for the compiler to
9505 use. All parts of the compiler which generate or optimize
9506 floating-point calculations must use these macros. They may evaluate
9507 their operands more than once, so operands must not have side effects.
9509 @defmac REAL_VALUE_TYPE
9510 The C data type to be used to hold a floating point value in the target
9511 machine's format. Typically this is a @code{struct} containing an
9512 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9516 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9517 Compares for equality the two values, @var{x} and @var{y}. If the target
9518 floating point format supports negative zeroes and/or NaNs,
9519 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9520 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9523 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9524 Tests whether @var{x} is less than @var{y}.
9527 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9528 Truncates @var{x} to a signed integer, rounding toward zero.
9531 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9532 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9533 @var{x} is negative, returns zero.
9536 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9537 Converts @var{string} into a floating point number in the target machine's
9538 representation for mode @var{mode}. This routine can handle both
9539 decimal and hexadecimal floating point constants, using the syntax
9540 defined by the C language for both.
9543 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9544 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9547 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9548 Determines whether @var{x} represents infinity (positive or negative).
9551 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9552 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9555 @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})
9556 Calculates an arithmetic operation on the two floating point values
9557 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9560 The operation to be performed is specified by @var{code}. Only the
9561 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9562 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9564 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9565 target's floating point format cannot represent infinity, it will call
9566 @code{abort}. Callers should check for this situation first, using
9567 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9570 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9571 Returns the negative of the floating point value @var{x}.
9574 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9575 Returns the absolute value of @var{x}.
9578 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9579 Truncates the floating point value @var{x} to fit in @var{mode}. The
9580 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9581 appropriate bit pattern to be output as a floating constant whose
9582 precision accords with mode @var{mode}.
9585 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9586 Converts a floating point value @var{x} into a double-precision integer
9587 which is then stored into @var{low} and @var{high}. If the value is not
9588 integral, it is truncated.
9591 @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})
9592 Converts a double-precision integer found in @var{low} and @var{high},
9593 into a floating point value which is then stored into @var{x}. The
9594 value is truncated to fit in mode @var{mode}.
9597 @node Mode Switching
9598 @section Mode Switching Instructions
9599 @cindex mode switching
9600 The following macros control mode switching optimizations:
9602 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9603 Define this macro if the port needs extra instructions inserted for mode
9604 switching in an optimizing compilation.
9606 For an example, the SH4 can perform both single and double precision
9607 floating point operations, but to perform a single precision operation,
9608 the FPSCR PR bit has to be cleared, while for a double precision
9609 operation, this bit has to be set. Changing the PR bit requires a general
9610 purpose register as a scratch register, hence these FPSCR sets have to
9611 be inserted before reload, i.e.@: you can't put this into instruction emitting
9612 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9614 You can have multiple entities that are mode-switched, and select at run time
9615 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9616 return nonzero for any @var{entity} that needs mode-switching.
9617 If you define this macro, you also have to define
9618 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9619 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9620 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9624 @defmac NUM_MODES_FOR_MODE_SWITCHING
9625 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9626 initializer for an array of integers. Each initializer element
9627 N refers to an entity that needs mode switching, and specifies the number
9628 of different modes that might need to be set for this entity.
9629 The position of the initializer in the initializer---starting counting at
9630 zero---determines the integer that is used to refer to the mode-switched
9632 In macros that take mode arguments / yield a mode result, modes are
9633 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9634 switch is needed / supplied.
9637 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9638 @var{entity} is an integer specifying a mode-switched entity. If
9639 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9640 return an integer value not larger than the corresponding element in
9641 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9642 be switched into prior to the execution of @var{insn}.
9645 @defmac MODE_AFTER (@var{mode}, @var{insn})
9646 If this macro is defined, it is evaluated for every @var{insn} during
9647 mode switching. It determines the mode that an insn results in (if
9648 different from the incoming mode).
9651 @defmac MODE_ENTRY (@var{entity})
9652 If this macro is defined, it is evaluated for every @var{entity} that needs
9653 mode switching. It should evaluate to an integer, which is a mode that
9654 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9655 is defined then @code{MODE_EXIT} must be defined.
9658 @defmac MODE_EXIT (@var{entity})
9659 If this macro is defined, it is evaluated for every @var{entity} that needs
9660 mode switching. It should evaluate to an integer, which is a mode that
9661 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9662 is defined then @code{MODE_ENTRY} must be defined.
9665 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9666 This macro specifies the order in which modes for @var{entity} are processed.
9667 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9668 lowest. The value of the macro should be an integer designating a mode
9669 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9670 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9671 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9674 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9675 Generate one or more insns to set @var{entity} to @var{mode}.
9676 @var{hard_reg_live} is the set of hard registers live at the point where
9677 the insn(s) are to be inserted.
9680 @node Target Attributes
9681 @section Defining target-specific uses of @code{__attribute__}
9682 @cindex target attributes
9683 @cindex machine attributes
9684 @cindex attributes, target-specific
9686 Target-specific attributes may be defined for functions, data and types.
9687 These are described using the following target hooks; they also need to
9688 be documented in @file{extend.texi}.
9690 @hook TARGET_ATTRIBUTE_TABLE
9691 If defined, this target hook points to an array of @samp{struct
9692 attribute_spec} (defined in @file{tree.h}) specifying the machine
9693 specific attributes for this target and some of the restrictions on the
9694 entities to which these attributes are applied and the arguments they
9698 @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
9699 If defined, this target hook is a function which returns true if the
9700 machine-specific attribute named @var{name} expects an identifier
9701 given as its first argument to be passed on as a plain identifier, not
9702 subjected to name lookup. If this is not defined, the default is
9703 false for all machine-specific attributes.
9706 @hook TARGET_COMP_TYPE_ATTRIBUTES
9707 If defined, this target hook is a function which returns zero if the attributes on
9708 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9709 and two if they are nearly compatible (which causes a warning to be
9710 generated). If this is not defined, machine-specific attributes are
9711 supposed always to be compatible.
9714 @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
9715 If defined, this target hook is a function which assigns default attributes to
9716 the newly defined @var{type}.
9719 @hook TARGET_MERGE_TYPE_ATTRIBUTES
9720 Define this target hook if the merging of type attributes needs special
9721 handling. If defined, the result is a list of the combined
9722 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9723 that @code{comptypes} has already been called and returned 1. This
9724 function may call @code{merge_attributes} to handle machine-independent
9728 @hook TARGET_MERGE_DECL_ATTRIBUTES
9729 Define this target hook if the merging of decl attributes needs special
9730 handling. If defined, the result is a list of the combined
9731 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9732 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9733 when this is needed are when one attribute overrides another, or when an
9734 attribute is nullified by a subsequent definition. This function may
9735 call @code{merge_attributes} to handle machine-independent merging.
9737 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9738 If the only target-specific handling you require is @samp{dllimport}
9739 for Microsoft Windows targets, you should define the macro
9740 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9741 will then define a function called
9742 @code{merge_dllimport_decl_attributes} which can then be defined as
9743 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9744 add @code{handle_dll_attribute} in the attribute table for your port
9745 to perform initial processing of the @samp{dllimport} and
9746 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9747 @file{i386/i386.c}, for example.
