1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Target Description Macros and Functions
9 @cindex machine description macros
10 @cindex target description macros
11 @cindex macros, target description
12 @cindex @file{tm.h} macros
14 In addition to the file @file{@var{machine}.md}, a machine description
15 includes a C header file conventionally given the name
16 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
17 The header file defines numerous macros that convey the information
18 about the target machine that does not fit into the scheme of the
19 @file{.md} file. The file @file{tm.h} should be a link to
20 @file{@var{machine}.h}. The header file @file{config.h} includes
21 @file{tm.h} and most compiler source files include @file{config.h}. The
22 source file defines a variable @code{targetm}, which is a structure
23 containing pointers to functions and data relating to the target
24 machine. @file{@var{machine}.c} should also contain their definitions,
25 if they are not defined elsewhere in GCC, and other functions called
26 through the macros defined in the @file{.h} file.
29 * Target Structure:: The @code{targetm} variable.
30 * Driver:: Controlling how the driver runs the compilation passes.
31 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
32 * Per-Function Data:: Defining data structures for per-function information.
33 * Storage Layout:: Defining sizes and alignments of data.
34 * Type Layout:: Defining sizes and properties of basic user data types.
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
37 * Old Constraints:: The old way to define machine-specific constraints.
38 * Stack and Calling:: Defining which way the stack grows and by how much.
39 * Varargs:: Defining the varargs macros.
40 * Trampolines:: Code set up at run time to enter a nested function.
41 * Library Calls:: Controlling how library routines are implicitly called.
42 * Addressing Modes:: Defining addressing modes valid for memory operands.
43 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
44 * Condition Code:: Defining how insns update the condition code.
45 * Costs:: Defining relative costs of different operations.
46 * Scheduling:: Adjusting the behavior of the instruction scheduler.
47 * Sections:: Dividing storage into text, data, and other sections.
48 * PIC:: Macros for position independent code.
49 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
50 * Debugging Info:: Defining the format of debugging output.
51 * Floating Point:: Handling floating point for cross-compilers.
52 * Mode Switching:: Insertion of mode-switching instructions.
53 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
54 * Emulated TLS:: Emulated TLS support.
55 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
56 * PCH Target:: Validity checking for precompiled headers.
57 * C++ ABI:: Controlling C++ ABI changes.
58 * Named Address Spaces:: Adding support for named address spaces
59 * Misc:: Everything else.
62 @node Target Structure
63 @section The Global @code{targetm} Variable
65 @cindex target functions
67 @deftypevar {struct gcc_target} targetm
68 The target @file{.c} file must define the global @code{targetm} variable
69 which contains pointers to functions and data relating to the target
70 machine. The variable is declared in @file{target.h};
71 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
72 used to initialize the variable, and macros for the default initializers
73 for elements of the structure. The @file{.c} file should override those
74 macros for which the default definition is inappropriate. For example:
77 #include "target-def.h"
79 /* @r{Initialize the GCC target structure.} */
81 #undef TARGET_COMP_TYPE_ATTRIBUTES
82 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
84 struct gcc_target targetm = TARGET_INITIALIZER;
88 Where a macro should be defined in the @file{.c} file in this manner to
89 form part of the @code{targetm} structure, it is documented below as a
90 ``Target Hook'' with a prototype. Many macros will change in future
91 from being defined in the @file{.h} file to being part of the
92 @code{targetm} structure.
94 Similarly, there is a @code{targetcm} variable for hooks that are
95 specific to front ends for C-family languages, documented as ``C
96 Target Hook''. This is declared in @file{c-family/c-target.h}, the
97 initializer @code{TARGETCM_INITIALIZER} in
98 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
99 themselves, they should set @code{target_has_targetcm=yes} in
100 @file{config.gcc}; otherwise a default definition is used.
102 Similarly, there is a @code{targetm_common} variable for hooks that
103 are shared between the compiler driver and the compilers proper,
104 documented as ``Common Target Hook''. This is declared in
105 @file{common/common-target.h}, the initializer
106 @code{TARGETM_COMMON_INITIALIZER} in
107 @file{common/common-target-def.h}. If targets initialize
108 @code{targetm_common} themselves, they should set
109 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
110 default definition is used.
113 @section Controlling the Compilation Driver, @file{gcc}
115 @cindex controlling the compilation driver
117 @c prevent bad page break with this line
118 You can control the compilation driver.
120 @defmac DRIVER_SELF_SPECS
121 A list of specs for the driver itself. It should be a suitable
122 initializer for an array of strings, with no surrounding braces.
124 The driver applies these specs to its own command line between loading
125 default @file{specs} files (but not command-line specified ones) and
126 choosing the multilib directory or running any subcommands. It
127 applies them in the order given, so each spec can depend on the
128 options added by earlier ones. It is also possible to remove options
129 using @samp{%<@var{option}} in the usual way.
131 This macro can be useful when a port has several interdependent target
132 options. It provides a way of standardizing the command line so
133 that the other specs are easier to write.
135 Do not define this macro if it does not need to do anything.
138 @defmac OPTION_DEFAULT_SPECS
139 A list of specs used to support configure-time default options (i.e.@:
140 @option{--with} options) in the driver. It should be a suitable initializer
141 for an array of structures, each containing two strings, without the
142 outermost pair of surrounding braces.
144 The first item in the pair is the name of the default. This must match
145 the code in @file{config.gcc} for the target. The second item is a spec
146 to apply if a default with this name was specified. The string
147 @samp{%(VALUE)} in the spec will be replaced by the value of the default
148 everywhere it occurs.
150 The driver will apply these specs to its own command line between loading
151 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
152 the same mechanism as @code{DRIVER_SELF_SPECS}.
154 Do not define this macro if it does not need to do anything.
158 A C string constant that tells the GCC driver program options to
159 pass to CPP@. It can also specify how to translate options you
160 give to GCC into options for GCC to pass to the CPP@.
162 Do not define this macro if it does not need to do anything.
165 @defmac CPLUSPLUS_CPP_SPEC
166 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
167 than C@. If you do not define this macro, then the value of
168 @code{CPP_SPEC} (if any) will be used instead.
172 A C string constant that tells the GCC driver program options to
173 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
175 It can also specify how to translate options you give to GCC into options
176 for GCC to pass to front ends.
178 Do not define this macro if it does not need to do anything.
182 A C string constant that tells the GCC driver program options to
183 pass to @code{cc1plus}. It can also specify how to translate options you
184 give to GCC into options for GCC to pass to the @code{cc1plus}.
186 Do not define this macro if it does not need to do anything.
187 Note that everything defined in CC1_SPEC is already passed to
188 @code{cc1plus} so there is no need to duplicate the contents of
189 CC1_SPEC in CC1PLUS_SPEC@.
193 A C string constant that tells the GCC driver program options to
194 pass to the assembler. It can also specify how to translate options
195 you give to GCC into options for GCC to pass to the assembler.
196 See the file @file{sun3.h} for an example of this.
198 Do not define this macro if it does not need to do anything.
201 @defmac ASM_FINAL_SPEC
202 A C string constant that tells the GCC driver program how to
203 run any programs which cleanup after the normal assembler.
204 Normally, this is not needed. See the file @file{mips.h} for
207 Do not define this macro if it does not need to do anything.
210 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
211 Define this macro, with no value, if the driver should give the assembler
212 an argument consisting of a single dash, @option{-}, to instruct it to
213 read from its standard input (which will be a pipe connected to the
214 output of the compiler proper). This argument is given after any
215 @option{-o} option specifying the name of the output file.
217 If you do not define this macro, the assembler is assumed to read its
218 standard input if given no non-option arguments. If your assembler
219 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
220 see @file{mips.h} for instance.
224 A C string constant that tells the GCC driver program options to
225 pass to the linker. It can also specify how to translate options you
226 give to GCC into options for GCC to pass to the linker.
228 Do not define this macro if it does not need to do anything.
232 Another C string constant used much like @code{LINK_SPEC}. The difference
233 between the two is that @code{LIB_SPEC} is used at the end of the
234 command given to the linker.
236 If this macro is not defined, a default is provided that
237 loads the standard C library from the usual place. See @file{gcc.c}.
241 Another C string constant that tells the GCC driver program
242 how and when to place a reference to @file{libgcc.a} into the
243 linker command line. This constant is placed both before and after
244 the value of @code{LIB_SPEC}.
246 If this macro is not defined, the GCC driver provides a default that
247 passes the string @option{-lgcc} to the linker.
250 @defmac REAL_LIBGCC_SPEC
251 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
252 @code{LIBGCC_SPEC} is not directly used by the driver program but is
253 instead modified to refer to different versions of @file{libgcc.a}
254 depending on the values of the command line flags @option{-static},
255 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
256 targets where these modifications are inappropriate, define
257 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
258 driver how to place a reference to @file{libgcc} on the link command
259 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
262 @defmac USE_LD_AS_NEEDED
263 A macro that controls the modifications to @code{LIBGCC_SPEC}
264 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
265 generated that uses --as-needed and the shared libgcc in place of the
266 static exception handler library, when linking without any of
267 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
271 If defined, this C string constant is added to @code{LINK_SPEC}.
272 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
273 the modifications to @code{LIBGCC_SPEC} mentioned in
274 @code{REAL_LIBGCC_SPEC}.
277 @defmac STARTFILE_SPEC
278 Another C string constant used much like @code{LINK_SPEC}. The
279 difference between the two is that @code{STARTFILE_SPEC} is used at
280 the very beginning of the command given to the linker.
282 If this macro is not defined, a default is provided that loads the
283 standard C startup file from the usual place. See @file{gcc.c}.
287 Another C string constant used much like @code{LINK_SPEC}. The
288 difference between the two is that @code{ENDFILE_SPEC} is used at
289 the very end of the command given to the linker.
291 Do not define this macro if it does not need to do anything.
294 @defmac THREAD_MODEL_SPEC
295 GCC @code{-v} will print the thread model GCC was configured to use.
296 However, this doesn't work on platforms that are multilibbed on thread
297 models, such as AIX 4.3. On such platforms, define
298 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
299 blanks that names one of the recognized thread models. @code{%*}, the
300 default value of this macro, will expand to the value of
301 @code{thread_file} set in @file{config.gcc}.
304 @defmac SYSROOT_SUFFIX_SPEC
305 Define this macro to add a suffix to the target sysroot when GCC is
306 configured with a sysroot. This will cause GCC to search for usr/lib,
307 et al, within sysroot+suffix.
310 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
311 Define this macro to add a headers_suffix to the target sysroot when
312 GCC is configured with a sysroot. This will cause GCC to pass the
313 updated sysroot+headers_suffix to CPP, causing it to search for
314 usr/include, et al, within sysroot+headers_suffix.
318 Define this macro to provide additional specifications to put in the
319 @file{specs} file that can be used in various specifications like
322 The definition should be an initializer for an array of structures,
323 containing a string constant, that defines the specification name, and a
324 string constant that provides the specification.
326 Do not define this macro if it does not need to do anything.
328 @code{EXTRA_SPECS} is useful when an architecture contains several
329 related targets, which have various @code{@dots{}_SPECS} which are similar
330 to each other, and the maintainer would like one central place to keep
333 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
334 define either @code{_CALL_SYSV} when the System V calling sequence is
335 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
338 The @file{config/rs6000/rs6000.h} target file defines:
341 #define EXTRA_SPECS \
342 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
344 #define CPP_SYS_DEFAULT ""
347 The @file{config/rs6000/sysv.h} target file defines:
351 "%@{posix: -D_POSIX_SOURCE @} \
352 %@{mcall-sysv: -D_CALL_SYSV @} \
353 %@{!mcall-sysv: %(cpp_sysv_default) @} \
354 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
356 #undef CPP_SYSV_DEFAULT
357 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
360 while the @file{config/rs6000/eabiaix.h} target file defines
361 @code{CPP_SYSV_DEFAULT} as:
364 #undef CPP_SYSV_DEFAULT
365 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
369 @defmac LINK_LIBGCC_SPECIAL_1
370 Define this macro if the driver program should find the library
371 @file{libgcc.a}. If you do not define this macro, the driver program will pass
372 the argument @option{-lgcc} to tell the linker to do the search.
375 @defmac LINK_GCC_C_SEQUENCE_SPEC
376 The sequence in which libgcc and libc are specified to the linker.
377 By default this is @code{%G %L %G}.
380 @defmac LINK_COMMAND_SPEC
381 A C string constant giving the complete command line need to execute the
382 linker. When you do this, you will need to update your port each time a
383 change is made to the link command line within @file{gcc.c}. Therefore,
384 define this macro only if you need to completely redefine the command
385 line for invoking the linker and there is no other way to accomplish
386 the effect you need. Overriding this macro may be avoidable by overriding
387 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
390 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
391 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
392 directories from linking commands. Do not give it a nonzero value if
393 removing duplicate search directories changes the linker's semantics.
396 @hook TARGET_ALWAYS_STRIP_DOTDOT
398 @defmac MULTILIB_DEFAULTS
399 Define this macro as a C expression for the initializer of an array of
400 string to tell the driver program which options are defaults for this
401 target and thus do not need to be handled specially when using
402 @code{MULTILIB_OPTIONS}.
404 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
405 the target makefile fragment or if none of the options listed in
406 @code{MULTILIB_OPTIONS} are set by default.
407 @xref{Target Fragment}.
410 @defmac RELATIVE_PREFIX_NOT_LINKDIR
411 Define this macro to tell @command{gcc} that it should only translate
412 a @option{-B} prefix into a @option{-L} linker option if the prefix
413 indicates an absolute file name.
416 @defmac MD_EXEC_PREFIX
417 If defined, this macro is an additional prefix to try after
418 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
419 when the compiler is built as a cross
420 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
421 to the list of directories used to find the assembler in @file{configure.in}.
424 @defmac STANDARD_STARTFILE_PREFIX
425 Define this macro as a C string constant if you wish to override the
426 standard choice of @code{libdir} as the default prefix to
427 try when searching for startup files such as @file{crt0.o}.
428 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
429 is built as a cross compiler.
432 @defmac STANDARD_STARTFILE_PREFIX_1
433 Define this macro as a C string constant if you wish to override the
434 standard choice of @code{/lib} as a prefix to try after the default prefix
435 when searching for startup files such as @file{crt0.o}.
436 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
437 is built as a cross compiler.
440 @defmac STANDARD_STARTFILE_PREFIX_2
441 Define this macro as a C string constant if you wish to override the
442 standard choice of @code{/lib} as yet another prefix to try after the
443 default prefix when searching for startup files such as @file{crt0.o}.
444 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
445 is built as a cross compiler.
448 @defmac MD_STARTFILE_PREFIX
449 If defined, this macro supplies an additional prefix to try after the
450 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
451 compiler is built as a cross compiler.
454 @defmac MD_STARTFILE_PREFIX_1
455 If defined, this macro supplies yet another prefix to try after the
456 standard prefixes. It is not searched when the compiler is built as a
460 @defmac INIT_ENVIRONMENT
461 Define this macro as a C string constant if you wish to set environment
462 variables for programs called by the driver, such as the assembler and
463 loader. The driver passes the value of this macro to @code{putenv} to
464 initialize the necessary environment variables.
467 @defmac LOCAL_INCLUDE_DIR
468 Define this macro as a C string constant if you wish to override the
469 standard choice of @file{/usr/local/include} as the default prefix to
470 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
471 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
472 @file{config.gcc}, normally @file{/usr/include}) in the search order.
474 Cross compilers do not search either @file{/usr/local/include} or its
478 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
479 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
480 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
481 If you do not define this macro, no component is used.
484 @defmac INCLUDE_DEFAULTS
485 Define this macro if you wish to override the entire default search path
486 for include files. For a native compiler, the default search path
487 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
488 @code{GPLUSPLUS_INCLUDE_DIR}, and
489 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
490 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
491 and specify private search areas for GCC@. The directory
492 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
494 The definition should be an initializer for an array of structures.
495 Each array element should have four elements: the directory name (a
496 string constant), the component name (also a string constant), a flag
497 for C++-only directories,
498 and a flag showing that the includes in the directory don't need to be
499 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
500 the array with a null element.
502 The component name denotes what GNU package the include file is part of,
503 if any, in all uppercase letters. For example, it might be @samp{GCC}
504 or @samp{BINUTILS}. If the package is part of a vendor-supplied
505 operating system, code the component name as @samp{0}.
507 For example, here is the definition used for VAX/VMS:
510 #define INCLUDE_DEFAULTS \
512 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
513 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
514 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
521 Here is the order of prefixes tried for exec files:
525 Any prefixes specified by the user with @option{-B}.
528 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
529 is not set and the compiler has not been installed in the configure-time
530 @var{prefix}, the location in which the compiler has actually been installed.
533 The directories specified by the environment variable @code{COMPILER_PATH}.
536 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
537 in the configured-time @var{prefix}.
540 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
543 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
546 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
550 Here is the order of prefixes tried for startfiles:
554 Any prefixes specified by the user with @option{-B}.
557 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
558 value based on the installed toolchain location.
561 The directories specified by the environment variable @code{LIBRARY_PATH}
562 (or port-specific name; native only, cross compilers do not use this).
565 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
566 in the configured @var{prefix} or this is a native compiler.
569 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
572 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
576 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
577 native compiler, or we have a target system root.
580 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
581 native compiler, or we have a target system root.
584 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
585 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
586 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
589 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
590 compiler, or we have a target system root. The default for this macro is
594 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
595 compiler, or we have a target system root. The default for this macro is
599 @node Run-time Target
600 @section Run-time Target Specification
601 @cindex run-time target specification
602 @cindex predefined macros
603 @cindex target specifications
605 @c prevent bad page break with this line
606 Here are run-time target specifications.
608 @defmac TARGET_CPU_CPP_BUILTINS ()
609 This function-like macro expands to a block of code that defines
610 built-in preprocessor macros and assertions for the target CPU, using
611 the functions @code{builtin_define}, @code{builtin_define_std} and
612 @code{builtin_assert}. When the front end
613 calls this macro it provides a trailing semicolon, and since it has
614 finished command line option processing your code can use those
617 @code{builtin_assert} takes a string in the form you pass to the
618 command-line option @option{-A}, such as @code{cpu=mips}, and creates
619 the assertion. @code{builtin_define} takes a string in the form
620 accepted by option @option{-D} and unconditionally defines the macro.
622 @code{builtin_define_std} takes a string representing the name of an
623 object-like macro. If it doesn't lie in the user's namespace,
624 @code{builtin_define_std} defines it unconditionally. Otherwise, it
625 defines a version with two leading underscores, and another version
626 with two leading and trailing underscores, and defines the original
627 only if an ISO standard was not requested on the command line. For
628 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
629 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
630 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
631 defines only @code{_ABI64}.
633 You can also test for the C dialect being compiled. The variable
634 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
635 or @code{clk_objective_c}. Note that if we are preprocessing
636 assembler, this variable will be @code{clk_c} but the function-like
637 macro @code{preprocessing_asm_p()} will return true, so you might want
638 to check for that first. If you need to check for strict ANSI, the
639 variable @code{flag_iso} can be used. The function-like macro
640 @code{preprocessing_trad_p()} can be used to check for traditional
644 @defmac TARGET_OS_CPP_BUILTINS ()
645 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
646 and is used for the target operating system instead.
649 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
650 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
651 and is used for the target object format. @file{elfos.h} uses this
652 macro to define @code{__ELF__}, so you probably do not need to define
656 @deftypevar {extern int} target_flags
657 This variable is declared in @file{options.h}, which is included before
658 any target-specific headers.
661 @hook TARGET_DEFAULT_TARGET_FLAGS
662 This variable specifies the initial value of @code{target_flags}.
663 Its default setting is 0.
666 @cindex optional hardware or system features
667 @cindex features, optional, in system conventions
669 @hook TARGET_HANDLE_OPTION
670 This hook is called whenever the user specifies one of the
671 target-specific options described by the @file{.opt} definition files
672 (@pxref{Options}). It has the opportunity to do some option-specific
673 processing and should return true if the option is valid. The default
674 definition does nothing but return true.
676 @var{decoded} specifies the option and its arguments. @var{opts} and
677 @var{opts_set} are the @code{gcc_options} structures to be used for
678 storing option state, and @var{loc} is the location at which the
679 option was passed (@code{UNKNOWN_LOCATION} except for options passed
683 @hook TARGET_HANDLE_C_OPTION
684 This target hook is called whenever the user specifies one of the
685 target-specific C language family options described by the @file{.opt}
686 definition files(@pxref{Options}). It has the opportunity to do some
687 option-specific processing and should return true if the option is
688 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
689 default definition does nothing but return false.
691 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
692 options. However, if processing an option requires routines that are
693 only available in the C (and related language) front ends, then you
694 should use @code{TARGET_HANDLE_C_OPTION} instead.
697 @hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT
699 @hook TARGET_STRING_OBJECT_REF_TYPE_P
701 @hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG
703 @hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE
704 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
705 but is called when the optimize level is changed via an attribute or
706 pragma or when it is reset at the end of the code affected by the
707 attribute or pragma. It is not called at the beginning of compilation
708 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
709 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
710 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
713 @defmac C_COMMON_OVERRIDE_OPTIONS
714 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
715 but is only used in the C
716 language frontends (C, Objective-C, C++, Objective-C++) and so can be
717 used to alter option flag variables which only exist in those
721 @hook TARGET_OPTION_OPTIMIZATION_TABLE
722 Some machines may desire to change what optimizations are performed for
723 various optimization levels. This variable, if defined, describes
724 options to enable at particular sets of optimization levels. These
725 options are processed once
726 just after the optimization level is determined and before the remainder
727 of the command options have been parsed, so may be overridden by other
728 options passed explicitly.
730 This processing is run once at program startup and when the optimization
731 options are changed via @code{#pragma GCC optimize} or by using the
732 @code{optimize} attribute.
735 @hook TARGET_OPTION_INIT_STRUCT
737 @hook TARGET_OPTION_DEFAULT_PARAMS
739 @defmac SWITCHABLE_TARGET
740 Some targets need to switch between substantially different subtargets
741 during compilation. For example, the MIPS target has one subtarget for
742 the traditional MIPS architecture and another for MIPS16. Source code
743 can switch between these two subarchitectures using the @code{mips16}
744 and @code{nomips16} attributes.
746 Such subtargets can differ in things like the set of available
747 registers, the set of available instructions, the costs of various
748 operations, and so on. GCC caches a lot of this type of information
749 in global variables, and recomputing them for each subtarget takes a
750 significant amount of time. The compiler therefore provides a facility
751 for maintaining several versions of the global variables and quickly
752 switching between them; see @file{target-globals.h} for details.
754 Define this macro to 1 if your target needs this facility. The default
758 @node Per-Function Data
759 @section Defining data structures for per-function information.
760 @cindex per-function data
761 @cindex data structures
763 If the target needs to store information on a per-function basis, GCC
764 provides a macro and a couple of variables to allow this. Note, just
765 using statics to store the information is a bad idea, since GCC supports
766 nested functions, so you can be halfway through encoding one function
767 when another one comes along.
769 GCC defines a data structure called @code{struct function} which
770 contains all of the data specific to an individual function. This
771 structure contains a field called @code{machine} whose type is
772 @code{struct machine_function *}, which can be used by targets to point
773 to their own specific data.
775 If a target needs per-function specific data it should define the type
776 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
777 This macro should be used to initialize the function pointer
778 @code{init_machine_status}. This pointer is explained below.
780 One typical use of per-function, target specific data is to create an
781 RTX to hold the register containing the function's return address. This
782 RTX can then be used to implement the @code{__builtin_return_address}
783 function, for level 0.
785 Note---earlier implementations of GCC used a single data area to hold
786 all of the per-function information. Thus when processing of a nested
787 function began the old per-function data had to be pushed onto a
788 stack, and when the processing was finished, it had to be popped off the
789 stack. GCC used to provide function pointers called
790 @code{save_machine_status} and @code{restore_machine_status} to handle
791 the saving and restoring of the target specific information. Since the
792 single data area approach is no longer used, these pointers are no
795 @defmac INIT_EXPANDERS
796 Macro called to initialize any target specific information. This macro
797 is called once per function, before generation of any RTL has begun.
798 The intention of this macro is to allow the initialization of the
799 function pointer @code{init_machine_status}.
802 @deftypevar {void (*)(struct function *)} init_machine_status
803 If this function pointer is non-@code{NULL} it will be called once per
804 function, before function compilation starts, in order to allow the
805 target to perform any target specific initialization of the
806 @code{struct function} structure. It is intended that this would be
807 used to initialize the @code{machine} of that structure.
809 @code{struct machine_function} structures are expected to be freed by GC@.
810 Generally, any memory that they reference must be allocated by using
811 GC allocation, including the structure itself.
815 @section Storage Layout
816 @cindex storage layout
818 Note that the definitions of the macros in this table which are sizes or
819 alignments measured in bits do not need to be constant. They can be C
820 expressions that refer to static variables, such as the @code{target_flags}.
821 @xref{Run-time Target}.
823 @defmac BITS_BIG_ENDIAN
824 Define this macro to have the value 1 if the most significant bit in a
825 byte has the lowest number; otherwise define it to have the value zero.
826 This means that bit-field instructions count from the most significant
827 bit. If the machine has no bit-field instructions, then this must still
828 be defined, but it doesn't matter which value it is defined to. This
829 macro need not be a constant.
831 This macro does not affect the way structure fields are packed into
832 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
835 @defmac BYTES_BIG_ENDIAN
836 Define this macro to have the value 1 if the most significant byte in a
837 word has the lowest number. This macro need not be a constant.
840 @defmac WORDS_BIG_ENDIAN
841 Define this macro to have the value 1 if, in a multiword object, the
842 most significant word has the lowest number. This applies to both
843 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
844 order of words in memory is not the same as the order in registers. This
845 macro need not be a constant.
848 @defmac REG_WORDS_BIG_ENDIAN
849 On some machines, the order of words in a multiword object differs between
850 registers in memory. In such a situation, define this macro to describe
851 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
852 the order of words in memory.
855 @defmac FLOAT_WORDS_BIG_ENDIAN
856 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
857 @code{TFmode} floating point numbers are stored in memory with the word
858 containing the sign bit at the lowest address; otherwise define it to
859 have the value 0. This macro need not be a constant.
861 You need not define this macro if the ordering is the same as for
865 @defmac BITS_PER_UNIT
866 Define this macro to be the number of bits in an addressable storage
867 unit (byte). If you do not define this macro the default is 8.
870 @defmac BITS_PER_WORD
871 Number of bits in a word. If you do not define this macro, the default
872 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
875 @defmac MAX_BITS_PER_WORD
876 Maximum number of bits in a word. If this is undefined, the default is
877 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
878 largest value that @code{BITS_PER_WORD} can have at run-time.
881 @defmac UNITS_PER_WORD
882 Number of storage units in a word; normally the size of a general-purpose
883 register, a power of two from 1 or 8.
886 @defmac MIN_UNITS_PER_WORD
887 Minimum number of units in a word. If this is undefined, the default is
888 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
889 smallest value that @code{UNITS_PER_WORD} can have at run-time.
893 Width of a pointer, in bits. You must specify a value no wider than the
894 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
895 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
896 a value the default is @code{BITS_PER_WORD}.
899 @defmac POINTERS_EXTEND_UNSIGNED
900 A C expression that determines how pointers should be extended from
901 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
902 greater than zero if pointers should be zero-extended, zero if they
903 should be sign-extended, and negative if some other sort of conversion
904 is needed. In the last case, the extension is done by the target's
905 @code{ptr_extend} instruction.
907 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
908 and @code{word_mode} are all the same width.
911 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
912 A macro to update @var{m} and @var{unsignedp} when an object whose type
913 is @var{type} and which has the specified mode and signedness is to be
914 stored in a register. This macro is only called when @var{type} is a
917 On most RISC machines, which only have operations that operate on a full
918 register, define this macro to set @var{m} to @code{word_mode} if
919 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
920 cases, only integer modes should be widened because wider-precision
921 floating-point operations are usually more expensive than their narrower
924 For most machines, the macro definition does not change @var{unsignedp}.
925 However, some machines, have instructions that preferentially handle
926 either signed or unsigned quantities of certain modes. For example, on
927 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
928 sign-extend the result to 64 bits. On such machines, set
929 @var{unsignedp} according to which kind of extension is more efficient.