9750 @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
9752 @defmac TARGET_DECLSPEC
9753 Define this macro to a nonzero value if you want to treat
9754 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9755 default, this behavior is enabled only for targets that define
9756 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9757 of @code{__declspec} is via a built-in macro, but you should not rely
9758 on this implementation detail.
9761 @hook TARGET_INSERT_ATTRIBUTES
9762 Define this target hook if you want to be able to add attributes to a decl
9763 when it is being created. This is normally useful for back ends which
9764 wish to implement a pragma by using the attributes which correspond to
9765 the pragma's effect. The @var{node} argument is the decl which is being
9766 created. The @var{attr_ptr} argument is a pointer to the attribute list
9767 for this decl. The list itself should not be modified, since it may be
9768 shared with other decls, but attributes may be chained on the head of
9769 the list and @code{*@var{attr_ptr}} modified to point to the new
9770 attributes, or a copy of the list may be made if further changes are
9774 @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
9776 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9777 into the current function, despite its having target-specific
9778 attributes, @code{false} otherwise. By default, if a function has a
9779 target specific attribute attached to it, it will not be inlined.
9782 @hook TARGET_OPTION_VALID_ATTRIBUTE_P
9783 This hook is called to parse the @code{attribute(option("..."))}, and
9784 it allows the function to set different target machine compile time
9785 options for the current function that might be different than the
9786 options specified on the command line. The hook should return
9787 @code{true} if the options are valid.
9789 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9790 the function declaration to hold a pointer to a target specific
9791 @var{struct cl_target_option} structure.
9794 @hook TARGET_OPTION_SAVE
9795 This hook is called to save any additional target specific information
9796 in the @var{struct cl_target_option} structure for function specific
9798 @xref{Option file format}.
9801 @hook TARGET_OPTION_RESTORE
9802 This hook is called to restore any additional target specific
9803 information in the @var{struct cl_target_option} structure for
9804 function specific options.
9807 @hook TARGET_OPTION_PRINT
9808 This hook is called to print any additional target specific
9809 information in the @var{struct cl_target_option} structure for
9810 function specific options.
9813 @hook TARGET_OPTION_PRAGMA_PARSE
9814 This target hook parses the options for @code{#pragma GCC option} to
9815 set the machine specific options for functions that occur later in the
9816 input stream. The options should be the same as handled by the
9817 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9820 @hook TARGET_OPTION_OVERRIDE
9821 Sometimes certain combinations of command options do not make sense on
9822 a particular target machine. You can override the hook
9823 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9824 once just after all the command options have been parsed.
9826 Don't use this hook to turn on various extra optimizations for
9827 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9829 If you need to do something whenever the optimization level is
9830 changed via the optimize attribute or pragma, see
9831 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9834 @hook TARGET_CAN_INLINE_P
9835 This target hook returns @code{false} if the @var{caller} function
9836 cannot inline @var{callee}, based on target specific information. By
9837 default, inlining is not allowed if the callee function has function
9838 specific target options and the caller does not use the same options.
9842 @section Emulating TLS
9843 @cindex Emulated TLS
9845 For targets whose psABI does not provide Thread Local Storage via
9846 specific relocations and instruction sequences, an emulation layer is
9847 used. A set of target hooks allows this emulation layer to be
9848 configured for the requirements of a particular target. For instance
9849 the psABI may in fact specify TLS support in terms of an emulation
9852 The emulation layer works by creating a control object for every TLS
9853 object. To access the TLS object, a lookup function is provided
9854 which, when given the address of the control object, will return the
9855 address of the current thread's instance of the TLS object.
9857 @hook TARGET_EMUTLS_GET_ADDRESS
9858 Contains the name of the helper function that uses a TLS control
9859 object to locate a TLS instance. The default causes libgcc's
9860 emulated TLS helper function to be used.
9863 @hook TARGET_EMUTLS_REGISTER_COMMON
9864 Contains the name of the helper function that should be used at
9865 program startup to register TLS objects that are implicitly
9866 initialized to zero. If this is @code{NULL}, all TLS objects will
9867 have explicit initializers. The default causes libgcc's emulated TLS
9868 registration function to be used.
9871 @hook TARGET_EMUTLS_VAR_SECTION
9872 Contains the name of the section in which TLS control variables should
9873 be placed. The default of @code{NULL} allows these to be placed in
9877 @hook TARGET_EMUTLS_TMPL_SECTION
9878 Contains the name of the section in which TLS initializers should be
9879 placed. The default of @code{NULL} allows these to be placed in any
9883 @hook TARGET_EMUTLS_VAR_PREFIX
9884 Contains the prefix to be prepended to TLS control variable names.
9885 The default of @code{NULL} uses a target-specific prefix.
9888 @hook TARGET_EMUTLS_TMPL_PREFIX
9889 Contains the prefix to be prepended to TLS initializer objects. The
9890 default of @code{NULL} uses a target-specific prefix.
9893 @hook TARGET_EMUTLS_VAR_FIELDS
9894 Specifies a function that generates the FIELD_DECLs for a TLS control
9895 object type. @var{type} is the RECORD_TYPE the fields are for and
9896 @var{name} should be filled with the structure tag, if the default of
9897 @code{__emutls_object} is unsuitable. The default creates a type suitable
9898 for libgcc's emulated TLS function.
9901 @hook TARGET_EMUTLS_VAR_INIT
9902 Specifies a function that generates the CONSTRUCTOR to initialize a
9903 TLS control object. @var{var} is the TLS control object, @var{decl}
9904 is the TLS object and @var{tmpl_addr} is the address of the
9905 initializer. The default initializes libgcc's emulated TLS control object.
9908 @hook TARGET_EMUTLS_VAR_ALIGN_FIXED
9909 Specifies whether the alignment of TLS control variable objects is
9910 fixed and should not be increased as some backends may do to optimize
9911 single objects. The default is false.
9914 @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9915 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9916 may be used to describe emulated TLS control objects.
9919 @node MIPS Coprocessors
9920 @section Defining coprocessor specifics for MIPS targets.
9921 @cindex MIPS coprocessor-definition macros
9923 The MIPS specification allows MIPS implementations to have as many as 4
9924 coprocessors, each with as many as 32 private registers. GCC supports
9925 accessing these registers and transferring values between the registers
9926 and memory using asm-ized variables. For example:
9929 register unsigned int cp0count asm ("c0r1");
9935 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9936 names may be added as described below, or the default names may be
9937 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9939 Coprocessor registers are assumed to be epilogue-used; sets to them will
9940 be preserved even if it does not appear that the register is used again
9941 later in the function.
9943 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9944 the FPU@. One accesses COP1 registers through standard mips
9945 floating-point support; they are not included in this mechanism.
9947 There is one macro used in defining the MIPS coprocessor interface which
9948 you may want to override in subtargets; it is described below.
9950 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9951 A comma-separated list (with leading comma) of pairs describing the
9952 alternate names of coprocessor registers. The format of each entry should be
9954 @{ @var{alternatename}, @var{register_number}@}
9960 @section Parameters for Precompiled Header Validity Checking
9961 @cindex parameters, precompiled headers
9963 @hook TARGET_GET_PCH_VALIDITY
9964 This hook returns a pointer to the data needed by
9965 @code{TARGET_PCH_VALID_P} and sets
9966 @samp{*@var{sz}} to the size of the data in bytes.