931 Do not define this macro if it would never modify @var{m}.
934 @hook TARGET_PROMOTE_FUNCTION_MODE
935 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
936 function return values. The target hook should return the new mode
937 and possibly change @code{*@var{punsignedp}} if the promotion should
938 change signedness. This function is called only for scalar @emph{or
941 @var{for_return} allows to distinguish the promotion of arguments and
942 return values. If it is @code{1}, a return value is being promoted and
943 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
944 If it is @code{2}, the returned mode should be that of the register in
945 which an incoming parameter is copied, or the outgoing result is computed;
946 then the hook should return the same mode as @code{promote_mode}, though
947 the signedness may be different.
949 @var{type} can be NULL when promoting function arguments of libcalls.
951 The default is to not promote arguments and return values. You can
952 also define the hook to @code{default_promote_function_mode_always_promote}
953 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
956 @defmac PARM_BOUNDARY
957 Normal alignment required for function parameters on the stack, in
958 bits. All stack parameters receive at least this much alignment
959 regardless of data type. On most machines, this is the same as the
963 @defmac STACK_BOUNDARY
964 Define this macro to the minimum alignment enforced by hardware for the
965 stack pointer on this machine. The definition is a C expression for the
966 desired alignment (measured in bits). This value is used as a default
967 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
968 this should be the same as @code{PARM_BOUNDARY}.
971 @defmac PREFERRED_STACK_BOUNDARY
972 Define this macro if you wish to preserve a certain alignment for the
973 stack pointer, greater than what the hardware enforces. The definition
974 is a C expression for the desired alignment (measured in bits). This
975 macro must evaluate to a value equal to or larger than
976 @code{STACK_BOUNDARY}.
979 @defmac INCOMING_STACK_BOUNDARY
980 Define this macro if the incoming stack boundary may be different
981 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
982 to a value equal to or larger than @code{STACK_BOUNDARY}.
985 @defmac FUNCTION_BOUNDARY
986 Alignment required for a function entry point, in bits.
989 @defmac BIGGEST_ALIGNMENT
990 Biggest alignment that any data type can require on this machine, in
991 bits. Note that this is not the biggest alignment that is supported,
992 just the biggest alignment that, when violated, may cause a fault.
995 @defmac MALLOC_ABI_ALIGNMENT
996 Alignment, in bits, a C conformant malloc implementation has to
997 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1000 @defmac ATTRIBUTE_ALIGNED_VALUE
1001 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1002 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1005 @defmac MINIMUM_ATOMIC_ALIGNMENT
1006 If defined, the smallest alignment, in bits, that can be given to an
1007 object that can be referenced in one operation, without disturbing any
1008 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1009 on machines that don't have byte or half-word store operations.
1012 @defmac BIGGEST_FIELD_ALIGNMENT
1013 Biggest alignment that any structure or union field can require on this
1014 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1015 structure and union fields only, unless the field alignment has been set
1016 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1019 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1020 An expression for the alignment of a structure field @var{field} if the
1021 alignment computed in the usual way (including applying of
1022 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1023 alignment) is @var{computed}. It overrides alignment only if the
1024 field alignment has not been set by the
1025 @code{__attribute__ ((aligned (@var{n})))} construct.
1028 @defmac MAX_STACK_ALIGNMENT
1029 Biggest stack alignment guaranteed by the backend. Use this macro
1030 to specify the maximum alignment of a variable on stack.
1032 If not defined, the default value is @code{STACK_BOUNDARY}.
1034 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1035 @c But the fix for PR 32893 indicates that we can only guarantee
1036 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1037 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1040 @defmac MAX_OFILE_ALIGNMENT
1041 Biggest alignment supported by the object file format of this machine.
1042 Use this macro to limit the alignment which can be specified using the
1043 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1044 the default value is @code{BIGGEST_ALIGNMENT}.
1046 On systems that use ELF, the default (in @file{config/elfos.h}) is
1047 the largest supported 32-bit ELF section alignment representable on
1048 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1049 On 32-bit ELF the largest supported section alignment in bits is
1050 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1053 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1054 If defined, a C expression to compute the alignment for a variable in
1055 the static store. @var{type} is the data type, and @var{basic-align} is
1056 the alignment that the object would ordinarily have. The value of this
1057 macro is used instead of that alignment to align the object.
1059 If this macro is not defined, then @var{basic-align} is used.
1062 One use of this macro is to increase alignment of medium-size data to
1063 make it all fit in fewer cache lines. Another is to cause character
1064 arrays to be word-aligned so that @code{strcpy} calls that copy
1065 constants to character arrays can be done inline.
1068 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1069 If defined, a C expression to compute the alignment given to a constant
1070 that is being placed in memory. @var{constant} is the constant and
1071 @var{basic-align} is the alignment that the object would ordinarily
1072 have. The value of this macro is used instead of that alignment to
1075 If this macro is not defined, then @var{basic-align} is used.
1077 The typical use of this macro is to increase alignment for string
1078 constants to be word aligned so that @code{strcpy} calls that copy
1079 constants can be done inline.
1082 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1083 If defined, a C expression to compute the alignment for a variable in
1084 the local store. @var{type} is the data type, and @var{basic-align} is
1085 the alignment that the object would ordinarily have. The value of this
1086 macro is used instead of that alignment to align the object.
1088 If this macro is not defined, then @var{basic-align} is used.
1090 One use of this macro is to increase alignment of medium-size data to
1091 make it all fit in fewer cache lines.
1093 If the value of this macro has a type, it should be an unsigned type.
1096 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1097 If defined, a C expression to compute the alignment for stack slot.
1098 @var{type} is the data type, @var{mode} is the widest mode available,
1099 and @var{basic-align} is the alignment that the slot would ordinarily
1100 have. The value of this macro is used instead of that alignment to
1103 If this macro is not defined, then @var{basic-align} is used when
1104 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1107 This macro is to set alignment of stack slot to the maximum alignment
1108 of all possible modes which the slot may have.
1110 If the value of this macro has a type, it should be an unsigned type.
1113 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1114 If defined, a C expression to compute the alignment for a local
1115 variable @var{decl}.
1117 If this macro is not defined, then
1118 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1121 One use of this macro is to increase alignment of medium-size data to
1122 make it all fit in fewer cache lines.
1124 If the value of this macro has a type, it should be an unsigned type.
1127 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1128 If defined, a C expression to compute the minimum required alignment
1129 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1130 @var{mode}, assuming normal alignment @var{align}.
1132 If this macro is not defined, then @var{align} will be used.
1135 @defmac EMPTY_FIELD_BOUNDARY
1136 Alignment in bits to be given to a structure bit-field that follows an
1137 empty field such as @code{int : 0;}.
1139 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1142 @defmac STRUCTURE_SIZE_BOUNDARY
1143 Number of bits which any structure or union's size must be a multiple of.
1144 Each structure or union's size is rounded up to a multiple of this.
1146 If you do not define this macro, the default is the same as
1147 @code{BITS_PER_UNIT}.
1150 @defmac STRICT_ALIGNMENT
1151 Define this macro to be the value 1 if instructions will fail to work
1152 if given data not on the nominal alignment. If instructions will merely
1153 go slower in that case, define this macro as 0.
1156 @defmac PCC_BITFIELD_TYPE_MATTERS
1157 Define this if you wish to imitate the way many other C compilers handle
1158 alignment of bit-fields and the structures that contain them.
1160 The behavior is that the type written for a named bit-field (@code{int},
1161 @code{short}, or other integer type) imposes an alignment for the entire
1162 structure, as if the structure really did contain an ordinary field of
1163 that type. In addition, the bit-field is placed within the structure so
1164 that it would fit within such a field, not crossing a boundary for it.
1166 Thus, on most machines, a named bit-field whose type is written as
1167 @code{int} would not cross a four-byte boundary, and would force
1168 four-byte alignment for the whole structure. (The alignment used may
1169 not be four bytes; it is controlled by the other alignment parameters.)
1171 An unnamed bit-field will not affect the alignment of the containing
1174 If the macro is defined, its definition should be a C expression;
1175 a nonzero value for the expression enables this behavior.
1177 Note that if this macro is not defined, or its value is zero, some
1178 bit-fields may cross more than one alignment boundary. The compiler can
1179 support such references if there are @samp{insv}, @samp{extv}, and
1180 @samp{extzv} insns that can directly reference memory.
1182 The other known way of making bit-fields work is to define
1183 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1184 Then every structure can be accessed with fullwords.
1186 Unless the machine has bit-field instructions or you define
1187 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1188 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1190 If your aim is to make GCC use the same conventions for laying out
1191 bit-fields as are used by another compiler, here is how to investigate
1192 what the other compiler does. Compile and run this program:
1211 printf ("Size of foo1 is %d\n",
1212 sizeof (struct foo1));
1213 printf ("Size of foo2 is %d\n",
1214 sizeof (struct foo2));
1219 If this prints 2 and 5, then the compiler's behavior is what you would
1220 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1223 @defmac BITFIELD_NBYTES_LIMITED
1224 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1225 to aligning a bit-field within the structure.
1228 @hook TARGET_ALIGN_ANON_BITFIELD
1229 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1230 whether unnamed bitfields affect the alignment of the containing
1231 structure. The hook should return true if the structure should inherit
1232 the alignment requirements of an unnamed bitfield's type.
1235 @hook TARGET_NARROW_VOLATILE_BITFIELD
1236 This target hook should return @code{true} if accesses to volatile bitfields
1237 should use the narrowest mode possible. It should return @code{false} if
1238 these accesses should use the bitfield container type.
1240 The default is @code{!TARGET_STRICT_ALIGN}.
1243 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1244 Return 1 if a structure or array containing @var{field} should be accessed using
1247 If @var{field} is the only field in the structure, @var{mode} is its
1248 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1249 case where structures of one field would require the structure's mode to
1250 retain the field's mode.
1252 Normally, this is not needed.
1255 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1256 Define this macro as an expression for the alignment of a type (given
1257 by @var{type} as a tree node) if the alignment computed in the usual
1258 way is @var{computed} and the alignment explicitly specified was
1261 The default is to use @var{specified} if it is larger; otherwise, use
1262 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1265 @defmac MAX_FIXED_MODE_SIZE
1266 An integer expression for the size in bits of the largest integer
1267 machine mode that should actually be used. All integer machine modes of
1268 this size or smaller can be used for structures and unions with the
1269 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1270 (DImode)} is assumed.
1273 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1274 If defined, an expression of type @code{enum machine_mode} that
1275 specifies the mode of the save area operand of a
1276 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1277 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1278 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1279 having its mode specified.
1281 You need not define this macro if it always returns @code{Pmode}. You
1282 would most commonly define this macro if the
1283 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1287 @defmac STACK_SIZE_MODE
1288 If defined, an expression of type @code{enum machine_mode} that
1289 specifies the mode of the size increment operand of an
1290 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1292 You need not define this macro if it always returns @code{word_mode}.
1293 You would most commonly define this macro if the @code{allocate_stack}
1294 pattern needs to support both a 32- and a 64-bit mode.
1297 @hook TARGET_LIBGCC_CMP_RETURN_MODE
1298 This target hook should return the mode to be used for the return value
1299 of compare instructions expanded to libgcc calls. If not defined
1300 @code{word_mode} is returned which is the right choice for a majority of
1304 @hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1305 This target hook should return the mode to be used for the shift count operand
1306 of shift instructions expanded to libgcc calls. If not defined
1307 @code{word_mode} is returned which is the right choice for a majority of
1311 @hook TARGET_UNWIND_WORD_MODE
1312 Return machine mode to be used for @code{_Unwind_Word} type.
1313 The default is to use @code{word_mode}.
1316 @defmac ROUND_TOWARDS_ZERO
1317 If defined, this macro should be true if the prevailing rounding
1318 mode is towards zero.
1320 Defining this macro only affects the way @file{libgcc.a} emulates
1321 floating-point arithmetic.
1323 Not defining this macro is equivalent to returning zero.
1326 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1327 This macro should return true if floats with @var{size}
1328 bits do not have a NaN or infinity representation, but use the largest
1329 exponent for normal numbers instead.
1331 Defining this macro only affects the way @file{libgcc.a} emulates
1332 floating-point arithmetic.
1334 The default definition of this macro returns false for all sizes.
1337 @hook TARGET_MS_BITFIELD_LAYOUT_P
1338 This target hook returns @code{true} if bit-fields in the given
1339 @var{record_type} are to be laid out following the rules of Microsoft
1340 Visual C/C++, namely: (i) a bit-field won't share the same storage
1341 unit with the previous bit-field if their underlying types have
1342 different sizes, and the bit-field will be aligned to the highest
1343 alignment of the underlying types of itself and of the previous
1344 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1345 the whole enclosing structure, even if it is unnamed; except that
1346 (iii) a zero-sized bit-field will be disregarded unless it follows
1347 another bit-field of nonzero size. If this hook returns @code{true},
1348 other macros that control bit-field layout are ignored.
1350 When a bit-field is inserted into a packed record, the whole size
1351 of the underlying type is used by one or more same-size adjacent
1352 bit-fields (that is, if its long:3, 32 bits is used in the record,
1353 and any additional adjacent long bit-fields are packed into the same
1354 chunk of 32 bits. However, if the size changes, a new field of that
1355 size is allocated). In an unpacked record, this is the same as using
1356 alignment, but not equivalent when packing.
1358 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1359 the latter will take precedence. If @samp{__attribute__((packed))} is
1360 used on a single field when MS bit-fields are in use, it will take
1361 precedence for that field, but the alignment of the rest of the structure
1362 may affect its placement.
1365 @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1366 Returns true if the target supports decimal floating point.
1369 @hook TARGET_FIXED_POINT_SUPPORTED_P
1370 Returns true if the target supports fixed-point arithmetic.
1373 @hook TARGET_EXPAND_TO_RTL_HOOK
1374 This hook is called just before expansion into rtl, allowing the target
1375 to perform additional initializations or analysis before the expansion.
1376 For example, the rs6000 port uses it to allocate a scratch stack slot
1377 for use in copying SDmode values between memory and floating point
1378 registers whenever the function being expanded has any SDmode
1382 @hook TARGET_INSTANTIATE_DECLS
1383 This hook allows the backend to perform additional instantiations on rtl
1384 that are not actually in any insns yet, but will be later.
1387 @hook TARGET_MANGLE_TYPE
1388 If your target defines any fundamental types, or any types your target
1389 uses should be mangled differently from the default, define this hook
1390 to return the appropriate encoding for these types as part of a C++
1391 mangled name. The @var{type} argument is the tree structure representing
1392 the type to be mangled. The hook may be applied to trees which are
1393 not target-specific fundamental types; it should return @code{NULL}
1394 for all such types, as well as arguments it does not recognize. If the
1395 return value is not @code{NULL}, it must point to a statically-allocated
1398 Target-specific fundamental types might be new fundamental types or
1399 qualified versions of ordinary fundamental types. Encode new
1400 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1401 is the name used for the type in source code, and @var{n} is the
1402 length of @var{name} in decimal. Encode qualified versions of
1403 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1404 @var{name} is the name used for the type qualifier in source code,
1405 @var{n} is the length of @var{name} as above, and @var{code} is the
1406 code used to represent the unqualified version of this type. (See
1407 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1408 codes.) In both cases the spaces are for clarity; do not include any
1409 spaces in your string.
1411 This hook is applied to types prior to typedef resolution. If the mangled
1412 name for a particular type depends only on that type's main variant, you
1413 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1416 The default version of this hook always returns @code{NULL}, which is
1417 appropriate for a target that does not define any new fundamental
1422 @section Layout of Source Language Data Types
1424 These macros define the sizes and other characteristics of the standard
1425 basic data types used in programs being compiled. Unlike the macros in
1426 the previous section, these apply to specific features of C and related
1427 languages, rather than to fundamental aspects of storage layout.
1429 @defmac INT_TYPE_SIZE
1430 A C expression for the size in bits of the type @code{int} on the
1431 target machine. If you don't define this, the default is one word.
1434 @defmac SHORT_TYPE_SIZE
1435 A C expression for the size in bits of the type @code{short} on the
1436 target machine. If you don't define this, the default is half a word.
1437 (If this would be less than one storage unit, it is rounded up to one
1441 @defmac LONG_TYPE_SIZE
1442 A C expression for the size in bits of the type @code{long} on the
1443 target machine. If you don't define this, the default is one word.
1446 @defmac ADA_LONG_TYPE_SIZE
1447 On some machines, the size used for the Ada equivalent of the type
1448 @code{long} by a native Ada compiler differs from that used by C@. In
1449 that situation, define this macro to be a C expression to be used for
1450 the size of that type. If you don't define this, the default is the
1451 value of @code{LONG_TYPE_SIZE}.
1454 @defmac LONG_LONG_TYPE_SIZE
1455 A C expression for the size in bits of the type @code{long long} on the
1456 target machine. If you don't define this, the default is two
1457 words. If you want to support GNU Ada on your machine, the value of this
1458 macro must be at least 64.
1461 @defmac CHAR_TYPE_SIZE
1462 A C expression for the size in bits of the type @code{char} on the
1463 target machine. If you don't define this, the default is
1464 @code{BITS_PER_UNIT}.
1467 @defmac BOOL_TYPE_SIZE
1468 A C expression for the size in bits of the C++ type @code{bool} and
1469 C99 type @code{_Bool} on the target machine. If you don't define
1470 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1473 @defmac FLOAT_TYPE_SIZE
1474 A C expression for the size in bits of the type @code{float} on the
1475 target machine. If you don't define this, the default is one word.
1478 @defmac DOUBLE_TYPE_SIZE
1479 A C expression for the size in bits of the type @code{double} on the
1480 target machine. If you don't define this, the default is two
1484 @defmac LONG_DOUBLE_TYPE_SIZE
1485 A C expression for the size in bits of the type @code{long double} on
1486 the target machine. If you don't define this, the default is two
1490 @defmac SHORT_FRACT_TYPE_SIZE
1491 A C expression for the size in bits of the type @code{short _Fract} on
1492 the target machine. If you don't define this, the default is
1493 @code{BITS_PER_UNIT}.
1496 @defmac FRACT_TYPE_SIZE
1497 A C expression for the size in bits of the type @code{_Fract} on
1498 the target machine. If you don't define this, the default is
1499 @code{BITS_PER_UNIT * 2}.
1502 @defmac LONG_FRACT_TYPE_SIZE
1503 A C expression for the size in bits of the type @code{long _Fract} on
1504 the target machine. If you don't define this, the default is
1505 @code{BITS_PER_UNIT * 4}.
1508 @defmac LONG_LONG_FRACT_TYPE_SIZE
1509 A C expression for the size in bits of the type @code{long long _Fract} on
1510 the target machine. If you don't define this, the default is
1511 @code{BITS_PER_UNIT * 8}.
1514 @defmac SHORT_ACCUM_TYPE_SIZE
1515 A C expression for the size in bits of the type @code{short _Accum} on
1516 the target machine. If you don't define this, the default is
1517 @code{BITS_PER_UNIT * 2}.
1520 @defmac ACCUM_TYPE_SIZE
1521 A C expression for the size in bits of the type @code{_Accum} on
1522 the target machine. If you don't define this, the default is
1523 @code{BITS_PER_UNIT * 4}.
1526 @defmac LONG_ACCUM_TYPE_SIZE
1527 A C expression for the size in bits of the type @code{long _Accum} on
1528 the target machine. If you don't define this, the default is
1529 @code{BITS_PER_UNIT * 8}.
1532 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1533 A C expression for the size in bits of the type @code{long long _Accum} on
1534 the target machine. If you don't define this, the default is
1535 @code{BITS_PER_UNIT * 16}.
1538 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1539 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1540 if you want routines in @file{libgcc2.a} for a size other than
1541 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1542 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1545 @defmac LIBGCC2_HAS_DF_MODE
1546 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1547 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1548 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1549 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1550 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1554 @defmac LIBGCC2_HAS_XF_MODE
1555 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1556 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1557 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1558 is 80 then the default is 1, otherwise it is 0.
1561 @defmac LIBGCC2_HAS_TF_MODE
1562 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1563 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1564 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1565 is 128 then the default is 1, otherwise it is 0.
1568 @defmac LIBGCC2_GNU_PREFIX
1569 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1570 hook and should be defined if that hook is overriden to be true. It
1571 causes function names in libgcc to be changed to use a @code{__gnu_}
1572 prefix for their name rather than the default @code{__}. A port which
1573 uses this macro should also arrange to use @file{t-gnu-prefix} in
1574 the libgcc @file{config.host}.
1581 Define these macros to be the size in bits of the mantissa of
1582 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1583 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1584 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1585 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1586 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1587 @code{DOUBLE_TYPE_SIZE} or
1588 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1591 @defmac TARGET_FLT_EVAL_METHOD
1592 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1593 assuming, if applicable, that the floating-point control word is in its
1594 default state. If you do not define this macro the value of
1595 @code{FLT_EVAL_METHOD} will be zero.
1598 @defmac WIDEST_HARDWARE_FP_SIZE
1599 A C expression for the size in bits of the widest floating-point format
1600 supported by the hardware. If you define this macro, you must specify a
1601 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1602 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1606 @defmac DEFAULT_SIGNED_CHAR
1607 An expression whose value is 1 or 0, according to whether the type
1608 @code{char} should be signed or unsigned by default. The user can
1609 always override this default with the options @option{-fsigned-char}
1610 and @option{-funsigned-char}.
1613 @hook TARGET_DEFAULT_SHORT_ENUMS
1614 This target hook should return true if the compiler should give an
1615 @code{enum} type only as many bytes as it takes to represent the range
1616 of possible values of that type. It should return false if all
1617 @code{enum} types should be allocated like @code{int}.
1619 The default is to return false.
1623 A C expression for a string describing the name of the data type to use
1624 for size values. The typedef name @code{size_t} is defined using the
1625 contents of the string.
1627 The string can contain more than one keyword. If so, separate them with
1628 spaces, and write first any length keyword, then @code{unsigned} if
1629 appropriate, and finally @code{int}. The string must exactly match one
1630 of the data type names defined in the function
1631 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1632 omit @code{int} or change the order---that would cause the compiler to
1635 If you don't define this macro, the default is @code{"long unsigned
1639 @defmac PTRDIFF_TYPE
1640 A C expression for a string describing the name of the data type to use
1641 for the result of subtracting two pointers. The typedef name
1642 @code{ptrdiff_t} is defined using the contents of the string. See
1643 @code{SIZE_TYPE} above for more information.
1645 If you don't define this macro, the default is @code{"long int"}.
1649 A C expression for a string describing the name of the data type to use
1650 for wide characters. The typedef name @code{wchar_t} is defined using
1651 the contents of the string. See @code{SIZE_TYPE} above for more
1654 If you don't define this macro, the default is @code{"int"}.
1657 @defmac WCHAR_TYPE_SIZE
1658 A C expression for the size in bits of the data type for wide
1659 characters. This is used in @code{cpp}, which cannot make use of
1664 A C expression for a string describing the name of the data type to
1665 use for wide characters passed to @code{printf} and returned from
1666 @code{getwc}. The typedef name @code{wint_t} is defined using the
1667 contents of the string. See @code{SIZE_TYPE} above for more
1670 If you don't define this macro, the default is @code{"unsigned int"}.
1674 A C expression for a string describing the name of the data type that
1675 can represent any value of any standard or extended signed integer type.
1676 The typedef name @code{intmax_t} is defined using the contents of the
1677 string. See @code{SIZE_TYPE} above for more information.
1679 If you don't define this macro, the default is the first of
1680 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1681 much precision as @code{long long int}.
1684 @defmac UINTMAX_TYPE
1685 A C expression for a string describing the name of the data type that
1686 can represent any value of any standard or extended unsigned integer
1687 type. The typedef name @code{uintmax_t} is defined using the contents
1688 of the string. See @code{SIZE_TYPE} above for more information.
1690 If you don't define this macro, the default is the first of
1691 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1692 unsigned int"} that has as much precision as @code{long long unsigned
1696 @defmac SIG_ATOMIC_TYPE
1702 @defmacx UINT16_TYPE
1703 @defmacx UINT32_TYPE
1704 @defmacx UINT64_TYPE
1705 @defmacx INT_LEAST8_TYPE
1706 @defmacx INT_LEAST16_TYPE
1707 @defmacx INT_LEAST32_TYPE
1708 @defmacx INT_LEAST64_TYPE
1709 @defmacx UINT_LEAST8_TYPE
1710 @defmacx UINT_LEAST16_TYPE
1711 @defmacx UINT_LEAST32_TYPE
1712 @defmacx UINT_LEAST64_TYPE
1713 @defmacx INT_FAST8_TYPE
1714 @defmacx INT_FAST16_TYPE
1715 @defmacx INT_FAST32_TYPE
1716 @defmacx INT_FAST64_TYPE
1717 @defmacx UINT_FAST8_TYPE
1718 @defmacx UINT_FAST16_TYPE
1719 @defmacx UINT_FAST32_TYPE
1720 @defmacx UINT_FAST64_TYPE
1721 @defmacx INTPTR_TYPE
1722 @defmacx UINTPTR_TYPE
1723 C expressions for the standard types @code{sig_atomic_t},
1724 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1725 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1726 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1727 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1728 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1729 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1730 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1731 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1732 @code{SIZE_TYPE} above for more information.
1734 If any of these macros evaluates to a null pointer, the corresponding
1735 type is not supported; if GCC is configured to provide
1736 @code{<stdint.h>} in such a case, the header provided may not conform
1737 to C99, depending on the type in question. The defaults for all of
1738 these macros are null pointers.
1741 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1742 The C++ compiler represents a pointer-to-member-function with a struct
1749 ptrdiff_t vtable_index;
1756 The C++ compiler must use one bit to indicate whether the function that
1757 will be called through a pointer-to-member-function is virtual.
1758 Normally, we assume that the low-order bit of a function pointer must
1759 always be zero. Then, by ensuring that the vtable_index is odd, we can
1760 distinguish which variant of the union is in use. But, on some
1761 platforms function pointers can be odd, and so this doesn't work. In
1762 that case, we use the low-order bit of the @code{delta} field, and shift
1763 the remainder of the @code{delta} field to the left.
1765 GCC will automatically make the right selection about where to store
1766 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1767 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1768 set such that functions always start at even addresses, but the lowest
1769 bit of pointers to functions indicate whether the function at that
1770 address is in ARM or Thumb mode. If this is the case of your
1771 architecture, you should define this macro to
1772 @code{ptrmemfunc_vbit_in_delta}.
1774 In general, you should not have to define this macro. On architectures
1775 in which function addresses are always even, according to
1776 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1777 @code{ptrmemfunc_vbit_in_pfn}.
1780 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1781 Normally, the C++ compiler uses function pointers in vtables. This
1782 macro allows the target to change to use ``function descriptors''
1783 instead. Function descriptors are found on targets for whom a
1784 function pointer is actually a small data structure. Normally the
1785 data structure consists of the actual code address plus a data
1786 pointer to which the function's data is relative.
1788 If vtables are used, the value of this macro should be the number
1789 of words that the function descriptor occupies.
1792 @defmac TARGET_VTABLE_ENTRY_ALIGN
1793 By default, the vtable entries are void pointers, the so the alignment
1794 is the same as pointer alignment. The value of this macro specifies
1795 the alignment of the vtable entry in bits. It should be defined only
1796 when special alignment is necessary. */
1799 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1800 There are a few non-descriptor entries in the vtable at offsets below
1801 zero. If these entries must be padded (say, to preserve the alignment
1802 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1803 of words in each data entry.
1807 @section Register Usage
1808 @cindex register usage
1810 This section explains how to describe what registers the target machine
1811 has, and how (in general) they can be used.
1813 The description of which registers a specific instruction can use is
1814 done with register classes; see @ref{Register Classes}. For information
1815 on using registers to access a stack frame, see @ref{Frame Registers}.
1816 For passing values in registers, see @ref{Register Arguments}.
1817 For returning values in registers, see @ref{Scalar Return}.
1820 * Register Basics:: Number and kinds of registers.
1821 * Allocation Order:: Order in which registers are allocated.
1822 * Values in Registers:: What kinds of values each reg can hold.
1823 * Leaf Functions:: Renumbering registers for leaf functions.
1824 * Stack Registers:: Handling a register stack such as 80387.
1827 @node Register Basics
1828 @subsection Basic Characteristics of Registers
1830 @c prevent bad page break with this line
1831 Registers have various characteristics.
1833 @defmac FIRST_PSEUDO_REGISTER
1834 Number of hardware registers known to the compiler. They receive
1835 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1836 pseudo register's number really is assigned the number
1837 @code{FIRST_PSEUDO_REGISTER}.