9969 @hook TARGET_PCH_VALID_P
9970 This hook checks whether the options used to create a PCH file are
9971 compatible with the current settings. It returns @code{NULL}
9972 if so and a suitable error message if not. Error messages will
9973 be presented to the user and must be localized using @samp{_(@var{msg})}.
9975 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9976 when the PCH file was created and @var{sz} is the size of that data in bytes.
9977 It's safe to assume that the data was created by the same version of the
9978 compiler, so no format checking is needed.
9980 The default definition of @code{default_pch_valid_p} should be
9981 suitable for most targets.
9984 @hook TARGET_CHECK_PCH_TARGET_FLAGS
9985 If this hook is nonnull, the default implementation of
9986 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9987 of @code{target_flags}. @var{pch_flags} specifies the value that
9988 @code{target_flags} had when the PCH file was created. The return
9989 value is the same as for @code{TARGET_PCH_VALID_P}.
9992 @hook TARGET_PREPARE_PCH_SAVE
9995 @section C++ ABI parameters
9996 @cindex parameters, c++ abi
9998 @hook TARGET_CXX_GUARD_TYPE
9999 Define this hook to override the integer type used for guard variables.
10000 These are used to implement one-time construction of static objects. The
10001 default is long_long_integer_type_node.
10004 @hook TARGET_CXX_GUARD_MASK_BIT
10005 This hook determines how guard variables are used. It should return
10006 @code{false} (the default) if the first byte should be used. A return value of
10007 @code{true} indicates that only the least significant bit should be used.
10010 @hook TARGET_CXX_GET_COOKIE_SIZE
10011 This hook returns the size of the cookie to use when allocating an array
10012 whose elements have the indicated @var{type}. Assumes that it is already
10013 known that a cookie is needed. The default is
10014 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10015 IA64/Generic C++ ABI@.
10018 @hook TARGET_CXX_COOKIE_HAS_SIZE
10019 This hook should return @code{true} if the element size should be stored in
10020 array cookies. The default is to return @code{false}.
10023 @hook TARGET_CXX_IMPORT_EXPORT_CLASS
10024 If defined by a backend this hook allows the decision made to export
10025 class @var{type} to be overruled. Upon entry @var{import_export}
10026 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10027 to be imported and 0 otherwise. This function should return the
10028 modified value and perform any other actions necessary to support the
10029 backend's targeted operating system.
10032 @hook TARGET_CXX_CDTOR_RETURNS_THIS
10033 This hook should return @code{true} if constructors and destructors return
10034 the address of the object created/destroyed. The default is to return
10038 @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
10039 This hook returns true if the key method for a class (i.e., the method
10040 which, if defined in the current translation unit, causes the virtual
10041 table to be emitted) may be an inline function. Under the standard
10042 Itanium C++ ABI the key method may be an inline function so long as
10043 the function is not declared inline in the class definition. Under
10044 some variants of the ABI, an inline function can never be the key
10045 method. The default is to return @code{true}.
10048 @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
10050 @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
10051 This hook returns true (the default) if virtual tables and other
10052 similar implicit class data objects are always COMDAT if they have
10053 external linkage. If this hook returns false, then class data for
10054 classes whose virtual table will be emitted in only one translation
10055 unit will not be COMDAT.
10058 @hook TARGET_CXX_LIBRARY_RTTI_COMDAT
10059 This hook returns true (the default) if the RTTI information for
10060 the basic types which is defined in the C++ runtime should always
10061 be COMDAT, false if it should not be COMDAT.
10064 @hook TARGET_CXX_USE_AEABI_ATEXIT
10065 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10066 should be used to register static destructors when @option{-fuse-cxa-atexit}
10067 is in effect. The default is to return false to use @code{__cxa_atexit}.
10070 @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
10071 This hook returns true if the target @code{atexit} function can be used
10072 in the same manner as @code{__cxa_atexit} to register C++ static
10073 destructors. This requires that @code{atexit}-registered functions in
10074 shared libraries are run in the correct order when the libraries are
10075 unloaded. The default is to return false.
10078 @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
10080 @hook TARGET_CXX_DECL_MANGLING_CONTEXT
10082 @node Named Address Spaces
10083 @section Adding support for named address spaces
10084 @cindex named address spaces
10086 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10087 standards committee, @cite{Programming Languages - C - Extensions to
10088 support embedded processors}, specifies a syntax for embedded
10089 processors to specify alternate address spaces. You can configure a
10090 GCC port to support section 5.1 of the draft report to add support for
10091 address spaces other than the default address space. These address
10092 spaces are new keywords that are similar to the @code{volatile} and
10093 @code{const} type attributes.
10095 Pointers to named address spaces can have a different size than
10096 pointers to the generic address space.
10098 For example, the SPU port uses the @code{__ea} address space to refer
10099 to memory in the host processor, rather than memory local to the SPU
10100 processor. Access to memory in the @code{__ea} address space involves
10101 issuing DMA operations to move data between the host processor and the
10102 local processor memory address space. Pointers in the @code{__ea}
10103 address space are either 32 bits or 64 bits based on the
10104 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10107 Internally, address spaces are represented as a small integer in the
10108 range 0 to 15 with address space 0 being reserved for the generic
10111 To register a named address space qualifier keyword with the C front end,
10112 the target may call the @code{c_register_addr_space} routine. For example,
10113 the SPU port uses the following to declare @code{__ea} as the keyword for
10114 named address space #1:
10116 #define ADDR_SPACE_EA 1
10117 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10120 @hook TARGET_ADDR_SPACE_POINTER_MODE
10121 Define this to return the machine mode to use for pointers to
10122 @var{address_space} if the target supports named address spaces.
10123 The default version of this hook returns @code{ptr_mode} for the
10124 generic address space only.
10127 @hook TARGET_ADDR_SPACE_ADDRESS_MODE
10128 Define this to return the machine mode to use for addresses in
10129 @var{address_space} if the target supports named address spaces.
10130 The default version of this hook returns @code{Pmode} for the
10131 generic address space only.
10134 @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
10135 Define this to return nonzero if the port can handle pointers
10136 with machine mode @var{mode} to address space @var{as}. This target
10137 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10138 except that it includes explicit named address space support. The default
10139 version of this hook returns true for the modes returned by either the
10140 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10141 target hooks for the given address space.
10144 @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
10145 Define this to return true if @var{exp} is a valid address for mode
10146 @var{mode} in the named address space @var{as}. The @var{strict}
10147 parameter says whether strict addressing is in effect after reload has
10148 finished. This target hook is the same as the
10149 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10150 explicit named address space support.
10153 @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
10154 Define this to modify an invalid address @var{x} to be a valid address
10155 with mode @var{mode} in the named address space @var{as}. This target
10156 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10157 except that it includes explicit named address space support.
10160 @hook TARGET_ADDR_SPACE_SUBSET_P
10161 Define this to return whether the @var{subset} named address space is
10162 contained within the @var{superset} named address space. Pointers to
10163 a named address space that is a subset of another named address space
10164 will be converted automatically without a cast if used together in
10165 arithmetic operations. Pointers to a superset address space can be
10166 converted to pointers to a subset address space via explicit casts.
10169 @hook TARGET_ADDR_SPACE_CONVERT
10170 Define this to convert the pointer expression represented by the RTL
10171 @var{op} with type @var{from_type} that points to a named address
10172 space to a new pointer expression with type @var{to_type} that points
10173 to a different named address space. When this hook it called, it is
10174 guaranteed that one of the two address spaces is a subset of the other,
10175 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10179 @section Miscellaneous Parameters
10180 @cindex parameters, miscellaneous
10182 @c prevent bad page break with this line
10183 Here are several miscellaneous parameters.
10185 @defmac HAS_LONG_COND_BRANCH
10186 Define this boolean macro to indicate whether or not your architecture
10187 has conditional branches that can span all of memory. It is used in
10188 conjunction with an optimization that partitions hot and cold basic
10189 blocks into separate sections of the executable. If this macro is
10190 set to false, gcc will convert any conditional branches that attempt
10191 to cross between sections into unconditional branches or indirect jumps.
10194 @defmac HAS_LONG_UNCOND_BRANCH
10195 Define this boolean macro to indicate whether or not your architecture
10196 has unconditional branches that can span all of memory. It is used in
10197 conjunction with an optimization that partitions hot and cold basic
10198 blocks into separate sections of the executable. If this macro is
10199 set to false, gcc will convert any unconditional branches that attempt
10200 to cross between sections into indirect jumps.
10203 @defmac CASE_VECTOR_MODE
10204 An alias for a machine mode name. This is the machine mode that
10205 elements of a jump-table should have.
10208 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10209 Optional: return the preferred mode for an @code{addr_diff_vec}
10210 when the minimum and maximum offset are known. If you define this,
10211 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10212 To make this work, you also have to define @code{INSN_ALIGN} and
10213 make the alignment for @code{addr_diff_vec} explicit.
10214 The @var{body} argument is provided so that the offset_unsigned and scale
10215 flags can be updated.
10218 @defmac CASE_VECTOR_PC_RELATIVE
10219 Define this macro to be a C expression to indicate when jump-tables
10220 should contain relative addresses. You need not define this macro if
10221 jump-tables never contain relative addresses, or jump-tables should
10222 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10226 @hook TARGET_CASE_VALUES_THRESHOLD
10227 This function return the smallest number of different values for which it
10228 is best to use a jump-table instead of a tree of conditional branches.
10229 The default is four for machines with a @code{casesi} instruction and
10230 five otherwise. This is best for most machines.
10233 @defmac CASE_USE_BIT_TESTS
10234 Define this macro to be a C expression to indicate whether C switch
10235 statements may be implemented by a sequence of bit tests. This is
10236 advantageous on processors that can efficiently implement left shift
10237 of 1 by the number of bits held in a register, but inappropriate on
10238 targets that would require a loop. By default, this macro returns
10239 @code{true} if the target defines an @code{ashlsi3} pattern, and
10240 @code{false} otherwise.
10243 @defmac WORD_REGISTER_OPERATIONS
10244 Define this macro if operations between registers with integral mode
10245 smaller than a word are always performed on the entire register.
10246 Most RISC machines have this property and most CISC machines do not.
10249 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10250 Define this macro to be a C expression indicating when insns that read
10251 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10252 bits outside of @var{mem_mode} to be either the sign-extension or the
10253 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10254 of @var{mem_mode} for which the
10255 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10256 @code{UNKNOWN} for other modes.
10258 This macro is not called with @var{mem_mode} non-integral or with a width
10259 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10260 value in this case. Do not define this macro if it would always return
10261 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10262 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10264 You may return a non-@code{UNKNOWN} value even if for some hard registers
10265 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10266 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10267 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10268 integral mode larger than this but not larger than @code{word_mode}.
10270 You must return @code{UNKNOWN} if for some hard registers that allow this
10271 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10272 @code{word_mode}, but that they can change to another integral mode that
10273 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10276 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10277 Define this macro if loading short immediate values into registers sign
10281 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
10282 Define this macro if the same instructions that convert a floating
10283 point number to a signed fixed point number also convert validly to an
10287 @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
10288 When @option{-ffast-math} is in effect, GCC tries to optimize
10289 divisions by the same divisor, by turning them into multiplications by
10290 the reciprocal. This target hook specifies the minimum number of divisions
10291 that should be there for GCC to perform the optimization for a variable
10292 of mode @var{mode}. The default implementation returns 3 if the machine
10293 has an instruction for the division, and 2 if it does not.
10297 The maximum number of bytes that a single instruction can move quickly
10298 between memory and registers or between two memory locations.
10301 @defmac MAX_MOVE_MAX
10302 The maximum number of bytes that a single instruction can move quickly
10303 between memory and registers or between two memory locations. If this
10304 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10305 constant value that is the largest value that @code{MOVE_MAX} can have
10309 @defmac SHIFT_COUNT_TRUNCATED
10310 A C expression that is nonzero if on this machine the number of bits
10311 actually used for the count of a shift operation is equal to the number
10312 of bits needed to represent the size of the object being shifted. When
10313 this macro is nonzero, the compiler will assume that it is safe to omit
10314 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10315 truncates the count of a shift operation. On machines that have
10316 instructions that act on bit-fields at variable positions, which may
10317 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10318 also enables deletion of truncations of the values that serve as
10319 arguments to bit-field instructions.
10321 If both types of instructions truncate the count (for shifts) and
10322 position (for bit-field operations), or if no variable-position bit-field
10323 instructions exist, you should define this macro.
10325 However, on some machines, such as the 80386 and the 680x0, truncation
10326 only applies to shift operations and not the (real or pretended)
10327 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10328 such machines. Instead, add patterns to the @file{md} file that include
10329 the implied truncation of the shift instructions.
10331 You need not define this macro if it would always have the value of zero.
10334 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10335 @hook TARGET_SHIFT_TRUNCATION_MASK
10336 This function describes how the standard shift patterns for @var{mode}
10337 deal with shifts by negative amounts or by more than the width of the mode.
10338 @xref{shift patterns}.
10340 On many machines, the shift patterns will apply a mask @var{m} to the
10341 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10342 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10343 this is true for mode @var{mode}, the function should return @var{m},
10344 otherwise it should return 0. A return value of 0 indicates that no
10345 particular behavior is guaranteed.
10347 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10348 @emph{not} apply to general shift rtxes; it applies only to instructions
10349 that are generated by the named shift patterns.
10351 The default implementation of this function returns
10352 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10353 and 0 otherwise. This definition is always safe, but if
10354 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10355 nevertheless truncate the shift count, you may get better code
10359 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10360 A C expression which is nonzero if on this machine it is safe to
10361 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10362 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10363 operating on it as if it had only @var{outprec} bits.
10365 On many machines, this expression can be 1.
10367 @c rearranged this, removed the phrase "it is reported that". this was
10368 @c to fix an overfull hbox. --mew 10feb93
10369 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10370 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10371 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10372 such cases may improve things.
10375 @hook TARGET_MODE_REP_EXTENDED
10376 The representation of an integral mode can be such that the values
10377 are always extended to a wider integral mode. Return
10378 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10379 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10380 otherwise. (Currently, none of the targets use zero-extended
10381 representation this way so unlike @code{LOAD_EXTEND_OP},
10382 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10383 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10384 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10385 widest integral mode and currently we take advantage of this fact.)