1840 @defmac FIXED_REGISTERS
1841 @cindex fixed register
1842 An initializer that says which registers are used for fixed purposes
1843 all throughout the compiled code and are therefore not available for
1844 general allocation. These would include the stack pointer, the frame
1845 pointer (except on machines where that can be used as a general
1846 register when no frame pointer is needed), the program counter on
1847 machines where that is considered one of the addressable registers,
1848 and any other numbered register with a standard use.
1850 This information is expressed as a sequence of numbers, separated by
1851 commas and surrounded by braces. The @var{n}th number is 1 if
1852 register @var{n} is fixed, 0 otherwise.
1854 The table initialized from this macro, and the table initialized by
1855 the following one, may be overridden at run time either automatically,
1856 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1857 the user with the command options @option{-ffixed-@var{reg}},
1858 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1861 @defmac CALL_USED_REGISTERS
1862 @cindex call-used register
1863 @cindex call-clobbered register
1864 @cindex call-saved register
1865 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1866 clobbered (in general) by function calls as well as for fixed
1867 registers. This macro therefore identifies the registers that are not
1868 available for general allocation of values that must live across
1871 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1872 automatically saves it on function entry and restores it on function
1873 exit, if the register is used within the function.
1876 @defmac CALL_REALLY_USED_REGISTERS
1877 @cindex call-used register
1878 @cindex call-clobbered register
1879 @cindex call-saved register
1880 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1881 that the entire set of @code{FIXED_REGISTERS} be included.
1882 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1883 This macro is optional. If not specified, it defaults to the value
1884 of @code{CALL_USED_REGISTERS}.
1887 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1888 @cindex call-used register
1889 @cindex call-clobbered register
1890 @cindex call-saved register
1891 A C expression that is nonzero if it is not permissible to store a
1892 value of mode @var{mode} in hard register number @var{regno} across a
1893 call without some part of it being clobbered. For most machines this
1894 macro need not be defined. It is only required for machines that do not
1895 preserve the entire contents of a register across a call.
1899 @findex call_used_regs
1902 @findex reg_class_contents
1903 @hook TARGET_CONDITIONAL_REGISTER_USAGE
1904 This hook may conditionally modify five variables
1905 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1906 @code{reg_names}, and @code{reg_class_contents}, to take into account
1907 any dependence of these register sets on target flags. The first three
1908 of these are of type @code{char []} (interpreted as Boolean vectors).
1909 @code{global_regs} is a @code{const char *[]}, and
1910 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1911 called, @code{fixed_regs}, @code{call_used_regs},
1912 @code{reg_class_contents}, and @code{reg_names} have been initialized
1913 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1914 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1915 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1916 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1917 command options have been applied.
1919 @cindex disabling certain registers
1920 @cindex controlling register usage
1921 If the usage of an entire class of registers depends on the target
1922 flags, you may indicate this to GCC by using this macro to modify
1923 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1924 registers in the classes which should not be used by GCC@. Also define
1925 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1926 to return @code{NO_REGS} if it
1927 is called with a letter for a class that shouldn't be used.
1929 (However, if this class is not included in @code{GENERAL_REGS} and all
1930 of the insn patterns whose constraints permit this class are
1931 controlled by target switches, then GCC will automatically avoid using
1932 these registers when the target switches are opposed to them.)
1935 @defmac INCOMING_REGNO (@var{out})
1936 Define this macro if the target machine has register windows. This C
1937 expression returns the register number as seen by the called function
1938 corresponding to the register number @var{out} as seen by the calling
1939 function. Return @var{out} if register number @var{out} is not an
1943 @defmac OUTGOING_REGNO (@var{in})
1944 Define this macro if the target machine has register windows. This C
1945 expression returns the register number as seen by the calling function
1946 corresponding to the register number @var{in} as seen by the called
1947 function. Return @var{in} if register number @var{in} is not an inbound
1951 @defmac LOCAL_REGNO (@var{regno})
1952 Define this macro if the target machine has register windows. This C
1953 expression returns true if the register is call-saved but is in the
1954 register window. Unlike most call-saved registers, such registers
1955 need not be explicitly restored on function exit or during non-local
1960 If the program counter has a register number, define this as that
1961 register number. Otherwise, do not define it.
1964 @node Allocation Order
1965 @subsection Order of Allocation of Registers
1966 @cindex order of register allocation
1967 @cindex register allocation order
1969 @c prevent bad page break with this line
1970 Registers are allocated in order.
1972 @defmac REG_ALLOC_ORDER
1973 If defined, an initializer for a vector of integers, containing the
1974 numbers of hard registers in the order in which GCC should prefer
1975 to use them (from most preferred to least).
1977 If this macro is not defined, registers are used lowest numbered first
1978 (all else being equal).
1980 One use of this macro is on machines where the highest numbered
1981 registers must always be saved and the save-multiple-registers
1982 instruction supports only sequences of consecutive registers. On such
1983 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1984 the highest numbered allocable register first.
1987 @defmac ADJUST_REG_ALLOC_ORDER
1988 A C statement (sans semicolon) to choose the order in which to allocate
1989 hard registers for pseudo-registers local to a basic block.
1991 Store the desired register order in the array @code{reg_alloc_order}.
1992 Element 0 should be the register to allocate first; element 1, the next
1993 register; and so on.
1995 The macro body should not assume anything about the contents of
1996 @code{reg_alloc_order} before execution of the macro.
1998 On most machines, it is not necessary to define this macro.
2001 @defmac HONOR_REG_ALLOC_ORDER
2002 Normally, IRA tries to estimate the costs for saving a register in the
2003 prologue and restoring it in the epilogue. This discourages it from
2004 using call-saved registers. If a machine wants to ensure that IRA
2005 allocates registers in the order given by REG_ALLOC_ORDER even if some
2006 call-saved registers appear earlier than call-used ones, this macro
2010 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2011 In some case register allocation order is not enough for the
2012 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2013 If this macro is defined, it should return a floating point value
2014 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2015 be increased by approximately the pseudo's usage frequency times the
2016 value returned by this macro. Not defining this macro is equivalent
2017 to having it always return @code{0.0}.
2019 On most machines, it is not necessary to define this macro.
2022 @node Values in Registers
2023 @subsection How Values Fit in Registers
2025 This section discusses the macros that describe which kinds of values
2026 (specifically, which machine modes) each register can hold, and how many
2027 consecutive registers are needed for a given mode.
2029 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2030 A C expression for the number of consecutive hard registers, starting
2031 at register number @var{regno}, required to hold a value of mode
2032 @var{mode}. This macro must never return zero, even if a register
2033 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2034 and/or CANNOT_CHANGE_MODE_CLASS instead.
2036 On a machine where all registers are exactly one word, a suitable
2037 definition of this macro is
2040 #define HARD_REGNO_NREGS(REGNO, MODE) \
2041 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2046 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2047 A C expression that is nonzero if a value of mode @var{mode}, stored
2048 in memory, ends with padding that causes it to take up more space than
2049 in registers starting at register number @var{regno} (as determined by
2050 multiplying GCC's notion of the size of the register when containing
2051 this mode by the number of registers returned by
2052 @code{HARD_REGNO_NREGS}). By default this is zero.
2054 For example, if a floating-point value is stored in three 32-bit
2055 registers but takes up 128 bits in memory, then this would be
2058 This macros only needs to be defined if there are cases where
2059 @code{subreg_get_info}
2060 would otherwise wrongly determine that a @code{subreg} can be
2061 represented by an offset to the register number, when in fact such a
2062 @code{subreg} would contain some of the padding not stored in
2063 registers and so not be representable.
2066 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2067 For values of @var{regno} and @var{mode} for which
2068 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2069 returning the greater number of registers required to hold the value
2070 including any padding. In the example above, the value would be four.
2073 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2074 Define this macro if the natural size of registers that hold values
2075 of mode @var{mode} is not the word size. It is a C expression that
2076 should give the natural size in bytes for the specified mode. It is
2077 used by the register allocator to try to optimize its results. This
2078 happens for example on SPARC 64-bit where the natural size of
2079 floating-point registers is still 32-bit.
2082 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2083 A C expression that is nonzero if it is permissible to store a value
2084 of mode @var{mode} in hard register number @var{regno} (or in several
2085 registers starting with that one). For a machine where all registers
2086 are equivalent, a suitable definition is
2089 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2092 You need not include code to check for the numbers of fixed registers,
2093 because the allocation mechanism considers them to be always occupied.
2095 @cindex register pairs
2096 On some machines, double-precision values must be kept in even/odd
2097 register pairs. You can implement that by defining this macro to reject
2098 odd register numbers for such modes.
2100 The minimum requirement for a mode to be OK in a register is that the
2101 @samp{mov@var{mode}} instruction pattern support moves between the
2102 register and other hard register in the same class and that moving a
2103 value into the register and back out not alter it.
2105 Since the same instruction used to move @code{word_mode} will work for
2106 all narrower integer modes, it is not necessary on any machine for
2107 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2108 you define patterns @samp{movhi}, etc., to take advantage of this. This
2109 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2110 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2113 Many machines have special registers for floating point arithmetic.
2114 Often people assume that floating point machine modes are allowed only
2115 in floating point registers. This is not true. Any registers that
2116 can hold integers can safely @emph{hold} a floating point machine
2117 mode, whether or not floating arithmetic can be done on it in those
2118 registers. Integer move instructions can be used to move the values.
2120 On some machines, though, the converse is true: fixed-point machine
2121 modes may not go in floating registers. This is true if the floating
2122 registers normalize any value stored in them, because storing a
2123 non-floating value there would garble it. In this case,
2124 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2125 floating registers. But if the floating registers do not automatically
2126 normalize, if you can store any bit pattern in one and retrieve it
2127 unchanged without a trap, then any machine mode may go in a floating
2128 register, so you can define this macro to say so.
2130 The primary significance of special floating registers is rather that
2131 they are the registers acceptable in floating point arithmetic
2132 instructions. However, this is of no concern to
2133 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2134 constraints for those instructions.
2136 On some machines, the floating registers are especially slow to access,
2137 so that it is better to store a value in a stack frame than in such a
2138 register if floating point arithmetic is not being done. As long as the
2139 floating registers are not in class @code{GENERAL_REGS}, they will not
2140 be used unless some pattern's constraint asks for one.
2143 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2144 A C expression that is nonzero if it is OK to rename a hard register
2145 @var{from} to another hard register @var{to}.
2147 One common use of this macro is to prevent renaming of a register to
2148 another register that is not saved by a prologue in an interrupt
2151 The default is always nonzero.
2154 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2155 A C expression that is nonzero if a value of mode
2156 @var{mode1} is accessible in mode @var{mode2} without copying.
2158 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2159 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2160 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2161 should be nonzero. If they differ for any @var{r}, you should define
2162 this macro to return zero unless some other mechanism ensures the
2163 accessibility of the value in a narrower mode.
2165 You should define this macro to return nonzero in as many cases as
2166 possible since doing so will allow GCC to perform better register
2170 @hook TARGET_HARD_REGNO_SCRATCH_OK
2171 This target hook should return @code{true} if it is OK to use a hard register
2172 @var{regno} as scratch reg in peephole2.
2174 One common use of this macro is to prevent using of a register that
2175 is not saved by a prologue in an interrupt handler.
2177 The default version of this hook always returns @code{true}.
2180 @defmac AVOID_CCMODE_COPIES
2181 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2182 registers. You should only define this macro if support for copying to/from
2183 @code{CCmode} is incomplete.
2186 @node Leaf Functions
2187 @subsection Handling Leaf Functions
2189 @cindex leaf functions
2190 @cindex functions, leaf
2191 On some machines, a leaf function (i.e., one which makes no calls) can run
2192 more efficiently if it does not make its own register window. Often this
2193 means it is required to receive its arguments in the registers where they
2194 are passed by the caller, instead of the registers where they would
2197 The special treatment for leaf functions generally applies only when
2198 other conditions are met; for example, often they may use only those
2199 registers for its own variables and temporaries. We use the term ``leaf
2200 function'' to mean a function that is suitable for this special
2201 handling, so that functions with no calls are not necessarily ``leaf
2204 GCC assigns register numbers before it knows whether the function is
2205 suitable for leaf function treatment. So it needs to renumber the
2206 registers in order to output a leaf function. The following macros
2209 @defmac LEAF_REGISTERS
2210 Name of a char vector, indexed by hard register number, which
2211 contains 1 for a register that is allowable in a candidate for leaf
2214 If leaf function treatment involves renumbering the registers, then the
2215 registers marked here should be the ones before renumbering---those that
2216 GCC would ordinarily allocate. The registers which will actually be
2217 used in the assembler code, after renumbering, should not be marked with 1
2220 Define this macro only if the target machine offers a way to optimize
2221 the treatment of leaf functions.
2224 @defmac LEAF_REG_REMAP (@var{regno})
2225 A C expression whose value is the register number to which @var{regno}
2226 should be renumbered, when a function is treated as a leaf function.
2228 If @var{regno} is a register number which should not appear in a leaf
2229 function before renumbering, then the expression should yield @minus{}1, which
2230 will cause the compiler to abort.
2232 Define this macro only if the target machine offers a way to optimize the
2233 treatment of leaf functions, and registers need to be renumbered to do
2237 @findex current_function_is_leaf
2238 @findex current_function_uses_only_leaf_regs
2239 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2240 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2241 specially. They can test the C variable @code{current_function_is_leaf}
2242 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2243 set prior to local register allocation and is valid for the remaining
2244 compiler passes. They can also test the C variable
2245 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2246 functions which only use leaf registers.
2247 @code{current_function_uses_only_leaf_regs} is valid after all passes
2248 that modify the instructions have been run and is only useful if
2249 @code{LEAF_REGISTERS} is defined.
2250 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2251 @c of the next paragraph?! --mew 2feb93
2253 @node Stack Registers
2254 @subsection Registers That Form a Stack
2256 There are special features to handle computers where some of the
2257 ``registers'' form a stack. Stack registers are normally written by
2258 pushing onto the stack, and are numbered relative to the top of the
2261 Currently, GCC can only handle one group of stack-like registers, and
2262 they must be consecutively numbered. Furthermore, the existing
2263 support for stack-like registers is specific to the 80387 floating
2264 point coprocessor. If you have a new architecture that uses
2265 stack-like registers, you will need to do substantial work on
2266 @file{reg-stack.c} and write your machine description to cooperate
2267 with it, as well as defining these macros.
2270 Define this if the machine has any stack-like registers.
2273 @defmac STACK_REG_COVER_CLASS
2274 This is a cover class containing the stack registers. Define this if
2275 the machine has any stack-like registers.
2278 @defmac FIRST_STACK_REG
2279 The number of the first stack-like register. This one is the top
2283 @defmac LAST_STACK_REG
2284 The number of the last stack-like register. This one is the bottom of
2288 @node Register Classes
2289 @section Register Classes
2290 @cindex register class definitions
2291 @cindex class definitions, register
2293 On many machines, the numbered registers are not all equivalent.
2294 For example, certain registers may not be allowed for indexed addressing;
2295 certain registers may not be allowed in some instructions. These machine
2296 restrictions are described to the compiler using @dfn{register classes}.
2298 You define a number of register classes, giving each one a name and saying
2299 which of the registers belong to it. Then you can specify register classes
2300 that are allowed as operands to particular instruction patterns.
2304 In general, each register will belong to several classes. In fact, one
2305 class must be named @code{ALL_REGS} and contain all the registers. Another
2306 class must be named @code{NO_REGS} and contain no registers. Often the
2307 union of two classes will be another class; however, this is not required.
2309 @findex GENERAL_REGS
2310 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2311 terribly special about the name, but the operand constraint letters
2312 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2313 the same as @code{ALL_REGS}, just define it as a macro which expands
2316 Order the classes so that if class @var{x} is contained in class @var{y}
2317 then @var{x} has a lower class number than @var{y}.
2319 The way classes other than @code{GENERAL_REGS} are specified in operand
2320 constraints is through machine-dependent operand constraint letters.
2321 You can define such letters to correspond to various classes, then use
2322 them in operand constraints.
2324 You must define the narrowest register classes for allocatable
2325 registers, so that each class either has no subclasses, or that for
2326 some mode, the move cost between registers within the class is
2327 cheaper than moving a register in the class to or from memory
2330 You should define a class for the union of two classes whenever some
2331 instruction allows both classes. For example, if an instruction allows
2332 either a floating point (coprocessor) register or a general register for a
2333 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2334 which includes both of them. Otherwise you will get suboptimal code,
2335 or even internal compiler errors when reload cannot find a register in the
2336 class computed via @code{reg_class_subunion}.
2338 You must also specify certain redundant information about the register
2339 classes: for each class, which classes contain it and which ones are
2340 contained in it; for each pair of classes, the largest class contained
2343 When a value occupying several consecutive registers is expected in a
2344 certain class, all the registers used must belong to that class.
2345 Therefore, register classes cannot be used to enforce a requirement for
2346 a register pair to start with an even-numbered register. The way to
2347 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2349 Register classes used for input-operands of bitwise-and or shift
2350 instructions have a special requirement: each such class must have, for
2351 each fixed-point machine mode, a subclass whose registers can transfer that
2352 mode to or from memory. For example, on some machines, the operations for
2353 single-byte values (@code{QImode}) are limited to certain registers. When
2354 this is so, each register class that is used in a bitwise-and or shift
2355 instruction must have a subclass consisting of registers from which
2356 single-byte values can be loaded or stored. This is so that
2357 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2359 @deftp {Data type} {enum reg_class}
2360 An enumerated type that must be defined with all the register class names
2361 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2362 must be the last register class, followed by one more enumerated value,
2363 @code{LIM_REG_CLASSES}, which is not a register class but rather
2364 tells how many classes there are.
2366 Each register class has a number, which is the value of casting
2367 the class name to type @code{int}. The number serves as an index
2368 in many of the tables described below.
2371 @defmac N_REG_CLASSES
2372 The number of distinct register classes, defined as follows:
2375 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2379 @defmac REG_CLASS_NAMES
2380 An initializer containing the names of the register classes as C string
2381 constants. These names are used in writing some of the debugging dumps.
2384 @defmac REG_CLASS_CONTENTS
2385 An initializer containing the contents of the register classes, as integers
2386 which are bit masks. The @var{n}th integer specifies the contents of class
2387 @var{n}. The way the integer @var{mask} is interpreted is that
2388 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2390 When the machine has more than 32 registers, an integer does not suffice.
2391 Then the integers are replaced by sub-initializers, braced groupings containing
2392 several integers. Each sub-initializer must be suitable as an initializer
2393 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2394 In this situation, the first integer in each sub-initializer corresponds to
2395 registers 0 through 31, the second integer to registers 32 through 63, and
2399 @defmac REGNO_REG_CLASS (@var{regno})
2400 A C expression whose value is a register class containing hard register
2401 @var{regno}. In general there is more than one such class; choose a class
2402 which is @dfn{minimal}, meaning that no smaller class also contains the
2406 @defmac BASE_REG_CLASS
2407 A macro whose definition is the name of the class to which a valid
2408 base register must belong. A base register is one used in an address
2409 which is the register value plus a displacement.
2412 @defmac MODE_BASE_REG_CLASS (@var{mode})
2413 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2414 the selection of a base register in a mode dependent manner. If
2415 @var{mode} is VOIDmode then it should return the same value as
2416 @code{BASE_REG_CLASS}.
2419 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2420 A C expression whose value is the register class to which a valid
2421 base register must belong in order to be used in a base plus index
2422 register address. You should define this macro if base plus index
2423 addresses have different requirements than other base register uses.
2426 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2427 A C expression whose value is the register class to which a valid
2428 base register for a memory reference in mode @var{mode} to address
2429 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2430 define the context in which the base register occurs. @var{outer_code} is
2431 the code of the immediately enclosing expression (@code{MEM} for the top level
2432 of an address, @code{ADDRESS} for something that occurs in an
2433 @code{address_operand}). @var{index_code} is the code of the corresponding
2434 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2437 @defmac INDEX_REG_CLASS
2438 A macro whose definition is the name of the class to which a valid
2439 index register must belong. An index register is one used in an
2440 address where its value is either multiplied by a scale factor or
2441 added to another register (as well as added to a displacement).
2444 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2445 A C expression which is nonzero if register number @var{num} is
2446 suitable for use as a base register in operand addresses.
2449 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2450 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2451 that expression may examine the mode of the memory reference in
2452 @var{mode}. You should define this macro if the mode of the memory
2453 reference affects whether a register may be used as a base register. If
2454 you define this macro, the compiler will use it instead of
2455 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2456 addresses that appear outside a @code{MEM}, i.e., as an
2457 @code{address_operand}.
2460 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2461 A C expression which is nonzero if register number @var{num} is suitable for
2462 use as a base register in base plus index operand addresses, accessing
2463 memory in mode @var{mode}. It may be either a suitable hard register or a
2464 pseudo register that has been allocated such a hard register. You should
2465 define this macro if base plus index addresses have different requirements
2466 than other base register uses.
2468 Use of this macro is deprecated; please use the more general
2469 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2472 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2473 A C expression which is nonzero if register number @var{num} is
2474 suitable for use as a base register in operand addresses, accessing
2475 memory in mode @var{mode} in address space @var{address_space}.
2476 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2477 that that expression may examine the context in which the register
2478 appears in the memory reference. @var{outer_code} is the code of the
2479 immediately enclosing expression (@code{MEM} if at the top level of the
2480 address, @code{ADDRESS} for something that occurs in an
2481 @code{address_operand}). @var{index_code} is the code of the
2482 corresponding index expression if @var{outer_code} is @code{PLUS};
2483 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2484 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2487 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2488 A C expression which is nonzero if register number @var{num} is
2489 suitable for use as an index register in operand addresses. It may be
2490 either a suitable hard register or a pseudo register that has been
2491 allocated such a hard register.
2493 The difference between an index register and a base register is that
2494 the index register may be scaled. If an address involves the sum of
2495 two registers, neither one of them scaled, then either one may be
2496 labeled the ``base'' and the other the ``index''; but whichever
2497 labeling is used must fit the machine's constraints of which registers
2498 may serve in each capacity. The compiler will try both labelings,
2499 looking for one that is valid, and will reload one or both registers
2500 only if neither labeling works.
2503 @hook TARGET_PREFERRED_RENAME_CLASS
2505 @hook TARGET_PREFERRED_RELOAD_CLASS
2506 A target hook that places additional restrictions on the register class
2507 to use when it is necessary to copy value @var{x} into a register in class
2508 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2509 another, smaller class.
2511 The default version of this hook always returns value of @code{rclass} argument.
2513 Sometimes returning a more restrictive class makes better code. For
2514 example, on the 68000, when @var{x} is an integer constant that is in range
2515 for a @samp{moveq} instruction, the value of this macro is always
2516 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2517 Requiring a data register guarantees that a @samp{moveq} will be used.
2519 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2520 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2521 loaded into some register class. By returning @code{NO_REGS} you can
2522 force @var{x} into a memory location. For example, rs6000 can load
2523 immediate values into general-purpose registers, but does not have an
2524 instruction for loading an immediate value into a floating-point
2525 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2526 @var{x} is a floating-point constant. If the constant can't be loaded
2527 into any kind of register, code generation will be better if
2528 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2529 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2531 If an insn has pseudos in it after register allocation, reload will go
2532 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2533 to find the best one. Returning @code{NO_REGS}, in this case, makes
2534 reload add a @code{!} in front of the constraint: the x86 back-end uses
2535 this feature to discourage usage of 387 registers when math is done in
2536 the SSE registers (and vice versa).
2539 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2540 A C expression that places additional restrictions on the register class
2541 to use when it is necessary to copy value @var{x} into a register in class
2542 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2543 another, smaller class. On many machines, the following definition is
2547 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2550 Sometimes returning a more restrictive class makes better code. For
2551 example, on the 68000, when @var{x} is an integer constant that is in range
2552 for a @samp{moveq} instruction, the value of this macro is always
2553 @code{DATA_REGS} as long as @var{class} includes the data registers.
2554 Requiring a data register guarantees that a @samp{moveq} will be used.
2556 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2557 @var{class} is if @var{x} is a legitimate constant which cannot be
2558 loaded into some register class. By returning @code{NO_REGS} you can
2559 force @var{x} into a memory location. For example, rs6000 can load
2560 immediate values into general-purpose registers, but does not have an
2561 instruction for loading an immediate value into a floating-point
2562 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2563 @var{x} is a floating-point constant. If the constant can't be loaded
2564 into any kind of register, code generation will be better if
2565 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2566 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2568 If an insn has pseudos in it after register allocation, reload will go
2569 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2570 to find the best one. Returning @code{NO_REGS}, in this case, makes
2571 reload add a @code{!} in front of the constraint: the x86 back-end uses
2572 this feature to discourage usage of 387 registers when math is done in
2573 the SSE registers (and vice versa).
2576 @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2577 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2580 The default version of this hook always returns value of @code{rclass}
2583 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2584 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2587 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2588 A C expression that places additional restrictions on the register class
2589 to use when it is necessary to be able to hold a value of mode
2590 @var{mode} in a reload register for which class @var{class} would
2593 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2594 there are certain modes that simply can't go in certain reload classes.
2596 The value is a register class; perhaps @var{class}, or perhaps another,
2599 Don't define this macro unless the target machine has limitations which
2600 require the macro to do something nontrivial.
2603 @hook TARGET_SECONDARY_RELOAD
2604 Many machines have some registers that cannot be copied directly to or
2605 from memory or even from other types of registers. An example is the
2606 @samp{MQ} register, which on most machines, can only be copied to or
2607 from general registers, but not memory. Below, we shall be using the
2608 term 'intermediate register' when a move operation cannot be performed
2609 directly, but has to be done by copying the source into the intermediate
2610 register first, and then copying the intermediate register to the
2611 destination. An intermediate register always has the same mode as
2612 source and destination. Since it holds the actual value being copied,
2613 reload might apply optimizations to re-use an intermediate register
2614 and eliding the copy from the source when it can determine that the
2615 intermediate register still holds the required value.
2617 Another kind of secondary reload is required on some machines which
2618 allow copying all registers to and from memory, but require a scratch
2619 register for stores to some memory locations (e.g., those with symbolic
2620 address on the RT, and those with certain symbolic address on the SPARC
2621 when compiling PIC)@. Scratch registers need not have the same mode
2622 as the value being copied, and usually hold a different value than
2623 that being copied. Special patterns in the md file are needed to
2624 describe how the copy is performed with the help of the scratch register;
2625 these patterns also describe the number, register class(es) and mode(s)
2626 of the scratch register(s).
2628 In some cases, both an intermediate and a scratch register are required.
2630 For input reloads, this target hook is called with nonzero @var{in_p},
2631 and @var{x} is an rtx that needs to be copied to a register of class
2632 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2633 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2634 needs to be copied to rtx @var{x} in @var{reload_mode}.
2636 If copying a register of @var{reload_class} from/to @var{x} requires
2637 an intermediate register, the hook @code{secondary_reload} should
2638 return the register class required for this intermediate register.
2639 If no intermediate register is required, it should return NO_REGS.
2640 If more than one intermediate register is required, describe the one
2641 that is closest in the copy chain to the reload register.
2643 If scratch registers are needed, you also have to describe how to
2644 perform the copy from/to the reload register to/from this
2645 closest intermediate register. Or if no intermediate register is
2646 required, but still a scratch register is needed, describe the
2647 copy from/to the reload register to/from the reload operand @var{x}.
2649 You do this by setting @code{sri->icode} to the instruction code of a pattern
2650 in the md file which performs the move. Operands 0 and 1 are the output
2651 and input of this copy, respectively. Operands from operand 2 onward are
2652 for scratch operands. These scratch operands must have a mode, and a
2653 single-register-class
2654 @c [later: or memory]
2657 When an intermediate register is used, the @code{secondary_reload}
2658 hook will be called again to determine how to copy the intermediate
2659 register to/from the reload operand @var{x}, so your hook must also
2660 have code to handle the register class of the intermediate operand.
2662 @c [For later: maybe we'll allow multi-alternative reload patterns -
2663 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2664 @c and match the constraints of input and output to determine the required
2665 @c alternative. A restriction would be that constraints used to match
2666 @c against reloads registers would have to be written as register class
2667 @c constraints, or we need a new target macro / hook that tells us if an
2668 @c arbitrary constraint can match an unknown register of a given class.
2669 @c Such a macro / hook would also be useful in other places.]
2672 @var{x} might be a pseudo-register or a @code{subreg} of a
2673 pseudo-register, which could either be in a hard register or in memory.
2674 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2675 in memory and the hard register number if it is in a register.
2677 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2678 currently not supported. For the time being, you will have to continue
2679 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2681 @code{copy_cost} also uses this target hook to find out how values are
2682 copied. If you want it to include some extra cost for the need to allocate
2683 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2684 Or if two dependent moves are supposed to have a lower cost than the sum
2685 of the individual moves due to expected fortuitous scheduling and/or special
2686 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2689 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2690 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2691 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2692 These macros are obsolete, new ports should use the target hook
2693 @code{TARGET_SECONDARY_RELOAD} instead.
2695 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2696 target hook. Older ports still define these macros to indicate to the
2697 reload phase that it may
2698 need to allocate at least one register for a reload in addition to the
2699 register to contain the data. Specifically, if copying @var{x} to a
2700 register @var{class} in @var{mode} requires an intermediate register,
2701 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2702 largest register class all of whose registers can be used as
2703 intermediate registers or scratch registers.
2705 If copying a register @var{class} in @var{mode} to @var{x} requires an
2706 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2707 was supposed to be defined be defined to return the largest register
2708 class required. If the
2709 requirements for input and output reloads were the same, the macro
2710 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2713 The values returned by these macros are often @code{GENERAL_REGS}.
2714 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2715 can be directly copied to or from a register of @var{class} in
2716 @var{mode} without requiring a scratch register. Do not define this
2717 macro if it would always return @code{NO_REGS}.
2719 If a scratch register is required (either with or without an
2720 intermediate register), you were supposed to define patterns for
2721 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2722 (@pxref{Standard Names}. These patterns, which were normally
2723 implemented with a @code{define_expand}, should be similar to the
2724 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2727 These patterns need constraints for the reload register and scratch
2729 contain a single register class. If the original reload register (whose
2730 class is @var{class}) can meet the constraint given in the pattern, the
2731 value returned by these macros is used for the class of the scratch
2732 register. Otherwise, two additional reload registers are required.
2733 Their classes are obtained from the constraints in the insn pattern.
2735 @var{x} might be a pseudo-register or a @code{subreg} of a
2736 pseudo-register, which could either be in a hard register or in memory.
2737 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2738 in memory and the hard register number if it is in a register.
2740 These macros should not be used in the case where a particular class of
2741 registers can only be copied to memory and not to another class of
2742 registers. In that case, secondary reload registers are not needed and
2743 would not be helpful. Instead, a stack location must be used to perform
2744 the copy and the @code{mov@var{m}} pattern should use memory as an
2745 intermediate storage. This case often occurs between floating-point and
2749 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2750 Certain machines have the property that some registers cannot be copied
2751 to some other registers without using memory. Define this macro on
2752 those machines to be a C expression that is nonzero if objects of mode
2753 @var{m} in registers of @var{class1} can only be copied to registers of
2754 class @var{class2} by storing a register of @var{class1} into memory
2755 and loading that memory location into a register of @var{class2}.
2757 Do not define this macro if its value would always be zero.
2760 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2761 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2762 allocates a stack slot for a memory location needed for register copies.
2763 If this macro is defined, the compiler instead uses the memory location
2764 defined by this macro.
2766 Do not define this macro if you do not define
2767 @code{SECONDARY_MEMORY_NEEDED}.
2770 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2771 When the compiler needs a secondary memory location to copy between two
2772 registers of mode @var{mode}, it normally allocates sufficient memory to
2773 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2774 load operations in a mode that many bits wide and whose class is the
2775 same as that of @var{mode}.
2777 This is right thing to do on most machines because it ensures that all
2778 bits of the register are copied and prevents accesses to the registers
2779 in a narrower mode, which some machines prohibit for floating-point
2782 However, this default behavior is not correct on some machines, such as
2783 the DEC Alpha, that store short integers in floating-point registers
2784 differently than in integer registers. On those machines, the default
2785 widening will not work correctly and you must define this macro to
2786 suppress that widening in some cases. See the file @file{alpha.h} for
2789 Do not define this macro if you do not define
2790 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2791 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2794 @hook TARGET_CLASS_LIKELY_SPILLED_P
2795 A target hook which returns @code{true} if pseudos that have been assigned
2796 to registers of class @var{rclass} would likely be spilled because
2797 registers of @var{rclass} are needed for spill registers.
2799 The default version of this target hook returns @code{true} if @var{rclass}
2800 has exactly one register and @code{false} otherwise. On most machines, this
2801 default should be used. Only use this target hook to some other expression
2802 if pseudos allocated by @file{local-alloc.c} end up in memory because their
2803 hard registers were needed for spill registers. If this target hook returns
2804 @code{false} for those classes, those pseudos will only be allocated by
2805 @file{global.c}, which knows how to reallocate the pseudo to another
2806 register. If there would not be another register available for reallocation,
2807 you should not change the implementation of this target hook since
2808 the only effect of such implementation would be to slow down register
2812 @hook TARGET_CLASS_MAX_NREGS
2813 A target hook returns the maximum number of consecutive registers
2814 of class @var{rclass} needed to hold a value of mode @var{mode}.
2816 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2817 the value returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass},
2818 @var{mode})} target hook should be the maximum value of
2819 @code{HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno}
2820 values in the class @var{rclass}.
2822 This target hook helps control the handling of multiple-word values
2825 The default version of this target hook returns the size of @var{mode}
2829 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2830 A C expression for the maximum number of consecutive registers
2831 of class @var{class} needed to hold a value of mode @var{mode}.
2833 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2834 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2835 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2836 @var{mode})} for all @var{regno} values in the class @var{class}.
2838 This macro helps control the handling of multiple-word values
2842 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2843 If defined, a C expression that returns nonzero for a @var{class} for which
2844 a change from mode @var{from} to mode @var{to} is invalid.
2846 For the example, loading 32-bit integer or floating-point objects into
2847 floating-point registers on the Alpha extends them to 64 bits.
2848 Therefore loading a 64-bit object and then storing it as a 32-bit object
2849 does not store the low-order 32 bits, as would be the case for a normal
2850 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2854 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2855 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2856 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2860 @node Old Constraints
2861 @section Obsolete Macros for Defining Constraints
2862 @cindex defining constraints, obsolete method
2863 @cindex constraints, defining, obsolete method
2865 Machine-specific constraints can be defined with these macros instead
2866 of the machine description constructs described in @ref{Define
2867 Constraints}. This mechanism is obsolete. New ports should not use
2868 it; old ports should convert to the new mechanism.
2870 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2871 For the constraint at the start of @var{str}, which starts with the letter
2872 @var{c}, return the length. This allows you to have register class /
2873 constant / extra constraints that are longer than a single letter;
2874 you don't need to define this macro if you can do with single-letter
2875 constraints only. The definition of this macro should use
2876 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2877 to handle specially.
2878 There are some sanity checks in genoutput.c that check the constraint lengths
2879 for the md file, so you can also use this macro to help you while you are
2880 transitioning from a byzantine single-letter-constraint scheme: when you
2881 return a negative length for a constraint you want to re-use, genoutput
2882 will complain about every instance where it is used in the md file.
2885 @defmac REG_CLASS_FROM_LETTER (@var{char})
2886 A C expression which defines the machine-dependent operand constraint
2887 letters for register classes. If @var{char} is such a letter, the
2888 value should be the register class corresponding to it. Otherwise,
2889 the value should be @code{NO_REGS}. The register letter @samp{r},
2890 corresponding to class @code{GENERAL_REGS}, will not be passed
2891 to this macro; you do not need to handle it.
2894 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2895 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2896 passed in @var{str}, so that you can use suffixes to distinguish between
2900 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2901 A C expression that defines the machine-dependent operand constraint
2902 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2903 particular ranges of integer values. If @var{c} is one of those
2904 letters, the expression should check that @var{value}, an integer, is in
2905 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2906 not one of those letters, the value should be 0 regardless of
2910 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2911 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2912 string passed in @var{str}, so that you can use suffixes to distinguish
2913 between different variants.
2916 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2917 A C expression that defines the machine-dependent operand constraint
2918 letters that specify particular ranges of @code{const_double} values
2919 (@samp{G} or @samp{H}).
2921 If @var{c} is one of those letters, the expression should check that
2922 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2923 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2924 letters, the value should be 0 regardless of @var{value}.
2926 @code{const_double} is used for all floating-point constants and for
2927 @code{DImode} fixed-point constants. A given letter can accept either
2928 or both kinds of values. It can use @code{GET_MODE} to distinguish
2929 between these kinds.
2932 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2933 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2934 string passed in @var{str}, so that you can use suffixes to distinguish
2935 between different variants.
2938 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2939 A C expression that defines the optional machine-dependent constraint
2940 letters that can be used to segregate specific types of operands, usually
2941 memory references, for the target machine. Any letter that is not
2942 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2943 @code{REG_CLASS_FROM_CONSTRAINT}
2944 may be used. Normally this macro will not be defined.
2946 If it is required for a particular target machine, it should return 1
2947 if @var{value} corresponds to the operand type represented by the
2948 constraint letter @var{c}. If @var{c} is not defined as an extra
2949 constraint, the value returned should be 0 regardless of @var{value}.
2951 For example, on the ROMP, load instructions cannot have their output
2952 in r0 if the memory reference contains a symbolic address. Constraint
2953 letter @samp{Q} is defined as representing a memory address that does
2954 @emph{not} contain a symbolic address. An alternative is specified with
2955 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2956 alternative specifies @samp{m} on the input and a register class that
2957 does not include r0 on the output.
2960 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2961 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2962 in @var{str}, so that you can use suffixes to distinguish between different
2966 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2967 A C expression that defines the optional machine-dependent constraint
2968 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2969 be treated like memory constraints by the reload pass.
2971 It should return 1 if the operand type represented by the constraint
2972 at the start of @var{str}, the first letter of which is the letter @var{c},
2973 comprises a subset of all memory references including
2974 all those whose address is simply a base register. This allows the reload
2975 pass to reload an operand, if it does not directly correspond to the operand
2976 type of @var{c}, by copying its address into a base register.
2978 For example, on the S/390, some instructions do not accept arbitrary
2979 memory references, but only those that do not make use of an index
2980 register. The constraint letter @samp{Q} is defined via
2981 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2982 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2983 a @samp{Q} constraint can handle any memory operand, because the
2984 reload pass knows it can be reloaded by copying the memory address
2985 into a base register if required. This is analogous to the way
2986 an @samp{o} constraint can handle any memory operand.
2989 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2990 A C expression that defines the optional machine-dependent constraint
2991 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2992 @code{EXTRA_CONSTRAINT_STR}, that should
2993 be treated like address constraints by the reload pass.
2995 It should return 1 if the operand type represented by the constraint
2996 at the start of @var{str}, which starts with the letter @var{c}, comprises
2997 a subset of all memory addresses including
2998 all those that consist of just a base register. This allows the reload
2999 pass to reload an operand, if it does not directly correspond to the operand
3000 type of @var{str}, by copying it into a base register.
3002 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3003 be used with the @code{address_operand} predicate. It is treated
3004 analogously to the @samp{p} constraint.
3007 @node Stack and Calling
3008 @section Stack Layout and Calling Conventions
3009 @cindex calling conventions
3011 @c prevent bad page break with this line
3012 This describes the stack layout and calling conventions.
3016 * Exception Handling::
3021 * Register Arguments::
3023 * Aggregate Return::
3028 * Stack Smashing Protection::
3032 @subsection Basic Stack Layout
3033 @cindex stack frame layout
3034 @cindex frame layout
3036 @c prevent bad page break with this line
3037 Here is the basic stack layout.
3039 @defmac STACK_GROWS_DOWNWARD
3040 Define this macro if pushing a word onto the stack moves the stack
3041 pointer to a smaller address.
3043 When we say, ``define this macro if @dots{}'', it means that the
3044 compiler checks this macro only with @code{#ifdef} so the precise
3045 definition used does not matter.
3048 @defmac STACK_PUSH_CODE
3049 This macro defines the operation used when something is pushed
3050 on the stack. In RTL, a push operation will be
3051 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3053 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3054 and @code{POST_INC}. Which of these is correct depends on
3055 the stack direction and on whether the stack pointer points
3056 to the last item on the stack or whether it points to the
3057 space for the next item on the stack.
3059 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3060 defined, which is almost always right, and @code{PRE_INC} otherwise,
3061 which is often wrong.
3064 @defmac FRAME_GROWS_DOWNWARD
3065 Define this macro to nonzero value if the addresses of local variable slots
3066 are at negative offsets from the frame pointer.
3069 @defmac ARGS_GROW_DOWNWARD
3070 Define this macro if successive arguments to a function occupy decreasing
3071 addresses on the stack.
3074 @defmac STARTING_FRAME_OFFSET
3075 Offset from the frame pointer to the first local variable slot to be allocated.
3077 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3078 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3079 Otherwise, it is found by adding the length of the first slot to the
3080 value @code{STARTING_FRAME_OFFSET}.
3081 @c i'm not sure if the above is still correct.. had to change it to get
3082 @c rid of an overfull. --mew 2feb93
3085 @defmac STACK_ALIGNMENT_NEEDED
3086 Define to zero to disable final alignment of the stack during reload.
3087 The nonzero default for this macro is suitable for most ports.
3089 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3090 is a register save block following the local block that doesn't require
3091 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3092 stack alignment and do it in the backend.
3095 @defmac STACK_POINTER_OFFSET
3096 Offset from the stack pointer register to the first location at which
3097 outgoing arguments are placed. If not specified, the default value of
3098 zero is used. This is the proper value for most machines.
3100 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3101 the first location at which outgoing arguments are placed.
3104 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3105 Offset from the argument pointer register to the first argument's
3106 address. On some machines it may depend on the data type of the
3109 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3110 the first argument's address.
3113 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3114 Offset from the stack pointer register to an item dynamically allocated
3115 on the stack, e.g., by @code{alloca}.
3117 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3118 length of the outgoing arguments. The default is correct for most
3119 machines. See @file{function.c} for details.
3122 @defmac INITIAL_FRAME_ADDRESS_RTX
3123 A C expression whose value is RTL representing the address of the initial
3124 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3125 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3126 default value will be used. Define this macro in order to make frame pointer
3127 elimination work in the presence of @code{__builtin_frame_address (count)} and
3128 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3131 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3132 A C expression whose value is RTL representing the address in a stack
3133 frame where the pointer to the caller's frame is stored. Assume that
3134 @var{frameaddr} is an RTL expression for the address of the stack frame
3137 If you don't define this macro, the default is to return the value
3138 of @var{frameaddr}---that is, the stack frame address is also the
3139 address of the stack word that points to the previous frame.
3142 @defmac SETUP_FRAME_ADDRESSES
3143 If defined, a C expression that produces the machine-specific code to
3144 setup the stack so that arbitrary frames can be accessed. For example,
3145 on the SPARC, we must flush all of the register windows to the stack
3146 before we can access arbitrary stack frames. You will seldom need to
3150 @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
3151 This target hook should return an rtx that is used to store
3152 the address of the current frame into the built in @code{setjmp} buffer.
3153 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3154 machines. One reason you may need to define this target hook is if
3155 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3158 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3159 A C expression whose value is RTL representing the value of the frame
3160 address for the current frame. @var{frameaddr} is the frame pointer
3161 of the current frame. This is used for __builtin_frame_address.
3162 You need only define this macro if the frame address is not the same
3163 as the frame pointer. Most machines do not need to define it.
3166 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3167 A C expression whose value is RTL representing the value of the return
3168 address for the frame @var{count} steps up from the current frame, after
3169 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3170 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3171 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3173 The value of the expression must always be the correct address when
3174 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3175 determine the return address of other frames.
3178 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3179 Define this if the return address of a particular stack frame is accessed
3180 from the frame pointer of the previous stack frame.
3183 @defmac INCOMING_RETURN_ADDR_RTX
3184 A C expression whose value is RTL representing the location of the
3185 incoming return address at the beginning of any function, before the
3186 prologue. This RTL is either a @code{REG}, indicating that the return
3187 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3190 You only need to define this macro if you want to support call frame
3191 debugging information like that provided by DWARF 2.
3193 If this RTL is a @code{REG}, you should also define
3194 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3197 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3198 A C expression whose value is an integer giving a DWARF 2 column
3199 number that may be used as an alternative return column. The column
3200 must not correspond to any gcc hard register (that is, it must not
3201 be in the range of @code{DWARF_FRAME_REGNUM}).
3203 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3204 general register, but an alternative column needs to be used for signal
3205 frames. Some targets have also used different frame return columns
3209 @defmac DWARF_ZERO_REG
3210 A C expression whose value is an integer giving a DWARF 2 register
3211 number that is considered to always have the value zero. This should
3212 only be defined if the target has an architected zero register, and
3213 someone decided it was a good idea to use that register number to
3214 terminate the stack backtrace. New ports should avoid this.
3217 @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
3218 This target hook allows the backend to emit frame-related insns that
3219 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3220 info engine will invoke it on insns of the form
3222 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3226 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3228 to let the backend emit the call frame instructions. @var{label} is
3229 the CFI label attached to the insn, @var{pattern} is the pattern of
3230 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3233 @defmac INCOMING_FRAME_SP_OFFSET
3234 A C expression whose value is an integer giving the offset, in bytes,
3235 from the value of the stack pointer register to the top of the stack
3236 frame at the beginning of any function, before the prologue. The top of
3237 the frame is defined to be the value of the stack pointer in the
3238 previous frame, just before the call instruction.
3240 You only need to define this macro if you want to support call frame
3241 debugging information like that provided by DWARF 2.
3244 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3245 A C expression whose value is an integer giving the offset, in bytes,
3246 from the argument pointer to the canonical frame address (cfa). The
3247 final value should coincide with that calculated by
3248 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3249 during virtual register instantiation.
3251 The default value for this macro is
3252 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3253 which is correct for most machines; in general, the arguments are found
3254 immediately before the stack frame. Note that this is not the case on
3255 some targets that save registers into the caller's frame, such as SPARC
3256 and rs6000, and so such targets need to define this macro.
3258 You only need to define this macro if the default is incorrect, and you
3259 want to support call frame debugging information like that provided by
3263 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3264 If defined, a C expression whose value is an integer giving the offset
3265 in bytes from the frame pointer to the canonical frame address (cfa).
3266 The final value should coincide with that calculated by
3267 @code{INCOMING_FRAME_SP_OFFSET}.
3269 Normally the CFA is calculated as an offset from the argument pointer,
3270 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3271 variable due to the ABI, this may not be possible. If this macro is
3272 defined, it implies that the virtual register instantiation should be
3273 based on the frame pointer instead of the argument pointer. Only one
3274 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3278 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3279 If defined, a C expression whose value is an integer giving the offset
3280 in bytes from the canonical frame address (cfa) to the frame base used
3281 in DWARF 2 debug information. The default is zero. A different value
3282 may reduce the size of debug information on some ports.
3285 @node Exception Handling
3286 @subsection Exception Handling Support
3287 @cindex exception handling
3289 @defmac EH_RETURN_DATA_REGNO (@var{N})
3290 A C expression whose value is the @var{N}th register number used for
3291 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3292 @var{N} registers are usable.
3294 The exception handling library routines communicate with the exception
3295 handlers via a set of agreed upon registers. Ideally these registers
3296 should be call-clobbered; it is possible to use call-saved registers,
3297 but may negatively impact code size. The target must support at least
3298 2 data registers, but should define 4 if there are enough free registers.
3300 You must define this macro if you want to support call frame exception
3301 handling like that provided by DWARF 2.
3304 @defmac EH_RETURN_STACKADJ_RTX
3305 A C expression whose value is RTL representing a location in which
3306 to store a stack adjustment to be applied before function return.
3307 This is used to unwind the stack to an exception handler's call frame.
3308 It will be assigned zero on code paths that return normally.
3310 Typically this is a call-clobbered hard register that is otherwise
3311 untouched by the epilogue, but could also be a stack slot.
3313 Do not define this macro if the stack pointer is saved and restored
3314 by the regular prolog and epilog code in the call frame itself; in
3315 this case, the exception handling library routines will update the
3316 stack location to be restored in place. Otherwise, you must define
3317 this macro if you want to support call frame exception handling like
3318 that provided by DWARF 2.
3321 @defmac EH_RETURN_HANDLER_RTX
3322 A C expression whose value is RTL representing a location in which
3323 to store the address of an exception handler to which we should
3324 return. It will not be assigned on code paths that return normally.
3326 Typically this is the location in the call frame at which the normal
3327 return address is stored. For targets that return by popping an
3328 address off the stack, this might be a memory address just below
3329 the @emph{target} call frame rather than inside the current call
3330 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3331 been assigned, so it may be used to calculate the location of the
3334 Some targets have more complex requirements than storing to an
3335 address calculable during initial code generation. In that case
3336 the @code{eh_return} instruction pattern should be used instead.
3338 If you want to support call frame exception handling, you must
3339 define either this macro or the @code{eh_return} instruction pattern.
3342 @defmac RETURN_ADDR_OFFSET
3343 If defined, an integer-valued C expression for which rtl will be generated
3344 to add it to the exception handler address before it is searched in the
3345 exception handling tables, and to subtract it again from the address before
3346 using it to return to the exception handler.
3349 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3350 This macro chooses the encoding of pointers embedded in the exception
3351 handling sections. If at all possible, this should be defined such
3352 that the exception handling section will not require dynamic relocations,
3353 and so may be read-only.
3355 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3356 @var{global} is true if the symbol may be affected by dynamic relocations.
3357 The macro should return a combination of the @code{DW_EH_PE_*} defines
3358 as found in @file{dwarf2.h}.
3360 If this macro is not defined, pointers will not be encoded but
3361 represented directly.
3364 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3365 This macro allows the target to emit whatever special magic is required
3366 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3367 Generic code takes care of pc-relative and indirect encodings; this must
3368 be defined if the target uses text-relative or data-relative encodings.
3370 This is a C statement that branches to @var{done} if the format was
3371 handled. @var{encoding} is the format chosen, @var{size} is the number
3372 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3376 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3377 This macro allows the target to add CPU and operating system specific
3378 code to the call-frame unwinder for use when there is no unwind data
3379 available. The most common reason to implement this macro is to unwind
3380 through signal frames.
3382 This macro is called from @code{uw_frame_state_for} in
3383 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3384 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3385 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3386 for the address of the code being executed and @code{context->cfa} for
3387 the stack pointer value. If the frame can be decoded, the register
3388 save addresses should be updated in @var{fs} and the macro should
3389 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3390 the macro should evaluate to @code{_URC_END_OF_STACK}.
3392 For proper signal handling in Java this macro is accompanied by
3393 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3396 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3397 This macro allows the target to add operating system specific code to the
3398 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3399 usually used for signal or interrupt frames.
3401 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3402 @var{context} is an @code{_Unwind_Context};
3403 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3404 for the abi and context in the @code{.unwabi} directive. If the
3405 @code{.unwabi} directive can be handled, the register save addresses should
3406 be updated in @var{fs}.
3409 @defmac TARGET_USES_WEAK_UNWIND_INFO
3410 A C expression that evaluates to true if the target requires unwind
3411 info to be given comdat linkage. Define it to be @code{1} if comdat
3412 linkage is necessary. The default is @code{0}.
3415 @node Stack Checking
3416 @subsection Specifying How Stack Checking is Done
3418 GCC will check that stack references are within the boundaries of the
3419 stack, if the option @option{-fstack-check} is specified, in one of
3424 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3425 will assume that you have arranged for full stack checking to be done
3426 at appropriate places in the configuration files. GCC will not do
3427 other special processing.
3430 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3431 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3432 that you have arranged for static stack checking (checking of the
3433 static stack frame of functions) to be done at appropriate places
3434 in the configuration files. GCC will only emit code to do dynamic
3435 stack checking (checking on dynamic stack allocations) using the third
3439 If neither of the above are true, GCC will generate code to periodically
3440 ``probe'' the stack pointer using the values of the macros defined below.
3443 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3444 GCC will change its allocation strategy for large objects if the option
3445 @option{-fstack-check} is specified: they will always be allocated
3446 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3448 @defmac STACK_CHECK_BUILTIN
3449 A nonzero value if stack checking is done by the configuration files in a
3450 machine-dependent manner. You should define this macro if stack checking
3451 is required by the ABI of your machine or if you would like to do stack
3452 checking in some more efficient way than the generic approach. The default
3453 value of this macro is zero.
3456 @defmac STACK_CHECK_STATIC_BUILTIN
3457 A nonzero value if static stack checking is done by the configuration files
3458 in a machine-dependent manner. You should define this macro if you would
3459 like to do static stack checking in some more efficient way than the generic
3460 approach. The default value of this macro is zero.
3463 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3464 An integer specifying the interval at which GCC must generate stack probe
3465 instructions, defined as 2 raised to this integer. You will normally
3466 define this macro so that the interval be no larger than the size of
3467 the ``guard pages'' at the end of a stack area. The default value
3468 of 12 (4096-byte interval) is suitable for most systems.
3471 @defmac STACK_CHECK_MOVING_SP
3472 An integer which is nonzero if GCC should move the stack pointer page by page
3473 when doing probes. This can be necessary on systems where the stack pointer
3474 contains the bottom address of the memory area accessible to the executing
3475 thread at any point in time. In this situation an alternate signal stack
3476 is required in order to be able to recover from a stack overflow. The
3477 default value of this macro is zero.
3480 @defmac STACK_CHECK_PROTECT
3481 The number of bytes of stack needed to recover from a stack overflow, for
3482 languages where such a recovery is supported. The default value of 75 words
3483 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3484 8192 bytes with other exception handling mechanisms should be adequate for
3488 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3489 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3490 in the opposite case.
3492 @defmac STACK_CHECK_MAX_FRAME_SIZE
3493 The maximum size of a stack frame, in bytes. GCC will generate probe
3494 instructions in non-leaf functions to ensure at least this many bytes of
3495 stack are available. If a stack frame is larger than this size, stack
3496 checking will not be reliable and GCC will issue a warning. The
3497 default is chosen so that GCC only generates one instruction on most
3498 systems. You should normally not change the default value of this macro.
3501 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3502 GCC uses this value to generate the above warning message. It
3503 represents the amount of fixed frame used by a function, not including
3504 space for any callee-saved registers, temporaries and user variables.
3505 You need only specify an upper bound for this amount and will normally
3506 use the default of four words.
3509 @defmac STACK_CHECK_MAX_VAR_SIZE
3510 The maximum size, in bytes, of an object that GCC will place in the
3511 fixed area of the stack frame when the user specifies
3512 @option{-fstack-check}.
3513 GCC computed the default from the values of the above macros and you will
3514 normally not need to override that default.
3518 @node Frame Registers
3519 @subsection Registers That Address the Stack Frame
3521 @c prevent bad page break with this line
3522 This discusses registers that address the stack frame.
3524 @defmac STACK_POINTER_REGNUM
3525 The register number of the stack pointer register, which must also be a
3526 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3527 the hardware determines which register this is.
3530 @defmac FRAME_POINTER_REGNUM
3531 The register number of the frame pointer register, which is used to
3532 access automatic variables in the stack frame. On some machines, the
3533 hardware determines which register this is. On other machines, you can
3534 choose any register you wish for this purpose.
3537 @defmac HARD_FRAME_POINTER_REGNUM
3538 On some machines the offset between the frame pointer and starting
3539 offset of the automatic variables is not known until after register
3540 allocation has been done (for example, because the saved registers are
3541 between these two locations). On those machines, define
3542 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3543 be used internally until the offset is known, and define
3544 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3545 used for the frame pointer.
3547 You should define this macro only in the very rare circumstances when it
3548 is not possible to calculate the offset between the frame pointer and
3549 the automatic variables until after register allocation has been
3550 completed. When this macro is defined, you must also indicate in your
3551 definition of @code{ELIMINABLE_REGS} how to eliminate
3552 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3553 or @code{STACK_POINTER_REGNUM}.
3555 Do not define this macro if it would be the same as
3556 @code{FRAME_POINTER_REGNUM}.