10387 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10388 value even if the extension is not performed on certain hard registers
10389 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10390 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10392 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10393 describe two related properties. If you define
10394 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10395 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10398 In order to enforce the representation of @code{mode},
10399 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10403 @defmac STORE_FLAG_VALUE
10404 A C expression describing the value returned by a comparison operator
10405 with an integral mode and stored by a store-flag instruction
10406 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10407 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10408 comparison operators whose results have a @code{MODE_INT} mode.
10410 A value of 1 or @minus{}1 means that the instruction implementing the
10411 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10412 and 0 when the comparison is false. Otherwise, the value indicates
10413 which bits of the result are guaranteed to be 1 when the comparison is
10414 true. This value is interpreted in the mode of the comparison
10415 operation, which is given by the mode of the first operand in the
10416 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10417 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10420 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10421 generate code that depends only on the specified bits. It can also
10422 replace comparison operators with equivalent operations if they cause
10423 the required bits to be set, even if the remaining bits are undefined.
10424 For example, on a machine whose comparison operators return an
10425 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10426 @samp{0x80000000}, saying that just the sign bit is relevant, the
10430 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10434 can be converted to
10437 (ashift:SI @var{x} (const_int @var{n}))
10441 where @var{n} is the appropriate shift count to move the bit being
10442 tested into the sign bit.
10444 There is no way to describe a machine that always sets the low-order bit
10445 for a true value, but does not guarantee the value of any other bits,
10446 but we do not know of any machine that has such an instruction. If you
10447 are trying to port GCC to such a machine, include an instruction to
10448 perform a logical-and of the result with 1 in the pattern for the
10449 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10451 Often, a machine will have multiple instructions that obtain a value
10452 from a comparison (or the condition codes). Here are rules to guide the
10453 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10458 Use the shortest sequence that yields a valid definition for
10459 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10460 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10461 comparison operators to do so because there may be opportunities to
10462 combine the normalization with other operations.
10465 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10466 slightly preferred on machines with expensive jumps and 1 preferred on
10470 As a second choice, choose a value of @samp{0x80000001} if instructions
10471 exist that set both the sign and low-order bits but do not define the
10475 Otherwise, use a value of @samp{0x80000000}.
10478 Many machines can produce both the value chosen for
10479 @code{STORE_FLAG_VALUE} and its negation in the same number of
10480 instructions. On those machines, you should also define a pattern for
10481 those cases, e.g., one matching
10484 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10487 Some machines can also perform @code{and} or @code{plus} operations on
10488 condition code values with less instructions than the corresponding
10489 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10490 machines, define the appropriate patterns. Use the names @code{incscc}
10491 and @code{decscc}, respectively, for the patterns which perform
10492 @code{plus} or @code{minus} operations on condition code values. See
10493 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10494 find such instruction sequences on other machines.
10496 If this macro is not defined, the default value, 1, is used. You need
10497 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10498 instructions, or if the value generated by these instructions is 1.
10501 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10502 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10503 returned when comparison operators with floating-point results are true.
10504 Define this macro on machines that have comparison operations that return
10505 floating-point values. If there are no such operations, do not define
10509 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10510 A C expression that gives a rtx representing the nonzero true element
10511 for vector comparisons. The returned rtx should be valid for the inner
10512 mode of @var{mode} which is guaranteed to be a vector mode. Define
10513 this macro on machines that have vector comparison operations that
10514 return a vector result. If there are no such operations, do not define
10515 this macro. Typically, this macro is defined as @code{const1_rtx} or
10516 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10517 the compiler optimizing such vector comparison operations for the
10521 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10522 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10523 A C expression that indicates whether the architecture defines a value
10524 for @code{clz} or @code{ctz} with a zero operand.
10525 A result of @code{0} indicates the value is undefined.
10526 If the value is defined for only the RTL expression, the macro should
10527 evaluate to @code{1}; if the value applies also to the corresponding optab
10528 entry (which is normally the case if it expands directly into
10529 the corresponding RTL), then the macro should evaluate to @code{2}.
10530 In the cases where the value is defined, @var{value} should be set to
10533 If this macro is not defined, the value of @code{clz} or
10534 @code{ctz} at zero is assumed to be undefined.
10536 This macro must be defined if the target's expansion for @code{ffs}
10537 relies on a particular value to get correct results. Otherwise it
10538 is not necessary, though it may be used to optimize some corner cases, and
10539 to provide a default expansion for the @code{ffs} optab.
10541 Note that regardless of this macro the ``definedness'' of @code{clz}
10542 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10543 visible to the user. Thus one may be free to adjust the value at will
10544 to match the target expansion of these operations without fear of
10549 An alias for the machine mode for pointers. On most machines, define
10550 this to be the integer mode corresponding to the width of a hardware
10551 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10552 On some machines you must define this to be one of the partial integer
10553 modes, such as @code{PSImode}.
10555 The width of @code{Pmode} must be at least as large as the value of
10556 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10557 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10561 @defmac FUNCTION_MODE
10562 An alias for the machine mode used for memory references to functions
10563 being called, in @code{call} RTL expressions. On most CISC machines,
10564 where an instruction can begin at any byte address, this should be
10565 @code{QImode}. On most RISC machines, where all instructions have fixed
10566 size and alignment, this should be a mode with the same size and alignment
10567 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10570 @defmac STDC_0_IN_SYSTEM_HEADERS
10571 In normal operation, the preprocessor expands @code{__STDC__} to the
10572 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10573 hosts, like Solaris, the system compiler uses a different convention,
10574 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10575 strict conformance to the C Standard.
10577 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10578 convention when processing system header files, but when processing user
10579 files @code{__STDC__} will always expand to 1.
10582 @defmac NO_IMPLICIT_EXTERN_C
10583 Define this macro if the system header files support C++ as well as C@.
10584 This macro inhibits the usual method of using system header files in
10585 C++, which is to pretend that the file's contents are enclosed in
10586 @samp{extern "C" @{@dots{}@}}.
10591 @defmac REGISTER_TARGET_PRAGMAS ()
10592 Define this macro if you want to implement any target-specific pragmas.
10593 If defined, it is a C expression which makes a series of calls to
10594 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10595 for each pragma. The macro may also do any
10596 setup required for the pragmas.
10598 The primary reason to define this macro is to provide compatibility with
10599 other compilers for the same target. In general, we discourage
10600 definition of target-specific pragmas for GCC@.
10602 If the pragma can be implemented by attributes then you should consider
10603 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10605 Preprocessor macros that appear on pragma lines are not expanded. All
10606 @samp{#pragma} directives that do not match any registered pragma are
10607 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10610 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10611 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10613 Each call to @code{c_register_pragma} or
10614 @code{c_register_pragma_with_expansion} establishes one pragma. The
10615 @var{callback} routine will be called when the preprocessor encounters a
10619 #pragma [@var{space}] @var{name} @dots{}
10622 @var{space} is the case-sensitive namespace of the pragma, or
10623 @code{NULL} to put the pragma in the global namespace. The callback
10624 routine receives @var{pfile} as its first argument, which can be passed
10625 on to cpplib's functions if necessary. You can lex tokens after the
10626 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10627 callback will be silently ignored. The end of the line is indicated by
10628 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10629 arguments of pragmas registered with
10630 @code{c_register_pragma_with_expansion} but not on the arguments of
10631 pragmas registered with @code{c_register_pragma}.