3559 @defmac ARG_POINTER_REGNUM
3560 The register number of the arg pointer register, which is used to access
3561 the function's argument list. On some machines, this is the same as the
3562 frame pointer register. On some machines, the hardware determines which
3563 register this is. On other machines, you can choose any register you
3564 wish for this purpose. If this is not the same register as the frame
3565 pointer register, then you must mark it as a fixed register according to
3566 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3567 (@pxref{Elimination}).
3570 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3571 Define this to a preprocessor constant that is nonzero if
3572 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3573 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3574 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3575 definition is not suitable for use in preprocessor conditionals.
3578 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3579 Define this to a preprocessor constant that is nonzero if
3580 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3581 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3582 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3583 definition is not suitable for use in preprocessor conditionals.
3586 @defmac RETURN_ADDRESS_POINTER_REGNUM
3587 The register number of the return address pointer register, which is used to
3588 access the current function's return address from the stack. On some
3589 machines, the return address is not at a fixed offset from the frame
3590 pointer or stack pointer or argument pointer. This register can be defined
3591 to point to the return address on the stack, and then be converted by
3592 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3594 Do not define this macro unless there is no other way to get the return
3595 address from the stack.
3598 @defmac STATIC_CHAIN_REGNUM
3599 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3600 Register numbers used for passing a function's static chain pointer. If
3601 register windows are used, the register number as seen by the called
3602 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3603 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3604 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3607 The static chain register need not be a fixed register.
3609 If the static chain is passed in memory, these macros should not be
3610 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3613 @hook TARGET_STATIC_CHAIN
3614 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3615 targets that may use different static chain locations for different
3616 nested functions. This may be required if the target has function
3617 attributes that affect the calling conventions of the function and
3618 those calling conventions use different static chain locations.
3620 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3622 If the static chain is passed in memory, this hook should be used to
3623 provide rtx giving @code{mem} expressions that denote where they are stored.
3624 Often the @code{mem} expression as seen by the caller will be at an offset
3625 from the stack pointer and the @code{mem} expression as seen by the callee
3626 will be at an offset from the frame pointer.
3627 @findex stack_pointer_rtx
3628 @findex frame_pointer_rtx
3629 @findex arg_pointer_rtx
3630 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3631 @code{arg_pointer_rtx} will have been initialized and should be used
3632 to refer to those items.
3635 @defmac DWARF_FRAME_REGISTERS
3636 This macro specifies the maximum number of hard registers that can be
3637 saved in a call frame. This is used to size data structures used in
3638 DWARF2 exception handling.
3640 Prior to GCC 3.0, this macro was needed in order to establish a stable
3641 exception handling ABI in the face of adding new hard registers for ISA
3642 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3643 in the number of hard registers. Nevertheless, this macro can still be
3644 used to reduce the runtime memory requirements of the exception handling
3645 routines, which can be substantial if the ISA contains a lot of
3646 registers that are not call-saved.
3648 If this macro is not defined, it defaults to
3649 @code{FIRST_PSEUDO_REGISTER}.
3652 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3654 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3655 for backward compatibility in pre GCC 3.0 compiled code.
3657 If this macro is not defined, it defaults to
3658 @code{DWARF_FRAME_REGISTERS}.
3661 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3663 Define this macro if the target's representation for dwarf registers
3664 is different than the internal representation for unwind column.
3665 Given a dwarf register, this macro should return the internal unwind
3666 column number to use instead.
3668 See the PowerPC's SPE target for an example.
3671 @defmac DWARF_FRAME_REGNUM (@var{regno})
3673 Define this macro if the target's representation for dwarf registers
3674 used in .eh_frame or .debug_frame is different from that used in other
3675 debug info sections. Given a GCC hard register number, this macro
3676 should return the .eh_frame register number. The default is
3677 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3681 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3683 Define this macro to map register numbers held in the call frame info
3684 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3685 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3686 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3687 return @code{@var{regno}}.
3691 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3693 Define this macro if the target stores register values as
3694 @code{_Unwind_Word} type in unwind context. It should be defined if
3695 target register size is larger than the size of @code{void *}. The
3696 default is to store register values as @code{void *} type.
3700 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3702 Define this macro to be 1 if the target always uses extended unwind
3703 context with version, args_size and by_value fields. If it is undefined,
3704 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3705 defined and 0 otherwise.
3710 @subsection Eliminating Frame Pointer and Arg Pointer
3712 @c prevent bad page break with this line
3713 This is about eliminating the frame pointer and arg pointer.
3715 @hook TARGET_FRAME_POINTER_REQUIRED
3716 This target hook should return @code{true} if a function must have and use
3717 a frame pointer. This target hook is called in the reload pass. If its return
3718 value is @code{true} the function will have a frame pointer.
3720 This target hook can in principle examine the current function and decide
3721 according to the facts, but on most machines the constant @code{false} or the
3722 constant @code{true} suffices. Use @code{false} when the machine allows code
3723 to be generated with no frame pointer, and doing so saves some time or space.
3724 Use @code{true} when there is no possible advantage to avoiding a frame
3727 In certain cases, the compiler does not know how to produce valid code
3728 without a frame pointer. The compiler recognizes those cases and
3729 automatically gives the function a frame pointer regardless of what
3730 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3733 In a function that does not require a frame pointer, the frame pointer
3734 register can be allocated for ordinary usage, unless you mark it as a
3735 fixed register. See @code{FIXED_REGISTERS} for more information.
3737 Default return value is @code{false}.
3740 @findex get_frame_size
3741 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3742 A C statement to store in the variable @var{depth-var} the difference
3743 between the frame pointer and the stack pointer values immediately after
3744 the function prologue. The value would be computed from information
3745 such as the result of @code{get_frame_size ()} and the tables of
3746 registers @code{regs_ever_live} and @code{call_used_regs}.
3748 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3749 need not be defined. Otherwise, it must be defined even if
3750 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3751 case, you may set @var{depth-var} to anything.
3754 @defmac ELIMINABLE_REGS
3755 If defined, this macro specifies a table of register pairs used to
3756 eliminate unneeded registers that point into the stack frame. If it is not
3757 defined, the only elimination attempted by the compiler is to replace
3758 references to the frame pointer with references to the stack pointer.
3760 The definition of this macro is a list of structure initializations, each
3761 of which specifies an original and replacement register.
3763 On some machines, the position of the argument pointer is not known until
3764 the compilation is completed. In such a case, a separate hard register
3765 must be used for the argument pointer. This register can be eliminated by
3766 replacing it with either the frame pointer or the argument pointer,
3767 depending on whether or not the frame pointer has been eliminated.
3769 In this case, you might specify:
3771 #define ELIMINABLE_REGS \
3772 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3773 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3774 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3777 Note that the elimination of the argument pointer with the stack pointer is
3778 specified first since that is the preferred elimination.
3781 @hook TARGET_CAN_ELIMINATE
3782 This target hook should returns @code{true} if the compiler is allowed to
3783 try to replace register number @var{from_reg} with register number
3784 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3785 is defined, and will usually be @code{true}, since most of the cases
3786 preventing register elimination are things that the compiler already
3789 Default return value is @code{true}.
3792 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3793 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3794 specifies the initial difference between the specified pair of
3795 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3799 @node Stack Arguments
3800 @subsection Passing Function Arguments on the Stack
3801 @cindex arguments on stack
3802 @cindex stack arguments
3804 The macros in this section control how arguments are passed
3805 on the stack. See the following section for other macros that
3806 control passing certain arguments in registers.
3808 @hook TARGET_PROMOTE_PROTOTYPES
3809 This target hook returns @code{true} if an argument declared in a
3810 prototype as an integral type smaller than @code{int} should actually be
3811 passed as an @code{int}. In addition to avoiding errors in certain
3812 cases of mismatch, it also makes for better code on certain machines.
3813 The default is to not promote prototypes.
3817 A C expression. If nonzero, push insns will be used to pass
3819 If the target machine does not have a push instruction, set it to zero.
3820 That directs GCC to use an alternate strategy: to
3821 allocate the entire argument block and then store the arguments into
3822 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3825 @defmac PUSH_ARGS_REVERSED
3826 A C expression. If nonzero, function arguments will be evaluated from
3827 last to first, rather than from first to last. If this macro is not
3828 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3829 and args grow in opposite directions, and 0 otherwise.
3832 @defmac PUSH_ROUNDING (@var{npushed})
3833 A C expression that is the number of bytes actually pushed onto the
3834 stack when an instruction attempts to push @var{npushed} bytes.
3836 On some machines, the definition
3839 #define PUSH_ROUNDING(BYTES) (BYTES)
3843 will suffice. But on other machines, instructions that appear
3844 to push one byte actually push two bytes in an attempt to maintain
3845 alignment. Then the definition should be
3848 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3851 If the value of this macro has a type, it should be an unsigned type.
3854 @findex current_function_outgoing_args_size
3855 @defmac ACCUMULATE_OUTGOING_ARGS
3856 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3857 will be computed and placed into the variable
3858 @code{current_function_outgoing_args_size}. No space will be pushed
3859 onto the stack for each call; instead, the function prologue should
3860 increase the stack frame size by this amount.
3862 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3866 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3867 Define this macro if functions should assume that stack space has been
3868 allocated for arguments even when their values are passed in
3871 The value of this macro is the size, in bytes, of the area reserved for
3872 arguments passed in registers for the function represented by @var{fndecl},
3873 which can be zero if GCC is calling a library function.
3874 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3877 This space can be allocated by the caller, or be a part of the
3878 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3881 @c above is overfull. not sure what to do. --mew 5feb93 did
3882 @c something, not sure if it looks good. --mew 10feb93
3884 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3885 Define this to a nonzero value if it is the responsibility of the
3886 caller to allocate the area reserved for arguments passed in registers
3887 when calling a function of @var{fntype}. @var{fntype} may be NULL
3888 if the function called is a library function.
3890 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3891 whether the space for these arguments counts in the value of
3892 @code{current_function_outgoing_args_size}.
3895 @defmac STACK_PARMS_IN_REG_PARM_AREA
3896 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3897 stack parameters don't skip the area specified by it.
3898 @c i changed this, makes more sens and it should have taken care of the
3899 @c overfull.. not as specific, tho. --mew 5feb93
3901 Normally, when a parameter is not passed in registers, it is placed on the
3902 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3903 suppresses this behavior and causes the parameter to be passed on the
3904 stack in its natural location.
3907 @hook TARGET_RETURN_POPS_ARGS
3908 This target hook returns the number of bytes of its own arguments that
3909 a function pops on returning, or 0 if the function pops no arguments
3910 and the caller must therefore pop them all after the function returns.
3912 @var{fundecl} is a C variable whose value is a tree node that describes
3913 the function in question. Normally it is a node of type
3914 @code{FUNCTION_DECL} that describes the declaration of the function.
3915 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3917 @var{funtype} is a C variable whose value is a tree node that
3918 describes the function in question. Normally it is a node of type
3919 @code{FUNCTION_TYPE} that describes the data type of the function.
3920 From this it is possible to obtain the data types of the value and
3921 arguments (if known).
3923 When a call to a library function is being considered, @var{fundecl}
3924 will contain an identifier node for the library function. Thus, if
3925 you need to distinguish among various library functions, you can do so
3926 by their names. Note that ``library function'' in this context means
3927 a function used to perform arithmetic, whose name is known specially
3928 in the compiler and was not mentioned in the C code being compiled.
3930 @var{size} is the number of bytes of arguments passed on the
3931 stack. If a variable number of bytes is passed, it is zero, and
3932 argument popping will always be the responsibility of the calling function.
3934 On the VAX, all functions always pop their arguments, so the definition
3935 of this macro is @var{size}. On the 68000, using the standard
3936 calling convention, no functions pop their arguments, so the value of
3937 the macro is always 0 in this case. But an alternative calling
3938 convention is available in which functions that take a fixed number of
3939 arguments pop them but other functions (such as @code{printf}) pop
3940 nothing (the caller pops all). When this convention is in use,
3941 @var{funtype} is examined to determine whether a function takes a fixed
3942 number of arguments.
3945 @defmac CALL_POPS_ARGS (@var{cum})
3946 A C expression that should indicate the number of bytes a call sequence
3947 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3948 when compiling a function call.
3950 @var{cum} is the variable in which all arguments to the called function
3951 have been accumulated.
3953 On certain architectures, such as the SH5, a call trampoline is used
3954 that pops certain registers off the stack, depending on the arguments
3955 that have been passed to the function. Since this is a property of the
3956 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3960 @node Register Arguments
3961 @subsection Passing Arguments in Registers
3962 @cindex arguments in registers
3963 @cindex registers arguments
3965 This section describes the macros which let you control how various
3966 types of arguments are passed in registers or how they are arranged in
3969 @hook TARGET_FUNCTION_ARG
3970 Return an RTX indicating whether a function argument is passed in a
3971 register and if so, which register.
3973 The arguments are @var{ca}, which summarizes all the previous
3974 arguments; @var{mode}, the machine mode of the argument; @var{type},
3975 the data type of the argument as a tree node or 0 if that is not known
3976 (which happens for C support library functions); and @var{named},
3977 which is @code{true} for an ordinary argument and @code{false} for
3978 nameless arguments that correspond to @samp{@dots{}} in the called
3979 function's prototype. @var{type} can be an incomplete type if a
3980 syntax error has previously occurred.
3982 The return value is usually either a @code{reg} RTX for the hard
3983 register in which to pass the argument, or zero to pass the argument
3986 The value of the expression can also be a @code{parallel} RTX@. This is
3987 used when an argument is passed in multiple locations. The mode of the
3988 @code{parallel} should be the mode of the entire argument. The
3989 @code{parallel} holds any number of @code{expr_list} pairs; each one
3990 describes where part of the argument is passed. In each
3991 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3992 register in which to pass this part of the argument, and the mode of the
3993 register RTX indicates how large this part of the argument is. The
3994 second operand of the @code{expr_list} is a @code{const_int} which gives
3995 the offset in bytes into the entire argument of where this part starts.
3996 As a special exception the first @code{expr_list} in the @code{parallel}
3997 RTX may have a first operand of zero. This indicates that the entire
3998 argument is also stored on the stack.
4000 The last time this hook is called, it is called with @code{MODE ==
4001 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
4002 pattern as operands 2 and 3 respectively.
4004 @cindex @file{stdarg.h} and register arguments
4005 The usual way to make the ISO library @file{stdarg.h} work on a
4006 machine where some arguments are usually passed in registers, is to
4007 cause nameless arguments to be passed on the stack instead. This is
4008 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
4009 @var{named} is @code{false}.
4011 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
4012 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
4013 You may use the hook @code{targetm.calls.must_pass_in_stack}
4014 in the definition of this macro to determine if this argument is of a
4015 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4016 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
4017 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4018 defined, the argument will be computed in the stack and then loaded into
4022 @hook TARGET_MUST_PASS_IN_STACK
4023 This target hook should return @code{true} if we should not pass @var{type}
4024 solely in registers. The file @file{expr.h} defines a
4025 definition that is usually appropriate, refer to @file{expr.h} for additional
4029 @hook TARGET_FUNCTION_INCOMING_ARG
4030 Define this hook if the target machine has ``register windows'', so
4031 that the register in which a function sees an arguments is not
4032 necessarily the same as the one in which the caller passed the
4035 For such machines, @code{TARGET_FUNCTION_ARG} computes the register in
4036 which the caller passes the value, and
4037 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4038 fashion to tell the function being called where the arguments will
4041 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4042 @code{TARGET_FUNCTION_ARG} serves both purposes.
4045 @hook TARGET_ARG_PARTIAL_BYTES
4046 This target hook returns the number of bytes at the beginning of an
4047 argument that must be put in registers. The value must be zero for
4048 arguments that are passed entirely in registers or that are entirely
4049 pushed on the stack.
4051 On some machines, certain arguments must be passed partially in
4052 registers and partially in memory. On these machines, typically the
4053 first few words of arguments are passed in registers, and the rest
4054 on the stack. If a multi-word argument (a @code{double} or a
4055 structure) crosses that boundary, its first few words must be passed
4056 in registers and the rest must be pushed. This macro tells the
4057 compiler when this occurs, and how many bytes should go in registers.
4059 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4060 register to be used by the caller for this argument; likewise
4061 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4064 @hook TARGET_PASS_BY_REFERENCE
4065 This target hook should return @code{true} if an argument at the
4066 position indicated by @var{cum} should be passed by reference. This
4067 predicate is queried after target independent reasons for being
4068 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4070 If the hook returns true, a copy of that argument is made in memory and a
4071 pointer to the argument is passed instead of the argument itself.
4072 The pointer is passed in whatever way is appropriate for passing a pointer
4076 @hook TARGET_CALLEE_COPIES
4077 The function argument described by the parameters to this hook is
4078 known to be passed by reference. The hook should return true if the
4079 function argument should be copied by the callee instead of copied
4082 For any argument for which the hook returns true, if it can be
4083 determined that the argument is not modified, then a copy need
4086 The default version of this hook always returns false.
4089 @defmac CUMULATIVE_ARGS
4090 A C type for declaring a variable that is used as the first argument
4091 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4092 target machines, the type @code{int} suffices and can hold the number
4093 of bytes of argument so far.
4095 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4096 arguments that have been passed on the stack. The compiler has other
4097 variables to keep track of that. For target machines on which all
4098 arguments are passed on the stack, there is no need to store anything in
4099 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4100 should not be empty, so use @code{int}.
4103 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4104 If defined, this macro is called before generating any code for a
4105 function, but after the @var{cfun} descriptor for the function has been
4106 created. The back end may use this macro to update @var{cfun} to
4107 reflect an ABI other than that which would normally be used by default.
4108 If the compiler is generating code for a compiler-generated function,
4109 @var{fndecl} may be @code{NULL}.
4112 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4113 A C statement (sans semicolon) for initializing the variable
4114 @var{cum} for the state at the beginning of the argument list. The
4115 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4116 is the tree node for the data type of the function which will receive
4117 the args, or 0 if the args are to a compiler support library function.
4118 For direct calls that are not libcalls, @var{fndecl} contain the
4119 declaration node of the function. @var{fndecl} is also set when
4120 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4121 being compiled. @var{n_named_args} is set to the number of named
4122 arguments, including a structure return address if it is passed as a
4123 parameter, when making a call. When processing incoming arguments,
4124 @var{n_named_args} is set to @minus{}1.
4126 When processing a call to a compiler support library function,
4127 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4128 contains the name of the function, as a string. @var{libname} is 0 when
4129 an ordinary C function call is being processed. Thus, each time this
4130 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4131 never both of them at once.
4134 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4135 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4136 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4137 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4138 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4139 0)} is used instead.
4142 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4143 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4144 finding the arguments for the function being compiled. If this macro is
4145 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4147 The value passed for @var{libname} is always 0, since library routines
4148 with special calling conventions are never compiled with GCC@. The
4149 argument @var{libname} exists for symmetry with
4150 @code{INIT_CUMULATIVE_ARGS}.
4151 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4152 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4155 @hook TARGET_FUNCTION_ARG_ADVANCE
4156 This hook updates the summarizer variable pointed to by @var{ca} to
4157 advance past an argument in the argument list. The values @var{mode},
4158 @var{type} and @var{named} describe that argument. Once this is done,
4159 the variable @var{cum} is suitable for analyzing the @emph{following}
4160 argument with @code{TARGET_FUNCTION_ARG}, etc.
4162 This hook need not do anything if the argument in question was passed
4163 on the stack. The compiler knows how to track the amount of stack space
4164 used for arguments without any special help.
4167 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4168 If defined, a C expression that is the number of bytes to add to the
4169 offset of the argument passed in memory. This is needed for the SPU,
4170 which passes @code{char} and @code{short} arguments in the preferred
4171 slot that is in the middle of the quad word instead of starting at the
4175 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4176 If defined, a C expression which determines whether, and in which direction,
4177 to pad out an argument with extra space. The value should be of type
4178 @code{enum direction}: either @code{upward} to pad above the argument,
4179 @code{downward} to pad below, or @code{none} to inhibit padding.
4181 The @emph{amount} of padding is not controlled by this macro, but by the
4182 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
4183 always just enough to reach the next multiple of that boundary.
4185 This macro has a default definition which is right for most systems.
4186 For little-endian machines, the default is to pad upward. For
4187 big-endian machines, the default is to pad downward for an argument of
4188 constant size shorter than an @code{int}, and upward otherwise.
4191 @defmac PAD_VARARGS_DOWN
4192 If defined, a C expression which determines whether the default
4193 implementation of va_arg will attempt to pad down before reading the
4194 next argument, if that argument is smaller than its aligned space as
4195 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4196 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4199 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4200 Specify padding for the last element of a block move between registers and
4201 memory. @var{first} is nonzero if this is the only element. Defining this
4202 macro allows better control of register function parameters on big-endian
4203 machines, without using @code{PARALLEL} rtl. In particular,
4204 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4205 registers, as there is no longer a "wrong" part of a register; For example,
4206 a three byte aggregate may be passed in the high part of a register if so
4210 @hook TARGET_FUNCTION_ARG_BOUNDARY
4211 This hook returns the alignment boundary, in bits, of an argument
4212 with the specified mode and type. The default hook returns
4213 @code{PARM_BOUNDARY} for all arguments.
4216 @hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY
4218 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4219 A C expression that is nonzero if @var{regno} is the number of a hard
4220 register in which function arguments are sometimes passed. This does
4221 @emph{not} include implicit arguments such as the static chain and
4222 the structure-value address. On many machines, no registers can be
4223 used for this purpose since all function arguments are pushed on the
4227 @hook TARGET_SPLIT_COMPLEX_ARG
4228 This hook should return true if parameter of type @var{type} are passed
4229 as two scalar parameters. By default, GCC will attempt to pack complex
4230 arguments into the target's word size. Some ABIs require complex arguments
4231 to be split and treated as their individual components. For example, on
4232 AIX64, complex floats should be passed in a pair of floating point
4233 registers, even though a complex float would fit in one 64-bit floating
4236 The default value of this hook is @code{NULL}, which is treated as always
4240 @hook TARGET_BUILD_BUILTIN_VA_LIST
4241 This hook returns a type node for @code{va_list} for the target.
4242 The default version of the hook returns @code{void*}.
4245 @hook TARGET_ENUM_VA_LIST_P
4246 This target hook is used in function @code{c_common_nodes_and_builtins}
4247 to iterate through the target specific builtin types for va_list. The
4248 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4249 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4251 The arguments @var{pname} and @var{ptree} are used to store the result of
4252 this macro and are set to the name of the va_list builtin type and its
4254 If the return value of this macro is zero, then there is no more element.
4255 Otherwise the @var{IDX} should be increased for the next call of this
4256 macro to iterate through all types.
4259 @hook TARGET_FN_ABI_VA_LIST
4260 This hook returns the va_list type of the calling convention specified by
4262 The default version of this hook returns @code{va_list_type_node}.
4265 @hook TARGET_CANONICAL_VA_LIST_TYPE
4266 This hook returns the va_list type of the calling convention specified by the
4267 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4271 @hook TARGET_GIMPLIFY_VA_ARG_EXPR
4272 This hook performs target-specific gimplification of
4273 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4274 arguments to @code{va_arg}; the latter two are as in
4275 @code{gimplify.c:gimplify_expr}.
4278 @hook TARGET_VALID_POINTER_MODE
4279 Define this to return nonzero if the port can handle pointers
4280 with machine mode @var{mode}. The default version of this
4281 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4284 @hook TARGET_REF_MAY_ALIAS_ERRNO
4286 @hook TARGET_SCALAR_MODE_SUPPORTED_P
4287 Define this to return nonzero if the port is prepared to handle
4288 insns involving scalar mode @var{mode}. For a scalar mode to be
4289 considered supported, all the basic arithmetic and comparisons
4292 The default version of this hook returns true for any mode
4293 required to handle the basic C types (as defined by the port).
4294 Included here are the double-word arithmetic supported by the
4295 code in @file{optabs.c}.
4298 @hook TARGET_VECTOR_MODE_SUPPORTED_P
4299 Define this to return nonzero if the port is prepared to handle
4300 insns involving vector mode @var{mode}. At the very least, it
4301 must have move patterns for this mode.
4304 @hook TARGET_ARRAY_MODE_SUPPORTED_P
4306 @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
4307 Define this to return nonzero for machine modes for which the port has
4308 small register classes. If this target hook returns nonzero for a given
4309 @var{mode}, the compiler will try to minimize the lifetime of registers
4310 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4311 In this case, the hook is expected to return nonzero if it returns nonzero
4314 On some machines, it is risky to let hard registers live across arbitrary
4315 insns. Typically, these machines have instructions that require values
4316 to be in specific registers (like an accumulator), and reload will fail
4317 if the required hard register is used for another purpose across such an
4320 Passes before reload do not know which hard registers will be used
4321 in an instruction, but the machine modes of the registers set or used in
4322 the instruction are already known. And for some machines, register
4323 classes are small for, say, integer registers but not for floating point
4324 registers. For example, the AMD x86-64 architecture requires specific
4325 registers for the legacy x86 integer instructions, but there are many
4326 SSE registers for floating point operations. On such targets, a good
4327 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4328 machine modes but zero for the SSE register classes.
4330 The default version of this hook returns false for any mode. It is always
4331 safe to redefine this hook to return with a nonzero value. But if you
4332 unnecessarily define it, you will reduce the amount of optimizations
4333 that can be performed in some cases. If you do not define this hook
4334 to return a nonzero value when it is required, the compiler will run out
4335 of spill registers and print a fatal error message.
4338 @hook TARGET_FLAGS_REGNUM
4341 @subsection How Scalar Function Values Are Returned
4342 @cindex return values in registers
4343 @cindex values, returned by functions
4344 @cindex scalars, returned as values
4346 This section discusses the macros that control returning scalars as
4347 values---values that can fit in registers.
4349 @hook TARGET_FUNCTION_VALUE
4351 Define this to return an RTX representing the place where a function
4352 returns or receives a value of data type @var{ret_type}, a tree node
4353 representing a data type. @var{fn_decl_or_type} is a tree node
4354 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4355 function being called. If @var{outgoing} is false, the hook should
4356 compute the register in which the caller will see the return value.
4357 Otherwise, the hook should return an RTX representing the place where
4358 a function returns a value.
4360 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4361 (Actually, on most machines, scalar values are returned in the same
4362 place regardless of mode.) The value of the expression is usually a
4363 @code{reg} RTX for the hard register where the return value is stored.
4364 The value can also be a @code{parallel} RTX, if the return value is in
4365 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4366 @code{parallel} form. Note that the callee will populate every
4367 location specified in the @code{parallel}, but if the first element of
4368 the @code{parallel} contains the whole return value, callers will use
4369 that element as the canonical location and ignore the others. The m68k
4370 port uses this type of @code{parallel} to return pointers in both
4371 @samp{%a0} (the canonical location) and @samp{%d0}.
4373 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4374 the same promotion rules specified in @code{PROMOTE_MODE} if
4375 @var{valtype} is a scalar type.
4377 If the precise function being called is known, @var{func} is a tree
4378 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4379 pointer. This makes it possible to use a different value-returning
4380 convention for specific functions when all their calls are
4383 Some target machines have ``register windows'' so that the register in
4384 which a function returns its value is not the same as the one in which
4385 the caller sees the value. For such machines, you should return
4386 different RTX depending on @var{outgoing}.
4388 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4389 aggregate data types, because these are returned in another way. See
4390 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4393 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4394 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4395 a new target instead.
4398 @defmac LIBCALL_VALUE (@var{mode})
4399 A C expression to create an RTX representing the place where a library
4400 function returns a value of mode @var{mode}.
4402 Note that ``library function'' in this context means a compiler
4403 support routine, used to perform arithmetic, whose name is known
4404 specially by the compiler and was not mentioned in the C code being
4408 @hook TARGET_LIBCALL_VALUE
4409 Define this hook if the back-end needs to know the name of the libcall
4410 function in order to determine where the result should be returned.
4412 The mode of the result is given by @var{mode} and the name of the called
4413 library function is given by @var{fun}. The hook should return an RTX
4414 representing the place where the library function result will be returned.
4416 If this hook is not defined, then LIBCALL_VALUE will be used.
4419 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4420 A C expression that is nonzero if @var{regno} is the number of a hard
4421 register in which the values of called function may come back.
4423 A register whose use for returning values is limited to serving as the
4424 second of a pair (for a value of type @code{double}, say) need not be
4425 recognized by this macro. So for most machines, this definition
4429 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4432 If the machine has register windows, so that the caller and the called
4433 function use different registers for the return value, this macro
4434 should recognize only the caller's register numbers.
4436 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4437 for a new target instead.
4440 @hook TARGET_FUNCTION_VALUE_REGNO_P
4441 A target hook that return @code{true} if @var{regno} is the number of a hard
4442 register in which the values of called function may come back.
4444 A register whose use for returning values is limited to serving as the
4445 second of a pair (for a value of type @code{double}, say) need not be
4446 recognized by this target hook.
4448 If the machine has register windows, so that the caller and the called
4449 function use different registers for the return value, this target hook
4450 should recognize only the caller's register numbers.
4452 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4455 @defmac APPLY_RESULT_SIZE
4456 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4457 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4458 saving and restoring an arbitrary return value.
4461 @hook TARGET_RETURN_IN_MSB
4462 This hook should return true if values of type @var{type} are returned
4463 at the most significant end of a register (in other words, if they are
4464 padded at the least significant end). You can assume that @var{type}
4465 is returned in a register; the caller is required to check this.
4467 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4468 be able to hold the complete return value. For example, if a 1-, 2-
4469 or 3-byte structure is returned at the most significant end of a
4470 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4474 @node Aggregate Return
4475 @subsection How Large Values Are Returned
4476 @cindex aggregates as return values
4477 @cindex large return values
4478 @cindex returning aggregate values
4479 @cindex structure value address
4481 When a function value's mode is @code{BLKmode} (and in some other
4482 cases), the value is not returned according to
4483 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4484 caller passes the address of a block of memory in which the value
4485 should be stored. This address is called the @dfn{structure value
4488 This section describes how to control returning structure values in
4491 @hook TARGET_RETURN_IN_MEMORY
4492 This target hook should return a nonzero value to say to return the
4493 function value in memory, just as large structures are always returned.
4494 Here @var{type} will be the data type of the value, and @var{fntype}
4495 will be the type of the function doing the returning, or @code{NULL} for
4498 Note that values of mode @code{BLKmode} must be explicitly handled
4499 by this function. Also, the option @option{-fpcc-struct-return}
4500 takes effect regardless of this macro. On most systems, it is
4501 possible to leave the hook undefined; this causes a default
4502 definition to be used, whose value is the constant 1 for @code{BLKmode}
4503 values, and 0 otherwise.
4505 Do not use this hook to indicate that structures and unions should always
4506 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4510 @defmac DEFAULT_PCC_STRUCT_RETURN
4511 Define this macro to be 1 if all structure and union return values must be
4512 in memory. Since this results in slower code, this should be defined
4513 only if needed for compatibility with other compilers or with an ABI@.
4514 If you define this macro to be 0, then the conventions used for structure
4515 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4518 If not defined, this defaults to the value 1.
4521 @hook TARGET_STRUCT_VALUE_RTX
4522 This target hook should return the location of the structure value
4523 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4524 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4525 be @code{NULL}, for libcalls. You do not need to define this target
4526 hook if the address is always passed as an ``invisible'' first
4529 On some architectures the place where the structure value address
4530 is found by the called function is not the same place that the
4531 caller put it. This can be due to register windows, or it could
4532 be because the function prologue moves it to a different place.
4533 @var{incoming} is @code{1} or @code{2} when the location is needed in
4534 the context of the called function, and @code{0} in the context of
4537 If @var{incoming} is nonzero and the address is to be found on the
4538 stack, return a @code{mem} which refers to the frame pointer. If
4539 @var{incoming} is @code{2}, the result is being used to fetch the
4540 structure value address at the beginning of a function. If you need
4541 to emit adjusting code, you should do it at this point.
4544 @defmac PCC_STATIC_STRUCT_RETURN
4545 Define this macro if the usual system convention on the target machine
4546 for returning structures and unions is for the called function to return
4547 the address of a static variable containing the value.
4549 Do not define this if the usual system convention is for the caller to
4550 pass an address to the subroutine.
4552 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4553 nothing when you use @option{-freg-struct-return} mode.
4556 @hook TARGET_GET_RAW_RESULT_MODE
4558 @hook TARGET_GET_RAW_ARG_MODE
4561 @subsection Caller-Saves Register Allocation
4563 If you enable it, GCC can save registers around function calls. This
4564 makes it possible to use call-clobbered registers to hold variables that
4565 must live across calls.
4567 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4568 A C expression to determine whether it is worthwhile to consider placing
4569 a pseudo-register in a call-clobbered hard register and saving and
4570 restoring it around each function call. The expression should be 1 when
4571 this is worth doing, and 0 otherwise.
4573 If you don't define this macro, a default is used which is good on most
4574 machines: @code{4 * @var{calls} < @var{refs}}.
4577 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4578 A C expression specifying which mode is required for saving @var{nregs}
4579 of a pseudo-register in call-clobbered hard register @var{regno}. If
4580 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4581 returned. For most machines this macro need not be defined since GCC
4582 will select the smallest suitable mode.
4585 @node Function Entry
4586 @subsection Function Entry and Exit
4587 @cindex function entry and exit
4591 This section describes the macros that output function entry
4592 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4594 @hook TARGET_ASM_FUNCTION_PROLOGUE
4595 If defined, a function that outputs the assembler code for entry to a
4596 function. The prologue is responsible for setting up the stack frame,
4597 initializing the frame pointer register, saving registers that must be
4598 saved, and allocating @var{size} additional bytes of storage for the
4599 local variables. @var{size} is an integer. @var{file} is a stdio
4600 stream to which the assembler code should be output.
4602 The label for the beginning of the function need not be output by this
4603 macro. That has already been done when the macro is run.
4605 @findex regs_ever_live
4606 To determine which registers to save, the macro can refer to the array
4607 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4608 @var{r} is used anywhere within the function. This implies the function
4609 prologue should save register @var{r}, provided it is not one of the
4610 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4611 @code{regs_ever_live}.)
4613 On machines that have ``register windows'', the function entry code does
4614 not save on the stack the registers that are in the windows, even if
4615 they are supposed to be preserved by function calls; instead it takes
4616 appropriate steps to ``push'' the register stack, if any non-call-used
4617 registers are used in the function.
4619 @findex frame_pointer_needed
4620 On machines where functions may or may not have frame-pointers, the
4621 function entry code must vary accordingly; it must set up the frame
4622 pointer if one is wanted, and not otherwise. To determine whether a
4623 frame pointer is in wanted, the macro can refer to the variable
4624 @code{frame_pointer_needed}. The variable's value will be 1 at run
4625 time in a function that needs a frame pointer. @xref{Elimination}.
4627 The function entry code is responsible for allocating any stack space
4628 required for the function. This stack space consists of the regions
4629 listed below. In most cases, these regions are allocated in the
4630 order listed, with the last listed region closest to the top of the
4631 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4632 the highest address if it is not defined). You can use a different order
4633 for a machine if doing so is more convenient or required for
4634 compatibility reasons. Except in cases where required by standard
4635 or by a debugger, there is no reason why the stack layout used by GCC
4636 need agree with that used by other compilers for a machine.
4639 @hook TARGET_ASM_FUNCTION_END_PROLOGUE
4640 If defined, a function that outputs assembler code at the end of a
4641 prologue. This should be used when the function prologue is being
4642 emitted as RTL, and you have some extra assembler that needs to be
4643 emitted. @xref{prologue instruction pattern}.
4646 @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
4647 If defined, a function that outputs assembler code at the start of an
4648 epilogue. This should be used when the function epilogue is being
4649 emitted as RTL, and you have some extra assembler that needs to be
4650 emitted. @xref{epilogue instruction pattern}.
4653 @hook TARGET_ASM_FUNCTION_EPILOGUE
4654 If defined, a function that outputs the assembler code for exit from a
4655 function. The epilogue is responsible for restoring the saved
4656 registers and stack pointer to their values when the function was
4657 called, and returning control to the caller. This macro takes the
4658 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4659 registers to restore are determined from @code{regs_ever_live} and
4660 @code{CALL_USED_REGISTERS} in the same way.
4662 On some machines, there is a single instruction that does all the work
4663 of returning from the function. On these machines, give that
4664 instruction the name @samp{return} and do not define the macro
4665 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4667 Do not define a pattern named @samp{return} if you want the
4668 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4669 switches to control whether return instructions or epilogues are used,
4670 define a @samp{return} pattern with a validity condition that tests the
4671 target switches appropriately. If the @samp{return} pattern's validity
4672 condition is false, epilogues will be used.
4674 On machines where functions may or may not have frame-pointers, the
4675 function exit code must vary accordingly. Sometimes the code for these
4676 two cases is completely different. To determine whether a frame pointer
4677 is wanted, the macro can refer to the variable
4678 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4679 a function that needs a frame pointer.
4681 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4682 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4683 The C variable @code{current_function_is_leaf} is nonzero for such a
4684 function. @xref{Leaf Functions}.
4686 On some machines, some functions pop their arguments on exit while
4687 others leave that for the caller to do. For example, the 68020 when
4688 given @option{-mrtd} pops arguments in functions that take a fixed
4689 number of arguments.
4691 @findex current_function_pops_args
4692 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4693 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4694 needs to know what was decided. The number of bytes of the current
4695 function's arguments that this function should pop is available in
4696 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4701 @findex current_function_pretend_args_size
4702 A region of @code{current_function_pretend_args_size} bytes of
4703 uninitialized space just underneath the first argument arriving on the
4704 stack. (This may not be at the very start of the allocated stack region
4705 if the calling sequence has pushed anything else since pushing the stack
4706 arguments. But usually, on such machines, nothing else has been pushed
4707 yet, because the function prologue itself does all the pushing.) This
4708 region is used on machines where an argument may be passed partly in
4709 registers and partly in memory, and, in some cases to support the
4710 features in @code{<stdarg.h>}.
4713 An area of memory used to save certain registers used by the function.
4714 The size of this area, which may also include space for such things as
4715 the return address and pointers to previous stack frames, is
4716 machine-specific and usually depends on which registers have been used
4717 in the function. Machines with register windows often do not require
4721 A region of at least @var{size} bytes, possibly rounded up to an allocation
4722 boundary, to contain the local variables of the function. On some machines,
4723 this region and the save area may occur in the opposite order, with the
4724 save area closer to the top of the stack.
4727 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4728 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4729 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4730 argument lists of the function. @xref{Stack Arguments}.
4733 @defmac EXIT_IGNORE_STACK
4734 Define this macro as a C expression that is nonzero if the return
4735 instruction or the function epilogue ignores the value of the stack
4736 pointer; in other words, if it is safe to delete an instruction to
4737 adjust the stack pointer before a return from the function. The
4740 Note that this macro's value is relevant only for functions for which
4741 frame pointers are maintained. It is never safe to delete a final
4742 stack adjustment in a function that has no frame pointer, and the
4743 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4746 @defmac EPILOGUE_USES (@var{regno})
4747 Define this macro as a C expression that is nonzero for registers that are
4748 used by the epilogue or the @samp{return} pattern. The stack and frame
4749 pointer registers are already assumed to be used as needed.
4752 @defmac EH_USES (@var{regno})
4753 Define this macro as a C expression that is nonzero for registers that are
4754 used by the exception handling mechanism, and so should be considered live
4755 on entry to an exception edge.
4758 @defmac DELAY_SLOTS_FOR_EPILOGUE
4759 Define this macro if the function epilogue contains delay slots to which
4760 instructions from the rest of the function can be ``moved''. The
4761 definition should be a C expression whose value is an integer
4762 representing the number of delay slots there.
4765 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4766 A C expression that returns 1 if @var{insn} can be placed in delay
4767 slot number @var{n} of the epilogue.
4769 The argument @var{n} is an integer which identifies the delay slot now
4770 being considered (since different slots may have different rules of
4771 eligibility). It is never negative and is always less than the number
4772 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4773 If you reject a particular insn for a given delay slot, in principle, it
4774 may be reconsidered for a subsequent delay slot. Also, other insns may
4775 (at least in principle) be considered for the so far unfilled delay
4778 @findex current_function_epilogue_delay_list
4779 @findex final_scan_insn
4780 The insns accepted to fill the epilogue delay slots are put in an RTL
4781 list made with @code{insn_list} objects, stored in the variable
4782 @code{current_function_epilogue_delay_list}. The insn for the first
4783 delay slot comes first in the list. Your definition of the macro
4784 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4785 outputting the insns in this list, usually by calling
4786 @code{final_scan_insn}.
4788 You need not define this macro if you did not define
4789 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4792 @hook TARGET_ASM_OUTPUT_MI_THUNK
4793 A function that outputs the assembler code for a thunk
4794 function, used to implement C++ virtual function calls with multiple
4795 inheritance. The thunk acts as a wrapper around a virtual function,
4796 adjusting the implicit object parameter before handing control off to
4799 First, emit code to add the integer @var{delta} to the location that
4800 contains the incoming first argument. Assume that this argument
4801 contains a pointer, and is the one used to pass the @code{this} pointer
4802 in C++. This is the incoming argument @emph{before} the function prologue,
4803 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4804 all other incoming arguments.
4806 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4807 made after adding @code{delta}. In particular, if @var{p} is the
4808 adjusted pointer, the following adjustment should be made:
4811 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4814 After the additions, emit code to jump to @var{function}, which is a
4815 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4816 not touch the return address. Hence returning from @var{FUNCTION} will
4817 return to whoever called the current @samp{thunk}.
4819 The effect must be as if @var{function} had been called directly with
4820 the adjusted first argument. This macro is responsible for emitting all
4821 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4822 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4824 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4825 have already been extracted from it.) It might possibly be useful on
4826 some targets, but probably not.
4828 If you do not define this macro, the target-independent code in the C++
4829 front end will generate a less efficient heavyweight thunk that calls
4830 @var{function} instead of jumping to it. The generic approach does
4831 not support varargs.
4834 @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
4835 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4836 to output the assembler code for the thunk function specified by the
4837 arguments it is passed, and false otherwise. In the latter case, the
4838 generic approach will be used by the C++ front end, with the limitations
4843 @subsection Generating Code for Profiling
4844 @cindex profiling, code generation
4846 These macros will help you generate code for profiling.
4848 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4849 A C statement or compound statement to output to @var{file} some
4850 assembler code to call the profiling subroutine @code{mcount}.
4853 The details of how @code{mcount} expects to be called are determined by
4854 your operating system environment, not by GCC@. To figure them out,
4855 compile a small program for profiling using the system's installed C
4856 compiler and look at the assembler code that results.
4858 Older implementations of @code{mcount} expect the address of a counter
4859 variable to be loaded into some register. The name of this variable is
4860 @samp{LP} followed by the number @var{labelno}, so you would generate
4861 the name using @samp{LP%d} in a @code{fprintf}.
4864 @defmac PROFILE_HOOK
4865 A C statement or compound statement to output to @var{file} some assembly
4866 code to call the profiling subroutine @code{mcount} even the target does
4867 not support profiling.
4870 @defmac NO_PROFILE_COUNTERS
4871 Define this macro to be an expression with a nonzero value if the
4872 @code{mcount} subroutine on your system does not need a counter variable
4873 allocated for each function. This is true for almost all modern
4874 implementations. If you define this macro, you must not use the
4875 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4878 @defmac PROFILE_BEFORE_PROLOGUE
4879 Define this macro if the code for function profiling should come before
4880 the function prologue. Normally, the profiling code comes after.
4884 @subsection Permitting tail calls
4887 @hook TARGET_FUNCTION_OK_FOR_SIBCALL
4888 True if it is ok to do sibling call optimization for the specified
4889 call expression @var{exp}. @var{decl} will be the called function,
4890 or @code{NULL} if this is an indirect call.
4892 It is not uncommon for limitations of calling conventions to prevent
4893 tail calls to functions outside the current unit of translation, or
4894 during PIC compilation. The hook is used to enforce these restrictions,
4895 as the @code{sibcall} md pattern can not fail, or fall over to a
4896 ``normal'' call. The criteria for successful sibling call optimization
4897 may vary greatly between different architectures.
4900 @hook TARGET_EXTRA_LIVE_ON_ENTRY
4901 Add any hard registers to @var{regs} that are live on entry to the
4902 function. This hook only needs to be defined to provide registers that
4903 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4904 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4905 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4906 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4909 @node Stack Smashing Protection
4910 @subsection Stack smashing protection
4911 @cindex stack smashing protection
4913 @hook TARGET_STACK_PROTECT_GUARD
4914 This hook returns a @code{DECL} node for the external variable to use
4915 for the stack protection guard. This variable is initialized by the
4916 runtime to some random value and is used to initialize the guard value
4917 that is placed at the top of the local stack frame. The type of this
4918 variable must be @code{ptr_type_node}.
4920 The default version of this hook creates a variable called
4921 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4924 @hook TARGET_STACK_PROTECT_FAIL
4925 This hook returns a tree expression that alerts the runtime that the
4926 stack protect guard variable has been modified. This expression should
4927 involve a call to a @code{noreturn} function.
4929 The default version of this hook invokes a function called
4930 @samp{__stack_chk_fail}, taking no arguments. This function is
4931 normally defined in @file{libgcc2.c}.
4934 @hook TARGET_SUPPORTS_SPLIT_STACK
4937 @section Implementing the Varargs Macros
4938 @cindex varargs implementation
4940 GCC comes with an implementation of @code{<varargs.h>} and
4941 @code{<stdarg.h>} that work without change on machines that pass arguments
4942 on the stack. Other machines require their own implementations of
4943 varargs, and the two machine independent header files must have
4944 conditionals to include it.
4946 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4947 the calling convention for @code{va_start}. The traditional
4948 implementation takes just one argument, which is the variable in which
4949 to store the argument pointer. The ISO implementation of
4950 @code{va_start} takes an additional second argument. The user is
4951 supposed to write the last named argument of the function here.
4953 However, @code{va_start} should not use this argument. The way to find
4954 the end of the named arguments is with the built-in functions described
4957 @defmac __builtin_saveregs ()
4958 Use this built-in function to save the argument registers in memory so
4959 that the varargs mechanism can access them. Both ISO and traditional
4960 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4961 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4963 On some machines, @code{__builtin_saveregs} is open-coded under the
4964 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4965 other machines, it calls a routine written in assembler language,
4966 found in @file{libgcc2.c}.
4968 Code generated for the call to @code{__builtin_saveregs} appears at the
4969 beginning of the function, as opposed to where the call to
4970 @code{__builtin_saveregs} is written, regardless of what the code is.
4971 This is because the registers must be saved before the function starts
4972 to use them for its own purposes.
4973 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4977 @defmac __builtin_next_arg (@var{lastarg})
4978 This builtin returns the address of the first anonymous stack
4979 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4980 returns the address of the location above the first anonymous stack
4981 argument. Use it in @code{va_start} to initialize the pointer for
4982 fetching arguments from the stack. Also use it in @code{va_start} to
4983 verify that the second parameter @var{lastarg} is the last named argument
4984 of the current function.
4987 @defmac __builtin_classify_type (@var{object})
4988 Since each machine has its own conventions for which data types are
4989 passed in which kind of register, your implementation of @code{va_arg}
4990 has to embody these conventions. The easiest way to categorize the
4991 specified data type is to use @code{__builtin_classify_type} together
4992 with @code{sizeof} and @code{__alignof__}.
4994 @code{__builtin_classify_type} ignores the value of @var{object},
4995 considering only its data type. It returns an integer describing what
4996 kind of type that is---integer, floating, pointer, structure, and so on.
4998 The file @file{typeclass.h} defines an enumeration that you can use to
4999 interpret the values of @code{__builtin_classify_type}.
5002 These machine description macros help implement varargs:
5004 @hook TARGET_EXPAND_BUILTIN_SAVEREGS
5005 If defined, this hook produces the machine-specific code for a call to
5006 @code{__builtin_saveregs}. This code will be moved to the very
5007 beginning of the function, before any parameter access are made. The
5008 return value of this function should be an RTX that contains the value
5009 to use as the return of @code{__builtin_saveregs}.
5012 @hook TARGET_SETUP_INCOMING_VARARGS
5013 This target hook offers an alternative to using
5014 @code{__builtin_saveregs} and defining the hook
5015 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5016 register arguments into the stack so that all the arguments appear to
5017 have been passed consecutively on the stack. Once this is done, you can
5018 use the standard implementation of varargs that works for machines that
5019 pass all their arguments on the stack.
5021 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5022 structure, containing the values that are obtained after processing the
5023 named arguments. The arguments @var{mode} and @var{type} describe the
5024 last named argument---its machine mode and its data type as a tree node.
5026 The target hook should do two things: first, push onto the stack all the
5027 argument registers @emph{not} used for the named arguments, and second,
5028 store the size of the data thus pushed into the @code{int}-valued
5029 variable pointed to by @var{pretend_args_size}. The value that you
5030 store here will serve as additional offset for setting up the stack
5033 Because you must generate code to push the anonymous arguments at
5034 compile time without knowing their data types,
5035 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5036 have just a single category of argument register and use it uniformly
5039 If the argument @var{second_time} is nonzero, it means that the
5040 arguments of the function are being analyzed for the second time. This
5041 happens for an inline function, which is not actually compiled until the
5042 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5043 not generate any instructions in this case.
5046 @hook TARGET_STRICT_ARGUMENT_NAMING
5047 Define this hook to return @code{true} if the location where a function
5048 argument is passed depends on whether or not it is a named argument.
5050 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5051 is set for varargs and stdarg functions. If this hook returns
5052 @code{true}, the @var{named} argument is always true for named
5053 arguments, and false for unnamed arguments. If it returns @code{false},
5054 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5055 then all arguments are treated as named. Otherwise, all named arguments
5056 except the last are treated as named.
5058 You need not define this hook if it always returns @code{false}.
5061 @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
5062 If you need to conditionally change ABIs so that one works with
5063 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5064 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5065 defined, then define this hook to return @code{true} if
5066 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5067 Otherwise, you should not define this hook.
5071 @section Trampolines for Nested Functions
5072 @cindex trampolines for nested functions
5073 @cindex nested functions, trampolines for
5075 A @dfn{trampoline} is a small piece of code that is created at run time
5076 when the address of a nested function is taken. It normally resides on
5077 the stack, in the stack frame of the containing function. These macros
5078 tell GCC how to generate code to allocate and initialize a
5081 The instructions in the trampoline must do two things: load a constant
5082 address into the static chain register, and jump to the real address of
5083 the nested function. On CISC machines such as the m68k, this requires
5084 two instructions, a move immediate and a jump. Then the two addresses
5085 exist in the trampoline as word-long immediate operands. On RISC
5086 machines, it is often necessary to load each address into a register in
5087 two parts. Then pieces of each address form separate immediate
5090 The code generated to initialize the trampoline must store the variable
5091 parts---the static chain value and the function address---into the
5092 immediate operands of the instructions. On a CISC machine, this is
5093 simply a matter of copying each address to a memory reference at the
5094 proper offset from the start of the trampoline. On a RISC machine, it
5095 may be necessary to take out pieces of the address and store them
5098 @hook TARGET_ASM_TRAMPOLINE_TEMPLATE
5099 This hook is called by @code{assemble_trampoline_template} to output,
5100 on the stream @var{f}, assembler code for a block of data that contains
5101 the constant parts of a trampoline. This code should not include a
5102 label---the label is taken care of automatically.
5104 If you do not define this hook, it means no template is needed
5105 for the target. Do not define this hook on systems where the block move
5106 code to copy the trampoline into place would be larger than the code
5107 to generate it on the spot.
5110 @defmac TRAMPOLINE_SECTION
5111 Return the section into which the trampoline template is to be placed
5112 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5115 @defmac TRAMPOLINE_SIZE
5116 A C expression for the size in bytes of the trampoline, as an integer.
5119 @defmac TRAMPOLINE_ALIGNMENT
5120 Alignment required for trampolines, in bits.
5122 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5123 is used for aligning trampolines.
5126 @hook TARGET_TRAMPOLINE_INIT
5127 This hook is called to initialize a trampoline.
5128 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5129 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5130 RTX for the static chain value that should be passed to the function
5133 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5134 first thing this hook should do is emit a block move into @var{m_tramp}
5135 from the memory block returned by @code{assemble_trampoline_template}.
5136 Note that the block move need only cover the constant parts of the
5137 trampoline. If the target isolates the variable parts of the trampoline
5138 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5140 If the target requires any other actions, such as flushing caches or
5141 enabling stack execution, these actions should be performed after
5142 initializing the trampoline proper.
5145 @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
5146 This hook should perform any machine-specific adjustment in
5147 the address of the trampoline. Its argument contains the address of the
5148 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5149 the address to be used for a function call should be different from the
5150 address at which the template was stored, the different address should
5151 be returned; otherwise @var{addr} should be returned unchanged.
5152 If this hook is not defined, @var{addr} will be used for function calls.
5155 Implementing trampolines is difficult on many machines because they have
5156 separate instruction and data caches. Writing into a stack location
5157 fails to clear the memory in the instruction cache, so when the program
5158 jumps to that location, it executes the old contents.
5160 Here are two possible solutions. One is to clear the relevant parts of
5161 the instruction cache whenever a trampoline is set up. The other is to
5162 make all trampolines identical, by having them jump to a standard
5163 subroutine. The former technique makes trampoline execution faster; the
5164 latter makes initialization faster.
5166 To clear the instruction cache when a trampoline is initialized, define
5167 the following macro.
5169 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5170 If defined, expands to a C expression clearing the @emph{instruction
5171 cache} in the specified interval. The definition of this macro would
5172 typically be a series of @code{asm} statements. Both @var{beg} and
5173 @var{end} are both pointer expressions.
5176 To use a standard subroutine, define the following macro. In addition,
5177 you must make sure that the instructions in a trampoline fill an entire
5178 cache line with identical instructions, or else ensure that the
5179 beginning of the trampoline code is always aligned at the same point in
5180 its cache line. Look in @file{m68k.h} as a guide.
5182 @defmac TRANSFER_FROM_TRAMPOLINE
5183 Define this macro if trampolines need a special subroutine to do their
5184 work. The macro should expand to a series of @code{asm} statements
5185 which will be compiled with GCC@. They go in a library function named
5186 @code{__transfer_from_trampoline}.
5188 If you need to avoid executing the ordinary prologue code of a compiled
5189 C function when you jump to the subroutine, you can do so by placing a
5190 special label of your own in the assembler code. Use one @code{asm}
5191 statement to generate an assembler label, and another to make the label
5192 global. Then trampolines can use that label to jump directly to your
5193 special assembler code.
5197 @section Implicit Calls to Library Routines
5198 @cindex library subroutine names
5199 @cindex @file{libgcc.a}
5201 @c prevent bad page break with this line
5202 Here is an explanation of implicit calls to library routines.
5204 @defmac DECLARE_LIBRARY_RENAMES
5205 This macro, if defined, should expand to a piece of C code that will get
5206 expanded when compiling functions for libgcc.a. It can be used to
5207 provide alternate names for GCC's internal library functions if there
5208 are ABI-mandated names that the compiler should provide.
5211 @findex set_optab_libfunc
5212 @findex init_one_libfunc
5213 @hook TARGET_INIT_LIBFUNCS
5214 This hook should declare additional library routines or rename
5215 existing ones, using the functions @code{set_optab_libfunc} and
5216 @code{init_one_libfunc} defined in @file{optabs.c}.
5217 @code{init_optabs} calls this macro after initializing all the normal
5220 The default is to do nothing. Most ports don't need to define this hook.
5223 @hook TARGET_LIBFUNC_GNU_PREFIX
5225 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5226 This macro should return @code{true} if the library routine that
5227 implements the floating point comparison operator @var{comparison} in
5228 mode @var{mode} will return a boolean, and @var{false} if it will
5231 GCC's own floating point libraries return tristates from the
5232 comparison operators, so the default returns false always. Most ports
5233 don't need to define this macro.
5236 @defmac TARGET_LIB_INT_CMP_BIASED
5237 This macro should evaluate to @code{true} if the integer comparison
5238 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5239 operand is smaller than the second, 1 to indicate that they are equal,
5240 and 2 to indicate that the first operand is greater than the second.
5241 If this macro evaluates to @code{false} the comparison functions return
5242 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5243 in @file{libgcc.a}, you do not need to define this macro.
5246 @cindex @code{EDOM}, implicit usage
5249 The value of @code{EDOM} on the target machine, as a C integer constant
5250 expression. If you don't define this macro, GCC does not attempt to
5251 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5252 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5255 If you do not define @code{TARGET_EDOM}, then compiled code reports
5256 domain errors by calling the library function and letting it report the
5257 error. If mathematical functions on your system use @code{matherr} when
5258 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5259 that @code{matherr} is used normally.