10633 Note that the use of @code{pragma_lex} is specific to the C and C++
10634 compilers. It will not work in the Java or Fortran compilers, or any
10635 other language compilers for that matter. Thus if @code{pragma_lex} is going
10636 to be called from target-specific code, it must only be done so when
10637 building the C and C++ compilers. This can be done by defining the
10638 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10639 target entry in the @file{config.gcc} file. These variables should name
10640 the target-specific, language-specific object file which contains the
10641 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10642 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10643 how to build this object file.
10646 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10647 Define this macro if macros should be expanded in the
10648 arguments of @samp{#pragma pack}.
10651 @defmac TARGET_DEFAULT_PACK_STRUCT
10652 If your target requires a structure packing default other than 0 (meaning
10653 the machine default), define this macro to the necessary value (in bytes).
10654 This must be a value that would also be valid to use with
10655 @samp{#pragma pack()} (that is, a small power of two).
10658 @defmac DOLLARS_IN_IDENTIFIERS
10659 Define this macro to control use of the character @samp{$} in
10660 identifier names for the C family of languages. 0 means @samp{$} is
10661 not allowed by default; 1 means it is allowed. 1 is the default;
10662 there is no need to define this macro in that case.
10665 @defmac NO_DOLLAR_IN_LABEL
10666 Define this macro if the assembler does not accept the character
10667 @samp{$} in label names. By default constructors and destructors in
10668 G++ have @samp{$} in the identifiers. If this macro is defined,
10669 @samp{.} is used instead.
10672 @defmac NO_DOT_IN_LABEL
10673 Define this macro if the assembler does not accept the character
10674 @samp{.} in label names. By default constructors and destructors in G++
10675 have names that use @samp{.}. If this macro is defined, these names
10676 are rewritten to avoid @samp{.}.
10679 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10680 Define this macro as a C expression that is nonzero if it is safe for the
10681 delay slot scheduler to place instructions in the delay slot of @var{insn},
10682 even if they appear to use a resource set or clobbered in @var{insn}.
10683 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10684 every @code{call_insn} has this behavior. On machines where some @code{insn}
10685 or @code{jump_insn} is really a function call and hence has this behavior,
10686 you should define this macro.
10688 You need not define this macro if it would always return zero.
10691 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10692 Define this macro as a C expression that is nonzero if it is safe for the
10693 delay slot scheduler to place instructions in the delay slot of @var{insn},
10694 even if they appear to set or clobber a resource referenced in @var{insn}.
10695 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10696 some @code{insn} or @code{jump_insn} is really a function call and its operands
10697 are registers whose use is actually in the subroutine it calls, you should
10698 define this macro. Doing so allows the delay slot scheduler to move
10699 instructions which copy arguments into the argument registers into the delay
10700 slot of @var{insn}.
10702 You need not define this macro if it would always return zero.
10705 @defmac MULTIPLE_SYMBOL_SPACES
10706 Define this macro as a C expression that is nonzero if, in some cases,
10707 global symbols from one translation unit may not be bound to undefined
10708 symbols in another translation unit without user intervention. For
10709 instance, under Microsoft Windows symbols must be explicitly imported
10710 from shared libraries (DLLs).
10712 You need not define this macro if it would always evaluate to zero.
10715 @hook TARGET_MD_ASM_CLOBBERS
10716 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10717 any hard regs the port wishes to automatically clobber for an asm.
10718 It should return the result of the last @code{tree_cons} used to add a
10719 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10720 corresponding parameters to the asm and may be inspected to avoid
10721 clobbering a register that is an input or output of the asm. You can use
10722 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10723 for overlap with regards to asm-declared registers.
10726 @defmac MATH_LIBRARY
10727 Define this macro as a C string constant for the linker argument to link
10728 in the system math library, minus the initial @samp{"-l"}, or
10729 @samp{""} if the target does not have a
10730 separate math library.
10732 You need only define this macro if the default of @samp{"m"} is wrong.
10735 @defmac LIBRARY_PATH_ENV
10736 Define this macro as a C string constant for the environment variable that
10737 specifies where the linker should look for libraries.
10739 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10743 @defmac TARGET_POSIX_IO
10744 Define this macro if the target supports the following POSIX@ file
10745 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10746 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10747 to use file locking when exiting a program, which avoids race conditions
10748 if the program has forked. It will also create directories at run-time
10749 for cross-profiling.
10752 @defmac MAX_CONDITIONAL_EXECUTE
10754 A C expression for the maximum number of instructions to execute via
10755 conditional execution instructions instead of a branch. A value of
10756 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10757 1 if it does use cc0.
10760 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10761 Used if the target needs to perform machine-dependent modifications on the
10762 conditionals used for turning basic blocks into conditionally executed code.
10763 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10764 contains information about the currently processed blocks. @var{true_expr}
10765 and @var{false_expr} are the tests that are used for converting the
10766 then-block and the else-block, respectively. Set either @var{true_expr} or
10767 @var{false_expr} to a null pointer if the tests cannot be converted.
10770 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10771 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10772 if-statements into conditions combined by @code{and} and @code{or} operations.
10773 @var{bb} contains the basic block that contains the test that is currently
10774 being processed and about to be turned into a condition.
10777 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10778 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10779 be converted to conditional execution format. @var{ce_info} points to
10780 a data structure, @code{struct ce_if_block}, which contains information
10781 about the currently processed blocks.
10784 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10785 A C expression to perform any final machine dependent modifications in
10786 converting code to conditional execution. The involved basic blocks
10787 can be found in the @code{struct ce_if_block} structure that is pointed
10788 to by @var{ce_info}.
10791 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10792 A C expression to cancel any machine dependent modifications in
10793 converting code to conditional execution. The involved basic blocks
10794 can be found in the @code{struct ce_if_block} structure that is pointed
10795 to by @var{ce_info}.
10798 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10799 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10800 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10803 @defmac IFCVT_EXTRA_FIELDS
10804 If defined, it should expand to a set of field declarations that will be
10805 added to the @code{struct ce_if_block} structure. These should be initialized
10806 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10809 @hook TARGET_MACHINE_DEPENDENT_REORG
10810 If non-null, this hook performs a target-specific pass over the
10811 instruction stream. The compiler will run it at all optimization levels,
10812 just before the point at which it normally does delayed-branch scheduling.
10814 The exact purpose of the hook varies from target to target. Some use
10815 it to do transformations that are necessary for correctness, such as
10816 laying out in-function constant pools or avoiding hardware hazards.
10817 Others use it as an opportunity to do some machine-dependent optimizations.
10819 You need not implement the hook if it has nothing to do. The default
10820 definition is null.
10823 @hook TARGET_INIT_BUILTINS
10824 Define this hook if you have any machine-specific built-in functions
10825 that need to be defined. It should be a function that performs the
10828 Machine specific built-in functions can be useful to expand special machine
10829 instructions that would otherwise not normally be generated because
10830 they have no equivalent in the source language (for example, SIMD vector
10831 instructions or prefetch instructions).
10833 To create a built-in function, call the function
10834 @code{lang_hooks.builtin_function}
10835 which is defined by the language front end. You can use any type nodes set
10836 up by @code{build_common_tree_nodes};
10837 only language front ends that use those two functions will call
10838 @samp{TARGET_INIT_BUILTINS}.