5262 @cindex @code{errno}, implicit usage
5263 @defmac GEN_ERRNO_RTX
5264 Define this macro as a C expression to create an rtl expression that
5265 refers to the global ``variable'' @code{errno}. (On certain systems,
5266 @code{errno} may not actually be a variable.) If you don't define this
5267 macro, a reasonable default is used.
5270 @cindex C99 math functions, implicit usage
5271 @defmac TARGET_C99_FUNCTIONS
5272 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5273 @code{sinf} and similarly for other functions defined by C99 standard. The
5274 default is zero because a number of existing systems lack support for these
5275 functions in their runtime so this macro needs to be redefined to one on
5276 systems that do support the C99 runtime.
5279 @cindex sincos math function, implicit usage
5280 @defmac TARGET_HAS_SINCOS
5281 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5282 and @code{cos} with the same argument to a call to @code{sincos}. The
5283 default is zero. The target has to provide the following functions:
5285 void sincos(double x, double *sin, double *cos);
5286 void sincosf(float x, float *sin, float *cos);
5287 void sincosl(long double x, long double *sin, long double *cos);
5291 @defmac NEXT_OBJC_RUNTIME
5292 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
5293 by default. This calling convention involves passing the object, the selector
5294 and the method arguments all at once to the method-lookup library function.
5295 This is the usual setting when targeting Darwin/Mac OS X systems, which have
5296 the NeXT runtime installed.
5298 If the macro is set to 0, the "GNU" Objective-C message sending convention
5299 will be used by default. This convention passes just the object and the
5300 selector to the method-lookup function, which returns a pointer to the method.
5302 In either case, it remains possible to select code-generation for the alternate
5303 scheme, by means of compiler command line switches.
5306 @node Addressing Modes
5307 @section Addressing Modes
5308 @cindex addressing modes
5310 @c prevent bad page break with this line
5311 This is about addressing modes.
5313 @defmac HAVE_PRE_INCREMENT
5314 @defmacx HAVE_PRE_DECREMENT
5315 @defmacx HAVE_POST_INCREMENT
5316 @defmacx HAVE_POST_DECREMENT
5317 A C expression that is nonzero if the machine supports pre-increment,
5318 pre-decrement, post-increment, or post-decrement addressing respectively.
5321 @defmac HAVE_PRE_MODIFY_DISP
5322 @defmacx HAVE_POST_MODIFY_DISP
5323 A C expression that is nonzero if the machine supports pre- or
5324 post-address side-effect generation involving constants other than
5325 the size of the memory operand.
5328 @defmac HAVE_PRE_MODIFY_REG
5329 @defmacx HAVE_POST_MODIFY_REG
5330 A C expression that is nonzero if the machine supports pre- or
5331 post-address side-effect generation involving a register displacement.
5334 @defmac CONSTANT_ADDRESS_P (@var{x})
5335 A C expression that is 1 if the RTX @var{x} is a constant which
5336 is a valid address. On most machines the default definition of
5337 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5338 is acceptable, but a few machines are more restrictive as to which
5339 constant addresses are supported.
5342 @defmac CONSTANT_P (@var{x})
5343 @code{CONSTANT_P}, which is defined by target-independent code,
5344 accepts integer-values expressions whose values are not explicitly
5345 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5346 expressions and @code{const} arithmetic expressions, in addition to
5347 @code{const_int} and @code{const_double} expressions.
5350 @defmac MAX_REGS_PER_ADDRESS
5351 A number, the maximum number of registers that can appear in a valid
5352 memory address. Note that it is up to you to specify a value equal to
5353 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5357 @hook TARGET_LEGITIMATE_ADDRESS_P
5358 A function that returns whether @var{x} (an RTX) is a legitimate memory
5359 address on the target machine for a memory operand of mode @var{mode}.
5361 Legitimate addresses are defined in two variants: a strict variant and a
5362 non-strict one. The @var{strict} parameter chooses which variant is
5363 desired by the caller.
5365 The strict variant is used in the reload pass. It must be defined so
5366 that any pseudo-register that has not been allocated a hard register is
5367 considered a memory reference. This is because in contexts where some
5368 kind of register is required, a pseudo-register with no hard register
5369 must be rejected. For non-hard registers, the strict variant should look
5370 up the @code{reg_renumber} array; it should then proceed using the hard
5371 register number in the array, or treat the pseudo as a memory reference
5372 if the array holds @code{-1}.
5374 The non-strict variant is used in other passes. It must be defined to
5375 accept all pseudo-registers in every context where some kind of
5376 register is required.
5378 Normally, constant addresses which are the sum of a @code{symbol_ref}
5379 and an integer are stored inside a @code{const} RTX to mark them as
5380 constant. Therefore, there is no need to recognize such sums
5381 specifically as legitimate addresses. Normally you would simply
5382 recognize any @code{const} as legitimate.
5384 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5385 sums that are not marked with @code{const}. It assumes that a naked
5386 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5387 naked constant sums as illegitimate addresses, so that none of them will
5388 be given to @code{PRINT_OPERAND_ADDRESS}.
5390 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5391 On some machines, whether a symbolic address is legitimate depends on
5392 the section that the address refers to. On these machines, define the
5393 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5394 into the @code{symbol_ref}, and then check for it here. When you see a
5395 @code{const}, you will have to look inside it to find the
5396 @code{symbol_ref} in order to determine the section. @xref{Assembler
5399 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5400 Some ports are still using a deprecated legacy substitute for
5401 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5405 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5409 and should @code{goto @var{label}} if the address @var{x} is a valid
5410 address on the target machine for a memory operand of mode @var{mode}.
5412 @findex REG_OK_STRICT
5413 Compiler source files that want to use the strict variant of this
5414 macro define the macro @code{REG_OK_STRICT}. You should use an
5415 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5416 that case and the non-strict variant otherwise.
5418 Using the hook is usually simpler because it limits the number of
5419 files that are recompiled when changes are made.
5422 @defmac TARGET_MEM_CONSTRAINT
5423 A single character to be used instead of the default @code{'m'}
5424 character for general memory addresses. This defines the constraint
5425 letter which matches the memory addresses accepted by
5426 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5427 support new address formats in your back end without changing the
5428 semantics of the @code{'m'} constraint. This is necessary in order to
5429 preserve functionality of inline assembly constructs using the
5430 @code{'m'} constraint.
5433 @defmac FIND_BASE_TERM (@var{x})
5434 A C expression to determine the base term of address @var{x},
5435 or to provide a simplified version of @var{x} from which @file{alias.c}
5436 can easily find the base term. This macro is used in only two places:
5437 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5439 It is always safe for this macro to not be defined. It exists so
5440 that alias analysis can understand machine-dependent addresses.
5442 The typical use of this macro is to handle addresses containing
5443 a label_ref or symbol_ref within an UNSPEC@.
5446 @hook TARGET_LEGITIMIZE_ADDRESS
5447 This hook is given an invalid memory address @var{x} for an
5448 operand of mode @var{mode} and should try to return a valid memory
5451 @findex break_out_memory_refs
5452 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5453 and @var{oldx} will be the operand that was given to that function to produce
5456 The code of the hook should not alter the substructure of
5457 @var{x}. If it transforms @var{x} into a more legitimate form, it
5458 should return the new @var{x}.
5460 It is not necessary for this hook to come up with a legitimate address.
5461 The compiler has standard ways of doing so in all cases. In fact, it
5462 is safe to omit this hook or make it return @var{x} if it cannot find
5463 a valid way to legitimize the address. But often a machine-dependent
5464 strategy can generate better code.
5467 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5468 A C compound statement that attempts to replace @var{x}, which is an address
5469 that needs reloading, with a valid memory address for an operand of mode
5470 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5471 It is not necessary to define this macro, but it might be useful for
5472 performance reasons.
5474 For example, on the i386, it is sometimes possible to use a single
5475 reload register instead of two by reloading a sum of two pseudo
5476 registers into a register. On the other hand, for number of RISC
5477 processors offsets are limited so that often an intermediate address
5478 needs to be generated in order to address a stack slot. By defining
5479 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5480 generated for adjacent some stack slots can be made identical, and thus
5483 @emph{Note}: This macro should be used with caution. It is necessary
5484 to know something of how reload works in order to effectively use this,
5485 and it is quite easy to produce macros that build in too much knowledge
5486 of reload internals.
5488 @emph{Note}: This macro must be able to reload an address created by a
5489 previous invocation of this macro. If it fails to handle such addresses
5490 then the compiler may generate incorrect code or abort.
5493 The macro definition should use @code{push_reload} to indicate parts that
5494 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5495 suitable to be passed unaltered to @code{push_reload}.
5497 The code generated by this macro must not alter the substructure of
5498 @var{x}. If it transforms @var{x} into a more legitimate form, it
5499 should assign @var{x} (which will always be a C variable) a new value.
5500 This also applies to parts that you change indirectly by calling
5503 @findex strict_memory_address_p
5504 The macro definition may use @code{strict_memory_address_p} to test if
5505 the address has become legitimate.
5508 If you want to change only a part of @var{x}, one standard way of doing
5509 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5510 single level of rtl. Thus, if the part to be changed is not at the
5511 top level, you'll need to replace first the top level.
5512 It is not necessary for this macro to come up with a legitimate
5513 address; but often a machine-dependent strategy can generate better code.
5516 @hook TARGET_MODE_DEPENDENT_ADDRESS_P
5517 This hook returns @code{true} if memory address @var{addr} can have
5518 different meanings depending on the machine mode of the memory
5519 reference it is used for or if the address is valid for some modes
5522 Autoincrement and autodecrement addresses typically have mode-dependent
5523 effects because the amount of the increment or decrement is the size
5524 of the operand being addressed. Some machines have other mode-dependent
5525 addresses. Many RISC machines have no mode-dependent addresses.
5527 You may assume that @var{addr} is a valid address for the machine.
5529 The default version of this hook returns @code{false}.
5532 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5533 A C statement or compound statement with a conditional @code{goto
5534 @var{label};} executed if memory address @var{x} (an RTX) can have
5535 different meanings depending on the machine mode of the memory
5536 reference it is used for or if the address is valid for some modes
5539 Autoincrement and autodecrement addresses typically have mode-dependent
5540 effects because the amount of the increment or decrement is the size
5541 of the operand being addressed. Some machines have other mode-dependent
5542 addresses. Many RISC machines have no mode-dependent addresses.
5544 You may assume that @var{addr} is a valid address for the machine.
5546 These are obsolete macros, replaced by the
5547 @code{TARGET_MODE_DEPENDENT_ADDRESS_P} target hook.
5550 @hook TARGET_LEGITIMATE_CONSTANT_P
5551 This hook returns true if @var{x} is a legitimate constant for a
5552 @var{mode}-mode immediate operand on the target machine. You can assume that
5553 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5555 The default definition returns true.
5558 @hook TARGET_DELEGITIMIZE_ADDRESS
5559 This hook is used to undo the possibly obfuscating effects of the
5560 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5561 macros. Some backend implementations of these macros wrap symbol
5562 references inside an @code{UNSPEC} rtx to represent PIC or similar
5563 addressing modes. This target hook allows GCC's optimizers to understand
5564 the semantics of these opaque @code{UNSPEC}s by converting them back
5565 into their original form.
5568 @hook TARGET_CANNOT_FORCE_CONST_MEM
5569 This hook should return true if @var{x} is of a form that cannot (or
5570 should not) be spilled to the constant pool. @var{mode} is the mode
5573 The default version of this hook returns false.
5575 The primary reason to define this hook is to prevent reload from
5576 deciding that a non-legitimate constant would be better reloaded
5577 from the constant pool instead of spilling and reloading a register
5578 holding the constant. This restriction is often true of addresses
5579 of TLS symbols for various targets.
5582 @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
5583 This hook should return true if pool entries for constant @var{x} can
5584 be placed in an @code{object_block} structure. @var{mode} is the mode
5587 The default version returns false for all constants.
5590 @hook TARGET_BUILTIN_RECIPROCAL
5591 This hook should return the DECL of a function that implements reciprocal of
5592 the builtin function with builtin function code @var{fn}, or
5593 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5594 when @var{fn} is a code of a machine-dependent builtin function. When
5595 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5596 of a square root function are performed, and only reciprocals of @code{sqrt}
5600 @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
5601 This hook should return the DECL of a function @var{f} that given an
5602 address @var{addr} as an argument returns a mask @var{m} that can be
5603 used to extract from two vectors the relevant data that resides in
5604 @var{addr} in case @var{addr} is not properly aligned.
5606 The autovectorizer, when vectorizing a load operation from an address
5607 @var{addr} that may be unaligned, will generate two vector loads from
5608 the two aligned addresses around @var{addr}. It then generates a
5609 @code{REALIGN_LOAD} operation to extract the relevant data from the
5610 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5611 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5612 the third argument, @var{OFF}, defines how the data will be extracted
5613 from these two vectors: if @var{OFF} is 0, then the returned vector is
5614 @var{v2}; otherwise, the returned vector is composed from the last
5615 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5616 @var{OFF} elements of @var{v2}.
5618 If this hook is defined, the autovectorizer will generate a call
5619 to @var{f} (using the DECL tree that this hook returns) and will
5620 use the return value of @var{f} as the argument @var{OFF} to
5621 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5622 should comply with the semantics expected by @code{REALIGN_LOAD}
5624 If this hook is not defined, then @var{addr} will be used as
5625 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5626 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5629 @hook TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN
5630 This hook should return the DECL of a function @var{f} that implements
5631 widening multiplication of the even elements of two input vectors of type @var{x}.
5633 If this hook is defined, the autovectorizer will use it along with the
5634 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5635 widening multiplication in cases that the order of the results does not have to be
5636 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5637 @code{widen_mult_hi/lo} idioms will be used.
5640 @hook TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD
5641 This hook should return the DECL of a function @var{f} that implements
5642 widening multiplication of the odd elements of two input vectors of type @var{x}.
5644 If this hook is defined, the autovectorizer will use it along with the
5645 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5646 widening multiplication in cases that the order of the results does not have to be
5647 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5648 @code{widen_mult_hi/lo} idioms will be used.
5651 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
5652 Returns cost of different scalar or vector statements for vectorization cost model.
5653 For vector memory operations the cost may depend on type (@var{vectype}) and
5654 misalignment value (@var{misalign}).
5657 @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
5658 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5661 @hook TARGET_VECTORIZE_VEC_PERM_CONST_OK
5662 Return true if a vector created for @code{vec_perm_const} is valid.
5665 @hook TARGET_VECTORIZE_BUILTIN_CONVERSION
5666 This hook should return the DECL of a function that implements conversion of the
5667 input vector of type @var{src_type} to type @var{dest_type}.
5668 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5669 specifies how the conversion is to be applied
5670 (truncation, rounding, etc.).
5672 If this hook is defined, the autovectorizer will use the
5673 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5674 conversion. Otherwise, it will return @code{NULL_TREE}.
5677 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
5678 This hook should return the decl of a function that implements the
5679 vectorized variant of the builtin function with builtin function code
5680 @var{code} or @code{NULL_TREE} if such a function is not available.
5681 The value of @var{fndecl} is the builtin function declaration. The
5682 return type of the vectorized function shall be of vector type
5683 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5686 @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
5687 This hook should return true if the target supports misaligned vector
5688 store/load of a specific factor denoted in the @var{misalignment}
5689 parameter. The vector store/load should be of machine mode @var{mode} and
5690 the elements in the vectors should be of type @var{type}. @var{is_packed}
5691 parameter is true if the memory access is defined in a packed struct.
5694 @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
5695 This hook should return the preferred mode for vectorizing scalar
5696 mode @var{mode}. The default is
5697 equal to @code{word_mode}, because the vectorizer can do some
5698 transformations even in absence of specialized @acronym{SIMD} hardware.
5701 @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES
5702 This hook should return a mask of sizes that should be iterated over
5703 after trying to autovectorize using the vector size derived from the
5704 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5705 The default is zero which means to not iterate over other vector sizes.
5708 @hook TARGET_VECTORIZE_BUILTIN_TM_LOAD
5710 @hook TARGET_VECTORIZE_BUILTIN_TM_STORE
5712 @hook TARGET_VECTORIZE_BUILTIN_GATHER
5713 Target builtin that implements vector gather operation. @var{mem_vectype}
5714 is the vector type of the load and @var{index_type} is scalar type of
5715 the index, scaled by @var{scale}.
5716 The default is @code{NULL_TREE} which means to not vectorize gather
5720 @node Anchored Addresses
5721 @section Anchored Addresses
5722 @cindex anchored addresses
5723 @cindex @option{-fsection-anchors}
5725 GCC usually addresses every static object as a separate entity.
5726 For example, if we have:
5730 int foo (void) @{ return a + b + c; @}
5733 the code for @code{foo} will usually calculate three separate symbolic
5734 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5735 it would be better to calculate just one symbolic address and access
5736 the three variables relative to it. The equivalent pseudocode would
5742 register int *xr = &x;
5743 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5747 (which isn't valid C). We refer to shared addresses like @code{x} as
5748 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5750 The hooks below describe the target properties that GCC needs to know
5751 in order to make effective use of section anchors. It won't use
5752 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5753 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5755 @hook TARGET_MIN_ANCHOR_OFFSET
5756 The minimum offset that should be applied to a section anchor.
5757 On most targets, it should be the smallest offset that can be
5758 applied to a base register while still giving a legitimate address
5759 for every mode. The default value is 0.
5762 @hook TARGET_MAX_ANCHOR_OFFSET
5763 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5764 offset that should be applied to section anchors. The default
5768 @hook TARGET_ASM_OUTPUT_ANCHOR
5769 Write the assembly code to define section anchor @var{x}, which is a
5770 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5771 The hook is called with the assembly output position set to the beginning
5772 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5774 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5775 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5776 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5777 is @code{NULL}, which disables the use of section anchors altogether.
5780 @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
5781 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5782 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5783 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5785 The default version is correct for most targets, but you might need to
5786 intercept this hook to handle things like target-specific attributes
5787 or target-specific sections.
5790 @node Condition Code
5791 @section Condition Code Status
5792 @cindex condition code status
5794 The macros in this section can be split in two families, according to the
5795 two ways of representing condition codes in GCC.
5797 The first representation is the so called @code{(cc0)} representation
5798 (@pxref{Jump Patterns}), where all instructions can have an implicit
5799 clobber of the condition codes. The second is the condition code
5800 register representation, which provides better schedulability for
5801 architectures that do have a condition code register, but on which
5802 most instructions do not affect it. The latter category includes
5805 The implicit clobbering poses a strong restriction on the placement of
5806 the definition and use of the condition code, which need to be in adjacent
5807 insns for machines using @code{(cc0)}. This can prevent important
5808 optimizations on some machines. For example, on the IBM RS/6000, there
5809 is a delay for taken branches unless the condition code register is set
5810 three instructions earlier than the conditional branch. The instruction
5811 scheduler cannot perform this optimization if it is not permitted to
5812 separate the definition and use of the condition code register.
5814 For this reason, it is possible and suggested to use a register to
5815 represent the condition code for new ports. If there is a specific
5816 condition code register in the machine, use a hard register. If the
5817 condition code or comparison result can be placed in any general register,
5818 or if there are multiple condition registers, use a pseudo register.
5819 Registers used to store the condition code value will usually have a mode
5820 that is in class @code{MODE_CC}.
5822 Alternatively, you can use @code{BImode} if the comparison operator is
5823 specified already in the compare instruction. In this case, you are not
5824 interested in most macros in this section.
5827 * CC0 Condition Codes:: Old style representation of condition codes.
5828 * MODE_CC Condition Codes:: Modern representation of condition codes.
5829 * Cond Exec Macros:: Macros to control conditional execution.
5832 @node CC0 Condition Codes
5833 @subsection Representation of condition codes using @code{(cc0)}
5837 The file @file{conditions.h} defines a variable @code{cc_status} to
5838 describe how the condition code was computed (in case the interpretation of
5839 the condition code depends on the instruction that it was set by). This
5840 variable contains the RTL expressions on which the condition code is
5841 currently based, and several standard flags.
5843 Sometimes additional machine-specific flags must be defined in the machine
5844 description header file. It can also add additional machine-specific
5845 information by defining @code{CC_STATUS_MDEP}.
5847 @defmac CC_STATUS_MDEP
5848 C code for a data type which is used for declaring the @code{mdep}
5849 component of @code{cc_status}. It defaults to @code{int}.
5851 This macro is not used on machines that do not use @code{cc0}.
5854 @defmac CC_STATUS_MDEP_INIT
5855 A C expression to initialize the @code{mdep} field to ``empty''.
5856 The default definition does nothing, since most machines don't use
5857 the field anyway. If you want to use the field, you should probably
5858 define this macro to initialize it.
5860 This macro is not used on machines that do not use @code{cc0}.
5863 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5864 A C compound statement to set the components of @code{cc_status}
5865 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5866 this macro's responsibility to recognize insns that set the condition
5867 code as a byproduct of other activity as well as those that explicitly
5870 This macro is not used on machines that do not use @code{cc0}.
5872 If there are insns that do not set the condition code but do alter
5873 other machine registers, this macro must check to see whether they
5874 invalidate the expressions that the condition code is recorded as
5875 reflecting. For example, on the 68000, insns that store in address
5876 registers do not set the condition code, which means that usually
5877 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5878 insns. But suppose that the previous insn set the condition code
5879 based on location @samp{a4@@(102)} and the current insn stores a new
5880 value in @samp{a4}. Although the condition code is not changed by
5881 this, it will no longer be true that it reflects the contents of
5882 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5883 @code{cc_status} in this case to say that nothing is known about the
5884 condition code value.
5886 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5887 with the results of peephole optimization: insns whose patterns are
5888 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5889 constants which are just the operands. The RTL structure of these
5890 insns is not sufficient to indicate what the insns actually do. What
5891 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5892 @code{CC_STATUS_INIT}.
5894 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5895 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5896 @samp{cc}. This avoids having detailed information about patterns in
5897 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5900 @node MODE_CC Condition Codes
5901 @subsection Representation of condition codes using registers
5905 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5906 On many machines, the condition code may be produced by other instructions
5907 than compares, for example the branch can use directly the condition
5908 code set by a subtract instruction. However, on some machines
5909 when the condition code is set this way some bits (such as the overflow
5910 bit) are not set in the same way as a test instruction, so that a different
5911 branch instruction must be used for some conditional branches. When
5912 this happens, use the machine mode of the condition code register to
5913 record different formats of the condition code register. Modes can
5914 also be used to record which compare instruction (e.g. a signed or an
5915 unsigned comparison) produced the condition codes.
5917 If other modes than @code{CCmode} are required, add them to
5918 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
5919 a mode given an operand of a compare. This is needed because the modes
5920 have to be chosen not only during RTL generation but also, for example,
5921 by instruction combination. The result of @code{SELECT_CC_MODE} should
5922 be consistent with the mode used in the patterns; for example to support
5923 the case of the add on the SPARC discussed above, we have the pattern
5927 [(set (reg:CC_NOOV 0)
5929 (plus:SI (match_operand:SI 0 "register_operand" "%r")
5930 (match_operand:SI 1 "arith_operand" "rI"))
5937 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
5938 for comparisons whose argument is a @code{plus}:
5941 #define SELECT_CC_MODE(OP,X,Y) \
5942 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5943 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5944 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5945 || GET_CODE (X) == NEG) \
5946 ? CC_NOOVmode : CCmode))
5949 Another reason to use modes is to retain information on which operands
5950 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
5953 You should define this macro if and only if you define extra CC modes
5954 in @file{@var{machine}-modes.def}.
5957 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5958 On some machines not all possible comparisons are defined, but you can
5959 convert an invalid comparison into a valid one. For example, the Alpha
5960 does not have a @code{GT} comparison, but you can use an @code{LT}
5961 comparison instead and swap the order of the operands.
5963 On such machines, define this macro to be a C statement to do any
5964 required conversions. @var{code} is the initial comparison code
5965 and @var{op0} and @var{op1} are the left and right operands of the
5966 comparison, respectively. You should modify @var{code}, @var{op0}, and
5967 @var{op1} as required.
5969 GCC will not assume that the comparison resulting from this macro is
5970 valid but will see if the resulting insn matches a pattern in the
5973 You need not define this macro if it would never change the comparison
5977 @defmac REVERSIBLE_CC_MODE (@var{mode})
5978 A C expression whose value is one if it is always safe to reverse a
5979 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5980 can ever return @var{mode} for a floating-point inequality comparison,
5981 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5983 You need not define this macro if it would always returns zero or if the
5984 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5985 For example, here is the definition used on the SPARC, where floating-point
5986 inequality comparisons are always given @code{CCFPEmode}:
5989 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5993 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5994 A C expression whose value is reversed condition code of the @var{code} for
5995 comparison done in CC_MODE @var{mode}. The macro is used only in case
5996 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5997 machine has some non-standard way how to reverse certain conditionals. For
5998 instance in case all floating point conditions are non-trapping, compiler may
5999 freely convert unordered compares to ordered one. Then definition may look
6003 #define REVERSE_CONDITION(CODE, MODE) \
6004 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6005 : reverse_condition_maybe_unordered (CODE))
6009 @hook TARGET_FIXED_CONDITION_CODE_REGS
6010 On targets which do not use @code{(cc0)}, and which use a hard
6011 register rather than a pseudo-register to hold condition codes, the
6012 regular CSE passes are often not able to identify cases in which the
6013 hard register is set to a common value. Use this hook to enable a
6014 small pass which optimizes such cases. This hook should return true
6015 to enable this pass, and it should set the integers to which its
6016 arguments point to the hard register numbers used for condition codes.
6017 When there is only one such register, as is true on most systems, the
6018 integer pointed to by @var{p2} should be set to
6019 @code{INVALID_REGNUM}.
6021 The default version of this hook returns false.
6024 @hook TARGET_CC_MODES_COMPATIBLE
6025 On targets which use multiple condition code modes in class
6026 @code{MODE_CC}, it is sometimes the case that a comparison can be
6027 validly done in more than one mode. On such a system, define this
6028 target hook to take two mode arguments and to return a mode in which
6029 both comparisons may be validly done. If there is no such mode,
6030 return @code{VOIDmode}.
6032 The default version of this hook checks whether the modes are the
6033 same. If they are, it returns that mode. If they are different, it
6034 returns @code{VOIDmode}.
6037 @node Cond Exec Macros
6038 @subsection Macros to control conditional execution
6039 @findex conditional execution
6042 There is one macro that may need to be defined for targets
6043 supporting conditional execution, independent of how they
6044 represent conditional branches.
6046 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6047 A C expression that returns true if the conditional execution predicate
6048 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6049 versa. Define this to return 0 if the target has conditional execution
6050 predicates that cannot be reversed safely. There is no need to validate
6051 that the arguments of op1 and op2 are the same, this is done separately.
6052 If no expansion is specified, this macro is defined as follows:
6055 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6056 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6061 @section Describing Relative Costs of Operations
6062 @cindex costs of instructions
6063 @cindex relative costs
6064 @cindex speed of instructions
6066 These macros let you describe the relative speed of various operations
6067 on the target machine.
6069 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6070 A C expression for the cost of moving data of mode @var{mode} from a
6071 register in class @var{from} to one in class @var{to}. The classes are
6072 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6073 value of 2 is the default; other values are interpreted relative to
6076 It is not required that the cost always equal 2 when @var{from} is the
6077 same as @var{to}; on some machines it is expensive to move between
6078 registers if they are not general registers.
6080 If reload sees an insn consisting of a single @code{set} between two
6081 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6082 classes returns a value of 2, reload does not check to ensure that the
6083 constraints of the insn are met. Setting a cost of other than 2 will
6084 allow reload to verify that the constraints are met. You should do this
6085 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6087 These macros are obsolete, new ports should use the target hook
6088 @code{TARGET_REGISTER_MOVE_COST} instead.
6091 @hook TARGET_REGISTER_MOVE_COST
6092 This target hook should return the cost of moving data of mode @var{mode}
6093 from a register in class @var{from} to one in class @var{to}. The classes
6094 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6095 A value of 2 is the default; other values are interpreted relative to
6098 It is not required that the cost always equal 2 when @var{from} is the
6099 same as @var{to}; on some machines it is expensive to move between
6100 registers if they are not general registers.