10841 @hook TARGET_BUILTIN_DECL
10842 Define this hook if you have any machine-specific built-in functions
10843 that need to be defined. It should be a function that returns the
10844 builtin function declaration for the builtin function code @var{code}.
10845 If there is no such builtin and it cannot be initialized at this time
10846 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10847 If @var{code} is out of range the function should return
10848 @code{error_mark_node}.
10851 @hook TARGET_EXPAND_BUILTIN
10853 Expand a call to a machine specific built-in function that was set up by
10854 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10855 function call; the result should go to @var{target} if that is
10856 convenient, and have mode @var{mode} if that is convenient.
10857 @var{subtarget} may be used as the target for computing one of
10858 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10859 ignored. This function should return the result of the call to the
10863 @hook TARGET_RESOLVE_OVERLOADED_BUILTIN
10864 Select a replacement for a machine specific built-in function that
10865 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10866 @emph{before} regular type checking, and so allows the target to
10867 implement a crude form of function overloading. @var{fndecl} is the
10868 declaration of the built-in function. @var{arglist} is the list of
10869 arguments passed to the built-in function. The result is a
10870 complete expression that implements the operation, usually
10871 another @code{CALL_EXPR}.
10872 @var{arglist} really has type @samp{VEC(tree,gc)*}
10875 @hook TARGET_FOLD_BUILTIN
10876 Fold a call to a machine specific built-in function that was set up by
10877 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10878 built-in function. @var{n_args} is the number of arguments passed to
10879 the function; the arguments themselves are pointed to by @var{argp}.
10880 The result is another tree containing a simplified expression for the
10881 call's result. If @var{ignore} is true the value will be ignored.
10884 @hook TARGET_INVALID_WITHIN_DOLOOP
10886 Take an instruction in @var{insn} and return NULL if it is valid within a
10887 low-overhead loop, otherwise return a string explaining why doloop
10888 could not be applied.
10890 Many targets use special registers for low-overhead looping. For any
10891 instruction that clobbers these this function should return a string indicating
10892 the reason why the doloop could not be applied.
10893 By default, the RTL loop optimizer does not use a present doloop pattern for
10894 loops containing function calls or branch on table instructions.
10897 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10899 Take a branch insn in @var{branch1} and another in @var{branch2}.
10900 Return true if redirecting @var{branch1} to the destination of
10901 @var{branch2} is possible.
10903 On some targets, branches may have a limited range. Optimizing the
10904 filling of delay slots can result in branches being redirected, and this
10905 may in turn cause a branch offset to overflow.
10908 @hook TARGET_COMMUTATIVE_P
10909 This target hook returns @code{true} if @var{x} is considered to be commutative.
10910 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10911 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10912 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10915 @hook TARGET_ALLOCATE_INITIAL_VALUE
10917 When the initial value of a hard register has been copied in a pseudo
10918 register, it is often not necessary to actually allocate another register
10919 to this pseudo register, because the original hard register or a stack slot
10920 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10921 is called at the start of register allocation once for each hard register
10922 that had its initial value copied by using
10923 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10924 Possible values are @code{NULL_RTX}, if you don't want
10925 to do any special allocation, a @code{REG} rtx---that would typically be
10926 the hard register itself, if it is known not to be clobbered---or a
10928 If you are returning a @code{MEM}, this is only a hint for the allocator;
10929 it might decide to use another register anyways.
10930 You may use @code{current_function_leaf_function} in the hook, functions
10931 that use @code{REG_N_SETS}, to determine if the hard
10932 register in question will not be clobbered.
10933 The default value of this hook is @code{NULL}, which disables any special
10937 @hook TARGET_UNSPEC_MAY_TRAP_P
10938 This target hook returns nonzero if @var{x}, an @code{unspec} or
10939 @code{unspec_volatile} operation, might cause a trap. Targets can use
10940 this hook to enhance precision of analysis for @code{unspec} and
10941 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10942 to analyze inner elements of @var{x} in which case @var{flags} should be
10946 @hook TARGET_SET_CURRENT_FUNCTION
10947 The compiler invokes this hook whenever it changes its current function
10948 context (@code{cfun}). You can define this function if
10949 the back end needs to perform any initialization or reset actions on a
10950 per-function basis. For example, it may be used to implement function
10951 attributes that affect register usage or code generation patterns.
10952 The argument @var{decl} is the declaration for the new function context,
10953 and may be null to indicate that the compiler has left a function context
10954 and is returning to processing at the top level.
10955 The default hook function does nothing.
10957 GCC sets @code{cfun} to a dummy function context during initialization of
10958 some parts of the back end. The hook function is not invoked in this
10959 situation; you need not worry about the hook being invoked recursively,
10960 or when the back end is in a partially-initialized state.
10961 @code{cfun} might be @code{NULL} to indicate processing at top level,
10962 outside of any function scope.
10965 @defmac TARGET_OBJECT_SUFFIX
10966 Define this macro to be a C string representing the suffix for object
10967 files on your target machine. If you do not define this macro, GCC will
10968 use @samp{.o} as the suffix for object files.
10971 @defmac TARGET_EXECUTABLE_SUFFIX
10972 Define this macro to be a C string representing the suffix to be
10973 automatically added to executable files on your target machine. If you
10974 do not define this macro, GCC will use the null string as the suffix for
10978 @defmac COLLECT_EXPORT_LIST
10979 If defined, @code{collect2} will scan the individual object files
10980 specified on its command line and create an export list for the linker.
10981 Define this macro for systems like AIX, where the linker discards
10982 object files that are not referenced from @code{main} and uses export
10986 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
10987 Define this macro to a C expression representing a variant of the
10988 method call @var{mdecl}, if Java Native Interface (JNI) methods
10989 must be invoked differently from other methods on your target.
10990 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
10991 the @code{stdcall} calling convention and this macro is then
10992 defined as this expression:
10995 build_type_attribute_variant (@var{mdecl},
10997 (get_identifier ("stdcall"),
11002 @hook TARGET_CANNOT_MODIFY_JUMPS_P
11003 This target hook returns @code{true} past the point in which new jump
11004 instructions could be created. On machines that require a register for
11005 every jump such as the SHmedia ISA of SH5, this point would typically be
11006 reload, so this target hook should be defined to a function such as:
11010 cannot_modify_jumps_past_reload_p ()
11012 return (reload_completed || reload_in_progress);
11017 @hook TARGET_BRANCH_TARGET_REGISTER_CLASS
11018 This target hook returns a register class for which branch target register
11019 optimizations should be applied. All registers in this class should be
11020 usable interchangeably. After reload, registers in this class will be
11021 re-allocated and loads will be hoisted out of loops and be subjected
11022 to inter-block scheduling.
11025 @hook TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED
11026 Branch target register optimization will by default exclude callee-saved
11028 that are not already live during the current function; if this target hook
11029 returns true, they will be included. The target code must than make sure
11030 that all target registers in the class returned by
11031 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11032 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11033 epilogues have already been generated. Note, even if you only return
11034 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11035 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11036 to reserve space for caller-saved target registers.
11039 @hook TARGET_HAVE_CONDITIONAL_EXECUTION
11040 This target hook returns true if the target supports conditional execution.
11041 This target hook is required only when the target has several different
11042 modes and they have different conditional execution capability, such as ARM.
11045 @hook TARGET_LOOP_UNROLL_ADJUST
11046 This target hook returns a new value for the number of times @var{loop}
11047 should be unrolled. The parameter @var{nunroll} is the number of times
11048 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11049 the loop, which is going to be checked for unrolling. This target hook
11050 is required only when the target has special constraints like maximum
11051 number of memory accesses.
11054 @defmac POWI_MAX_MULTS
11055 If defined, this macro is interpreted as a signed integer C expression
11056 that specifies the maximum number of floating point multiplications
11057 that should be emitted when expanding exponentiation by an integer
11058 constant inline. When this value is defined, exponentiation requiring
11059 more than this number of multiplications is implemented by calling the
11060 system library's @code{pow}, @code{powf} or @code{powl} routines.
11061 The default value places no upper bound on the multiplication count.
11064 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11065 This target hook should register any extra include files for the
11066 target. The parameter @var{stdinc} indicates if normal include files
11067 are present. The parameter @var{sysroot} is the system root directory.
11068 The parameter @var{iprefix} is the prefix for the gcc directory.
11071 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11072 This target hook should register any extra include files for the
11073 target before any standard headers. The parameter @var{stdinc}
11074 indicates if normal include files are present. The parameter
11075 @var{sysroot} is the system root directory. The parameter
11076 @var{iprefix} is the prefix for the gcc directory.
11079 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11080 This target hook should register special include paths for the target.
11081 The parameter @var{path} is the include to register. On Darwin
11082 systems, this is used for Framework includes, which have semantics
11083 that are different from @option{-I}.
11086 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11087 This target macro returns @code{true} if it is safe to use a local alias
11088 for a virtual function @var{fndecl} when constructing thunks,
11089 @code{false} otherwise. By default, the macro returns @code{true} for all
11090 functions, if a target supports aliases (i.e.@: defines
11091 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11094 @defmac TARGET_FORMAT_TYPES
11095 If defined, this macro is the name of a global variable containing
11096 target-specific format checking information for the @option{-Wformat}
11097 option. The default is to have no target-specific format checks.
11100 @defmac TARGET_N_FORMAT_TYPES
11101 If defined, this macro is the number of entries in
11102 @code{TARGET_FORMAT_TYPES}.
11105 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11106 If defined, this macro is the name of a global variable containing
11107 target-specific format overrides for the @option{-Wformat} option. The
11108 default is to have no target-specific format overrides. If defined,
11109 @code{TARGET_FORMAT_TYPES} must be defined, too.
11112 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11113 If defined, this macro specifies the number of entries in
11114 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11117 @defmac TARGET_OVERRIDES_FORMAT_INIT
11118 If defined, this macro specifies the optional initialization
11119 routine for target specific customizations of the system printf
11120 and scanf formatter settings.
11123 @hook TARGET_RELAXED_ORDERING
11124 If set to @code{true}, means that the target's memory model does not
11125 guarantee that loads which do not depend on one another will access
11126 main memory in the order of the instruction stream; if ordering is
11127 important, an explicit memory barrier must be used. This is true of
11128 many recent processors which implement a policy of ``relaxed,''
11129 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11130 and ia64. The default is @code{false}.
11133 @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
11134 If defined, this macro returns the diagnostic message when it is
11135 illegal to pass argument @var{val} to function @var{funcdecl}
11136 with prototype @var{typelist}.
11139 @hook TARGET_INVALID_CONVERSION
11140 If defined, this macro returns the diagnostic message when it is
11141 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11142 if validity should be determined by the front end.
11145 @hook TARGET_INVALID_UNARY_OP
11146 If defined, this macro returns the diagnostic message when it is
11147 invalid to apply operation @var{op} (where unary plus is denoted by
11148 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11149 if validity should be determined by the front end.
11152 @hook TARGET_INVALID_BINARY_OP
11153 If defined, this macro returns the diagnostic message when it is
11154 invalid to apply operation @var{op} to operands of types @var{type1}
11155 and @var{type2}, or @code{NULL} if validity should be determined by
11159 @hook TARGET_INVALID_PARAMETER_TYPE
11160 If defined, this macro returns the diagnostic message when it is
11161 invalid for functions to include parameters of type @var{type},
11162 or @code{NULL} if validity should be determined by
11163 the front end. This is currently used only by the C and C++ front ends.
11166 @hook TARGET_INVALID_RETURN_TYPE
11167 If defined, this macro returns the diagnostic message when it is
11168 invalid for functions to have return type @var{type},
11169 or @code{NULL} if validity should be determined by
11170 the front end. This is currently used only by the C and C++ front ends.
11173 @hook TARGET_PROMOTED_TYPE
11174 If defined, this target hook returns the type to which values of
11175 @var{type} should be promoted when they appear in expressions,
11176 analogous to the integer promotions, or @code{NULL_TREE} to use the
11177 front end's normal promotion rules. This hook is useful when there are
11178 target-specific types with special promotion rules.
11179 This is currently used only by the C and C++ front ends.
11182 @hook TARGET_CONVERT_TO_TYPE
11183 If defined, this hook returns the result of converting @var{expr} to
11184 @var{type}. It should return the converted expression,
11185 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11186 This hook is useful when there are target-specific types with special
11188 This is currently used only by the C and C++ front ends.
11191 @defmac TARGET_USE_JCR_SECTION
11192 This macro determines whether to use the JCR section to register Java
11193 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11194 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11198 This macro determines the size of the objective C jump buffer for the
11199 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11202 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11203 Define this macro if any target-specific attributes need to be attached
11204 to the functions in @file{libgcc} that provide low-level support for
11205 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11206 and the associated definitions of those functions.
11209 @hook TARGET_UPDATE_STACK_BOUNDARY
11210 Define this macro to update the current function stack boundary if
11214 @hook TARGET_GET_DRAP_RTX
11215 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11216 different argument pointer register is needed to access the function's
11217 argument list due to stack realignment. Return @code{NULL} if no DRAP
11221 @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
11222 When optimization is disabled, this hook indicates whether or not
11223 arguments should be allocated to stack slots. Normally, GCC allocates
11224 stacks slots for arguments when not optimizing in order to make
11225 debugging easier. However, when a function is declared with
11226 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11227 cannot safely move arguments from the registers in which they are passed
11228 to the stack. Therefore, this hook should return true in general, but
11229 false for naked functions. The default implementation always returns true.
11232 @hook TARGET_CONST_ANCHOR
11233 On some architectures it can take multiple instructions to synthesize
11234 a constant. If there is another constant already in a register that
11235 is close enough in value then it is preferable that the new constant
11236 is computed from this register using immediate addition or
11237 subtraction. We accomplish this through CSE. Besides the value of
11238 the constant we also add a lower and an upper constant anchor to the
11239 available expressions. These are then queried when encountering new
11240 constants. The anchors are computed by rounding the constant up and
11241 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11242 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11243 accepted by immediate-add plus one. We currently assume that the
11244 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11245 MIPS, where add-immediate takes a 16-bit signed value,
11246 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11247 is zero, which disables this optimization. @end deftypevr
11249 @hook TARGET_MEMMODEL_CHECK
11250 Validate target specific memory model mask bits. When NULL no target specific
11251 memory model bits are allowed.
11254 @hook TARGET_ATOMIC_TEST_AND_SET_TRUEVAL