6102 If reload sees an insn consisting of a single @code{set} between two
6103 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6104 classes returns a value of 2, reload does not check to ensure that the
6105 constraints of the insn are met. Setting a cost of other than 2 will
6106 allow reload to verify that the constraints are met. You should do this
6107 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6109 The default version of this function returns 2.
6112 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6113 A C expression for the cost of moving data of mode @var{mode} between a
6114 register of class @var{class} and memory; @var{in} is zero if the value
6115 is to be written to memory, nonzero if it is to be read in. This cost
6116 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6117 registers and memory is more expensive than between two registers, you
6118 should define this macro to express the relative cost.
6120 If you do not define this macro, GCC uses a default cost of 4 plus
6121 the cost of copying via a secondary reload register, if one is
6122 needed. If your machine requires a secondary reload register to copy
6123 between memory and a register of @var{class} but the reload mechanism is
6124 more complex than copying via an intermediate, define this macro to
6125 reflect the actual cost of the move.
6127 GCC defines the function @code{memory_move_secondary_cost} if
6128 secondary reloads are needed. It computes the costs due to copying via
6129 a secondary register. If your machine copies from memory using a
6130 secondary register in the conventional way but the default base value of
6131 4 is not correct for your machine, define this macro to add some other
6132 value to the result of that function. The arguments to that function
6133 are the same as to this macro.
6135 These macros are obsolete, new ports should use the target hook
6136 @code{TARGET_MEMORY_MOVE_COST} instead.
6139 @hook TARGET_MEMORY_MOVE_COST
6140 This target hook should return the cost of moving data of mode @var{mode}
6141 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6142 if the value is to be written to memory, @code{true} if it is to be read in.
6143 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6144 If moving between registers and memory is more expensive than between two
6145 registers, you should add this target hook to express the relative cost.
6147 If you do not add this target hook, GCC uses a default cost of 4 plus
6148 the cost of copying via a secondary reload register, if one is
6149 needed. If your machine requires a secondary reload register to copy
6150 between memory and a register of @var{rclass} but the reload mechanism is
6151 more complex than copying via an intermediate, use this target hook to
6152 reflect the actual cost of the move.
6154 GCC defines the function @code{memory_move_secondary_cost} if
6155 secondary reloads are needed. It computes the costs due to copying via
6156 a secondary register. If your machine copies from memory using a
6157 secondary register in the conventional way but the default base value of
6158 4 is not correct for your machine, use this target hook to add some other
6159 value to the result of that function. The arguments to that function
6160 are the same as to this target hook.
6163 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6164 A C expression for the cost of a branch instruction. A value of 1 is
6165 the default; other values are interpreted relative to that. Parameter
6166 @var{speed_p} is true when the branch in question should be optimized
6167 for speed. When it is false, @code{BRANCH_COST} should return a value
6168 optimal for code size rather than performance. @var{predictable_p} is
6169 true for well-predicted branches. On many architectures the
6170 @code{BRANCH_COST} can be reduced then.
6173 Here are additional macros which do not specify precise relative costs,
6174 but only that certain actions are more expensive than GCC would
6177 @defmac SLOW_BYTE_ACCESS
6178 Define this macro as a C expression which is nonzero if accessing less
6179 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6180 faster than accessing a word of memory, i.e., if such access
6181 require more than one instruction or if there is no difference in cost
6182 between byte and (aligned) word loads.
6184 When this macro is not defined, the compiler will access a field by
6185 finding the smallest containing object; when it is defined, a fullword
6186 load will be used if alignment permits. Unless bytes accesses are
6187 faster than word accesses, using word accesses is preferable since it
6188 may eliminate subsequent memory access if subsequent accesses occur to
6189 other fields in the same word of the structure, but to different bytes.
6192 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6193 Define this macro to be the value 1 if memory accesses described by the
6194 @var{mode} and @var{alignment} parameters have a cost many times greater
6195 than aligned accesses, for example if they are emulated in a trap
6198 When this macro is nonzero, the compiler will act as if
6199 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6200 moves. This can cause significantly more instructions to be produced.
6201 Therefore, do not set this macro nonzero if unaligned accesses only add a
6202 cycle or two to the time for a memory access.
6204 If the value of this macro is always zero, it need not be defined. If
6205 this macro is defined, it should produce a nonzero value when
6206 @code{STRICT_ALIGNMENT} is nonzero.
6209 @defmac MOVE_RATIO (@var{speed})
6210 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6211 which a sequence of insns should be generated instead of a
6212 string move insn or a library call. Increasing the value will always
6213 make code faster, but eventually incurs high cost in increased code size.
6215 Note that on machines where the corresponding move insn is a
6216 @code{define_expand} that emits a sequence of insns, this macro counts
6217 the number of such sequences.
6219 The parameter @var{speed} is true if the code is currently being
6220 optimized for speed rather than size.
6222 If you don't define this, a reasonable default is used.
6225 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6226 A C expression used to determine whether @code{move_by_pieces} will be used to
6227 copy a chunk of memory, or whether some other block move mechanism
6228 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6229 than @code{MOVE_RATIO}.
6232 @defmac MOVE_MAX_PIECES
6233 A C expression used by @code{move_by_pieces} to determine the largest unit
6234 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6237 @defmac CLEAR_RATIO (@var{speed})
6238 The threshold of number of scalar move insns, @emph{below} which a sequence
6239 of insns should be generated to clear memory instead of a string clear insn
6240 or a library call. Increasing the value will always make code faster, but
6241 eventually incurs high cost in increased code size.
6243 The parameter @var{speed} is true if the code is currently being
6244 optimized for speed rather than size.
6246 If you don't define this, a reasonable default is used.
6249 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6250 A C expression used to determine whether @code{clear_by_pieces} will be used
6251 to clear a chunk of memory, or whether some other block clear mechanism
6252 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6253 than @code{CLEAR_RATIO}.
6256 @defmac SET_RATIO (@var{speed})
6257 The threshold of number of scalar move insns, @emph{below} which a sequence
6258 of insns should be generated to set memory to a constant value, instead of
6259 a block set insn or a library call.
6260 Increasing the value will always make code faster, but
6261 eventually incurs high cost in increased code size.
6263 The parameter @var{speed} is true if the code is currently being
6264 optimized for speed rather than size.
6266 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6269 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6270 A C expression used to determine whether @code{store_by_pieces} will be
6271 used to set a chunk of memory to a constant value, or whether some
6272 other mechanism will be used. Used by @code{__builtin_memset} when
6273 storing values other than constant zero.
6274 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6275 than @code{SET_RATIO}.
6278 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6279 A C expression used to determine whether @code{store_by_pieces} will be
6280 used to set a chunk of memory to a constant string value, or whether some
6281 other mechanism will be used. Used by @code{__builtin_strcpy} when
6282 called with a constant source string.
6283 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6284 than @code{MOVE_RATIO}.
6287 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6288 A C expression used to determine whether a load postincrement is a good
6289 thing to use for a given mode. Defaults to the value of
6290 @code{HAVE_POST_INCREMENT}.
6293 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6294 A C expression used to determine whether a load postdecrement is a good
6295 thing to use for a given mode. Defaults to the value of
6296 @code{HAVE_POST_DECREMENT}.
6299 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6300 A C expression used to determine whether a load preincrement is a good
6301 thing to use for a given mode. Defaults to the value of
6302 @code{HAVE_PRE_INCREMENT}.
6305 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6306 A C expression used to determine whether a load predecrement is a good
6307 thing to use for a given mode. Defaults to the value of
6308 @code{HAVE_PRE_DECREMENT}.
6311 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6312 A C expression used to determine whether a store postincrement is a good
6313 thing to use for a given mode. Defaults to the value of
6314 @code{HAVE_POST_INCREMENT}.
6317 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6318 A C expression used to determine whether a store postdecrement is a good
6319 thing to use for a given mode. Defaults to the value of
6320 @code{HAVE_POST_DECREMENT}.
6323 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6324 This macro is used to determine whether a store preincrement is a good
6325 thing to use for a given mode. Defaults to the value of
6326 @code{HAVE_PRE_INCREMENT}.
6329 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6330 This macro is used to determine whether a store predecrement is a good
6331 thing to use for a given mode. Defaults to the value of
6332 @code{HAVE_PRE_DECREMENT}.
6335 @defmac NO_FUNCTION_CSE
6336 Define this macro if it is as good or better to call a constant
6337 function address than to call an address kept in a register.
6340 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6341 Define this macro if a non-short-circuit operation produced by
6342 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6343 @code{BRANCH_COST} is greater than or equal to the value 2.
6346 @hook TARGET_RTX_COSTS
6347 This target hook describes the relative costs of RTL expressions.
6349 The cost may depend on the precise form of the expression, which is
6350 available for examination in @var{x}, and the fact that @var{x} appears
6351 as operand @var{opno} of an expression with rtx code @var{outer_code}.
6352 That is, the hook can assume that there is some rtx @var{y} such
6353 that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that
6354 either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or
6355 (b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}.
6357 @var{code} is @var{x}'s expression code---redundant, since it can be
6358 obtained with @code{GET_CODE (@var{x})}.
6360 In implementing this hook, you can use the construct
6361 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6364 On entry to the hook, @code{*@var{total}} contains a default estimate
6365 for the cost of the expression. The hook should modify this value as
6366 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6367 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6368 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6370 When optimizing for code size, i.e.@: when @code{speed} is
6371 false, this target hook should be used to estimate the relative
6372 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6374 The hook returns true when all subexpressions of @var{x} have been
6375 processed, and false when @code{rtx_cost} should recurse.
6378 @hook TARGET_ADDRESS_COST
6379 This hook computes the cost of an addressing mode that contains
6380 @var{address}. If not defined, the cost is computed from
6381 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6383 For most CISC machines, the default cost is a good approximation of the
6384 true cost of the addressing mode. However, on RISC machines, all
6385 instructions normally have the same length and execution time. Hence
6386 all addresses will have equal costs.
6388 In cases where more than one form of an address is known, the form with
6389 the lowest cost will be used. If multiple forms have the same, lowest,
6390 cost, the one that is the most complex will be used.
6392 For example, suppose an address that is equal to the sum of a register
6393 and a constant is used twice in the same basic block. When this macro
6394 is not defined, the address will be computed in a register and memory
6395 references will be indirect through that register. On machines where
6396 the cost of the addressing mode containing the sum is no higher than
6397 that of a simple indirect reference, this will produce an additional
6398 instruction and possibly require an additional register. Proper
6399 specification of this macro eliminates this overhead for such machines.
6401 This hook is never called with an invalid address.
6403 On machines where an address involving more than one register is as
6404 cheap as an address computation involving only one register, defining
6405 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6406 be live over a region of code where only one would have been if
6407 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6408 should be considered in the definition of this macro. Equivalent costs
6409 should probably only be given to addresses with different numbers of
6410 registers on machines with lots of registers.
6414 @section Adjusting the Instruction Scheduler
6416 The instruction scheduler may need a fair amount of machine-specific
6417 adjustment in order to produce good code. GCC provides several target
6418 hooks for this purpose. It is usually enough to define just a few of
6419 them: try the first ones in this list first.
6421 @hook TARGET_SCHED_ISSUE_RATE
6422 This hook returns the maximum number of instructions that can ever
6423 issue at the same time on the target machine. The default is one.
6424 Although the insn scheduler can define itself the possibility of issue
6425 an insn on the same cycle, the value can serve as an additional
6426 constraint to issue insns on the same simulated processor cycle (see
6427 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6428 This value must be constant over the entire compilation. If you need
6429 it to vary depending on what the instructions are, you must use
6430 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6433 @hook TARGET_SCHED_VARIABLE_ISSUE
6434 This hook is executed by the scheduler after it has scheduled an insn
6435 from the ready list. It should return the number of insns which can
6436 still be issued in the current cycle. The default is
6437 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6438 @code{USE}, which normally are not counted against the issue rate.
6439 You should define this hook if some insns take more machine resources
6440 than others, so that fewer insns can follow them in the same cycle.
6441 @var{file} is either a null pointer, or a stdio stream to write any
6442 debug output to. @var{verbose} is the verbose level provided by
6443 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6447 @hook TARGET_SCHED_ADJUST_COST
6448 This function corrects the value of @var{cost} based on the
6449 relationship between @var{insn} and @var{dep_insn} through the
6450 dependence @var{link}. It should return the new value. The default
6451 is to make no adjustment to @var{cost}. This can be used for example
6452 to specify to the scheduler using the traditional pipeline description
6453 that an output- or anti-dependence does not incur the same cost as a
6454 data-dependence. If the scheduler using the automaton based pipeline
6455 description, the cost of anti-dependence is zero and the cost of
6456 output-dependence is maximum of one and the difference of latency
6457 times of the first and the second insns. If these values are not
6458 acceptable, you could use the hook to modify them too. See also
6459 @pxref{Processor pipeline description}.
6462 @hook TARGET_SCHED_ADJUST_PRIORITY
6463 This hook adjusts the integer scheduling priority @var{priority} of
6464 @var{insn}. It should return the new priority. Increase the priority to
6465 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6466 later. Do not define this hook if you do not need to adjust the
6467 scheduling priorities of insns.
6470 @hook TARGET_SCHED_REORDER
6471 This hook is executed by the scheduler after it has scheduled the ready
6472 list, to allow the machine description to reorder it (for example to
6473 combine two small instructions together on @samp{VLIW} machines).
6474 @var{file} is either a null pointer, or a stdio stream to write any
6475 debug output to. @var{verbose} is the verbose level provided by
6476 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6477 list of instructions that are ready to be scheduled. @var{n_readyp} is
6478 a pointer to the number of elements in the ready list. The scheduler
6479 reads the ready list in reverse order, starting with
6480 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6481 is the timer tick of the scheduler. You may modify the ready list and
6482 the number of ready insns. The return value is the number of insns that
6483 can issue this cycle; normally this is just @code{issue_rate}. See also
6484 @samp{TARGET_SCHED_REORDER2}.
6487 @hook TARGET_SCHED_REORDER2
6488 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6489 function is called whenever the scheduler starts a new cycle. This one
6490 is called once per iteration over a cycle, immediately after
6491 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6492 return the number of insns to be scheduled in the same cycle. Defining
6493 this hook can be useful if there are frequent situations where
6494 scheduling one insn causes other insns to become ready in the same
6495 cycle. These other insns can then be taken into account properly.
6498 @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
6499 This hook is called after evaluation forward dependencies of insns in
6500 chain given by two parameter values (@var{head} and @var{tail}
6501 correspondingly) but before insns scheduling of the insn chain. For
6502 example, it can be used for better insn classification if it requires
6503 analysis of dependencies. This hook can use backward and forward
6504 dependencies of the insn scheduler because they are already
6508 @hook TARGET_SCHED_INIT
6509 This hook is executed by the scheduler at the beginning of each block of
6510 instructions that are to be scheduled. @var{file} is either a null
6511 pointer, or a stdio stream to write any debug output to. @var{verbose}
6512 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6513 @var{max_ready} is the maximum number of insns in the current scheduling
6514 region that can be live at the same time. This can be used to allocate
6515 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6518 @hook TARGET_SCHED_FINISH
6519 This hook is executed by the scheduler at the end of each block of
6520 instructions that are to be scheduled. It can be used to perform
6521 cleanup of any actions done by the other scheduling hooks. @var{file}
6522 is either a null pointer, or a stdio stream to write any debug output
6523 to. @var{verbose} is the verbose level provided by
6524 @option{-fsched-verbose-@var{n}}.
6527 @hook TARGET_SCHED_INIT_GLOBAL
6528 This hook is executed by the scheduler after function level initializations.
6529 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6530 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6531 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6534 @hook TARGET_SCHED_FINISH_GLOBAL
6535 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6536 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6537 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6540 @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
6541 The hook returns an RTL insn. The automaton state used in the
6542 pipeline hazard recognizer is changed as if the insn were scheduled
6543 when the new simulated processor cycle starts. Usage of the hook may
6544 simplify the automaton pipeline description for some @acronym{VLIW}
6545 processors. If the hook is defined, it is used only for the automaton
6546 based pipeline description. The default is not to change the state
6547 when the new simulated processor cycle starts.
6550 @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
6551 The hook can be used to initialize data used by the previous hook.
6554 @hook TARGET_SCHED_DFA_POST_CYCLE_INSN
6555 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6556 to changed the state as if the insn were scheduled when the new
6557 simulated processor cycle finishes.
6560 @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
6561 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6562 used to initialize data used by the previous hook.
6565 @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
6566 The hook to notify target that the current simulated cycle is about to finish.
6567 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6568 to change the state in more complicated situations - e.g., when advancing
6569 state on a single insn is not enough.
6572 @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
6573 The hook to notify target that new simulated cycle has just started.
6574 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6575 to change the state in more complicated situations - e.g., when advancing
6576 state on a single insn is not enough.
6579 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
6580 This hook controls better choosing an insn from the ready insn queue
6581 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6582 chooses the first insn from the queue. If the hook returns a positive
6583 value, an additional scheduler code tries all permutations of
6584 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6585 subsequent ready insns to choose an insn whose issue will result in
6586 maximal number of issued insns on the same cycle. For the
6587 @acronym{VLIW} processor, the code could actually solve the problem of
6588 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6589 rules of @acronym{VLIW} packing are described in the automaton.
6591 This code also could be used for superscalar @acronym{RISC}
6592 processors. Let us consider a superscalar @acronym{RISC} processor
6593 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6594 @var{B}, some insns can be executed only in pipelines @var{B} or
6595 @var{C}, and one insn can be executed in pipeline @var{B}. The
6596 processor may issue the 1st insn into @var{A} and the 2nd one into
6597 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6598 until the next cycle. If the scheduler issues the 3rd insn the first,
6599 the processor could issue all 3 insns per cycle.
6601 Actually this code demonstrates advantages of the automaton based
6602 pipeline hazard recognizer. We try quickly and easy many insn
6603 schedules to choose the best one.
6605 The default is no multipass scheduling.
6608 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
6610 This hook controls what insns from the ready insn queue will be
6611 considered for the multipass insn scheduling. If the hook returns
6612 zero for @var{insn}, the insn will be not chosen to
6615 The default is that any ready insns can be chosen to be issued.
6618 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
6619 This hook prepares the target backend for a new round of multipass
6623 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
6624 This hook is called when multipass scheduling evaluates instruction INSN.
6627 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
6628 This is called when multipass scheduling backtracks from evaluation of
6632 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
6633 This hook notifies the target about the result of the concluded current
6634 round of multipass scheduling.
6637 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
6638 This hook initializes target-specific data used in multipass scheduling.
6641 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
6642 This hook finalizes target-specific data used in multipass scheduling.
6645 @hook TARGET_SCHED_DFA_NEW_CYCLE
6646 This hook is called by the insn scheduler before issuing @var{insn}
6647 on cycle @var{clock}. If the hook returns nonzero,
6648 @var{insn} is not issued on this processor cycle. Instead,
6649 the processor cycle is advanced. If *@var{sort_p}
6650 is zero, the insn ready queue is not sorted on the new cycle
6651 start as usually. @var{dump} and @var{verbose} specify the file and
6652 verbosity level to use for debugging output.
6653 @var{last_clock} and @var{clock} are, respectively, the
6654 processor cycle on which the previous insn has been issued,
6655 and the current processor cycle.
6658 @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
6659 This hook is used to define which dependences are considered costly by
6660 the target, so costly that it is not advisable to schedule the insns that
6661 are involved in the dependence too close to one another. The parameters
6662 to this hook are as follows: The first parameter @var{_dep} is the dependence
6663 being evaluated. The second parameter @var{cost} is the cost of the
6664 dependence as estimated by the scheduler, and the third
6665 parameter @var{distance} is the distance in cycles between the two insns.
6666 The hook returns @code{true} if considering the distance between the two
6667 insns the dependence between them is considered costly by the target,
6668 and @code{false} otherwise.
6670 Defining this hook can be useful in multiple-issue out-of-order machines,
6671 where (a) it's practically hopeless to predict the actual data/resource
6672 delays, however: (b) there's a better chance to predict the actual grouping
6673 that will be formed, and (c) correctly emulating the grouping can be very
6674 important. In such targets one may want to allow issuing dependent insns
6675 closer to one another---i.e., closer than the dependence distance; however,
6676 not in cases of ``costly dependences'', which this hooks allows to define.
6679 @hook TARGET_SCHED_H_I_D_EXTENDED
6680 This hook is called by the insn scheduler after emitting a new instruction to
6681 the instruction stream. The hook notifies a target backend to extend its
6682 per instruction data structures.
6685 @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
6686 Return a pointer to a store large enough to hold target scheduling context.
6689 @hook TARGET_SCHED_INIT_SCHED_CONTEXT
6690 Initialize store pointed to by @var{tc} to hold target scheduling context.
6691 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6692 beginning of the block. Otherwise, copy the current context into @var{tc}.
6695 @hook TARGET_SCHED_SET_SCHED_CONTEXT
6696 Copy target scheduling context pointed to by @var{tc} to the current context.
6699 @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
6700 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6703 @hook TARGET_SCHED_FREE_SCHED_CONTEXT
6704 Deallocate a store for target scheduling context pointed to by @var{tc}.
6707 @hook TARGET_SCHED_SPECULATE_INSN
6708 This hook is called by the insn scheduler when @var{insn} has only
6709 speculative dependencies and therefore can be scheduled speculatively.
6710 The hook is used to check if the pattern of @var{insn} has a speculative
6711 version and, in case of successful check, to generate that speculative
6712 pattern. The hook should return 1, if the instruction has a speculative form,
6713 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6714 speculation. If the return value equals 1 then @var{new_pat} is assigned
6715 the generated speculative pattern.
6718 @hook TARGET_SCHED_NEEDS_BLOCK_P
6719 This hook is called by the insn scheduler during generation of recovery code
6720 for @var{insn}. It should return @code{true}, if the corresponding check
6721 instruction should branch to recovery code, or @code{false} otherwise.
6724 @hook TARGET_SCHED_GEN_SPEC_CHECK
6725 This hook is called by the insn scheduler to generate a pattern for recovery
6726 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6727 speculative instruction for which the check should be generated.
6728 @var{label} is either a label of a basic block, where recovery code should
6729 be emitted, or a null pointer, when requested check doesn't branch to
6730 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6731 a pattern for a branchy check corresponding to a simple check denoted by
6732 @var{insn} should be generated. In this case @var{label} can't be null.
6735 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC
6736 This hook is used as a workaround for
6737 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6738 called on the first instruction of the ready list. The hook is used to
6739 discard speculative instructions that stand first in the ready list from
6740 being scheduled on the current cycle. If the hook returns @code{false},
6741 @var{insn} will not be chosen to be issued.
6742 For non-speculative instructions,
6743 the hook should always return @code{true}. For example, in the ia64 backend
6744 the hook is used to cancel data speculative insns when the ALAT table
6748 @hook TARGET_SCHED_SET_SCHED_FLAGS
6749 This hook is used by the insn scheduler to find out what features should be
6751 The structure *@var{spec_info} should be filled in by the target.
6752 The structure describes speculation types that can be used in the scheduler.
6755 @hook TARGET_SCHED_SMS_RES_MII
6756 This hook is called by the swing modulo scheduler to calculate a
6757 resource-based lower bound which is based on the resources available in
6758 the machine and the resources required by each instruction. The target
6759 backend can use @var{g} to calculate such bound. A very simple lower
6760 bound will be used in case this hook is not implemented: the total number
6761 of instructions divided by the issue rate.
6764 @hook TARGET_SCHED_DISPATCH
6765 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6766 is supported in hardware and the condition specified in the parameter is true.
6769 @hook TARGET_SCHED_DISPATCH_DO
6770 This hook is called by Haifa Scheduler. It performs the operation specified
6771 in its second parameter.
6774 @hook TARGET_SCHED_EXPOSED_PIPELINE
6776 @hook TARGET_SCHED_REASSOCIATION_WIDTH
6779 @section Dividing the Output into Sections (Texts, Data, @dots{})
6780 @c the above section title is WAY too long. maybe cut the part between
6781 @c the (...)? --mew 10feb93
6783 An object file is divided into sections containing different types of
6784 data. In the most common case, there are three sections: the @dfn{text
6785 section}, which holds instructions and read-only data; the @dfn{data
6786 section}, which holds initialized writable data; and the @dfn{bss
6787 section}, which holds uninitialized data. Some systems have other kinds
6790 @file{varasm.c} provides several well-known sections, such as
6791 @code{text_section}, @code{data_section} and @code{bss_section}.
6792 The normal way of controlling a @code{@var{foo}_section} variable
6793 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6794 as described below. The macros are only read once, when @file{varasm.c}
6795 initializes itself, so their values must be run-time constants.
6796 They may however depend on command-line flags.
6798 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6799 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6800 to be string literals.
6802 Some assemblers require a different string to be written every time a
6803 section is selected. If your assembler falls into this category, you
6804 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6805 @code{get_unnamed_section} to set up the sections.
6807 You must always create a @code{text_section}, either by defining
6808 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6809 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6810 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6811 create a distinct @code{readonly_data_section}, the default is to
6812 reuse @code{text_section}.
6814 All the other @file{varasm.c} sections are optional, and are null
6815 if the target does not provide them.
6817 @defmac TEXT_SECTION_ASM_OP
6818 A C expression whose value is a string, including spacing, containing the
6819 assembler operation that should precede instructions and read-only data.
6820 Normally @code{"\t.text"} is right.
6823 @defmac HOT_TEXT_SECTION_NAME
6824 If defined, a C string constant for the name of the section containing most
6825 frequently executed functions of the program. If not defined, GCC will provide
6826 a default definition if the target supports named sections.
6829 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6830 If defined, a C string constant for the name of the section containing unlikely
6831 executed functions in the program.
6834 @defmac DATA_SECTION_ASM_OP
6835 A C expression whose value is a string, including spacing, containing the
6836 assembler operation to identify the following data as writable initialized
6837 data. Normally @code{"\t.data"} is right.
6840 @defmac SDATA_SECTION_ASM_OP
6841 If defined, a C expression whose value is a string, including spacing,
6842 containing the assembler operation to identify the following data as
6843 initialized, writable small data.
6846 @defmac READONLY_DATA_SECTION_ASM_OP
6847 A C expression whose value is a string, including spacing, containing the
6848 assembler operation to identify the following data as read-only initialized
6852 @defmac BSS_SECTION_ASM_OP
6853 If defined, a C expression whose value is a string, including spacing,
6854 containing the assembler operation to identify the following data as
6855 uninitialized global data. If not defined, and
6856 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
6857 uninitialized global data will be output in the data section if
6858 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6862 @defmac SBSS_SECTION_ASM_OP
6863 If defined, a C expression whose value is a string, including spacing,
6864 containing the assembler operation to identify the following data as
6865 uninitialized, writable small data.
6868 @defmac TLS_COMMON_ASM_OP
6869 If defined, a C expression whose value is a string containing the
6870 assembler operation to identify the following data as thread-local
6871 common data. The default is @code{".tls_common"}.
6874 @defmac TLS_SECTION_ASM_FLAG
6875 If defined, a C expression whose value is a character constant
6876 containing the flag used to mark a section as a TLS section. The
6877 default is @code{'T'}.
6880 @defmac INIT_SECTION_ASM_OP
6881 If defined, a C expression whose value is a string, including spacing,
6882 containing the assembler operation to identify the following data as
6883 initialization code. If not defined, GCC will assume such a section does
6884 not exist. This section has no corresponding @code{init_section}
6885 variable; it is used entirely in runtime code.
6888 @defmac FINI_SECTION_ASM_OP
6889 If defined, a C expression whose value is a string, including spacing,
6890 containing the assembler operation to identify the following data as
6891 finalization code. If not defined, GCC will assume such a section does
6892 not exist. This section has no corresponding @code{fini_section}
6893 variable; it is used entirely in runtime code.
6896 @defmac INIT_ARRAY_SECTION_ASM_OP
6897 If defined, a C expression whose value is a string, including spacing,
6898 containing the assembler operation to identify the following data as
6899 part of the @code{.init_array} (or equivalent) section. If not
6900 defined, GCC will assume such a section does not exist. Do not define
6901 both this macro and @code{INIT_SECTION_ASM_OP}.
6904 @defmac FINI_ARRAY_SECTION_ASM_OP
6905 If defined, a C expression whose value is a string, including spacing,
6906 containing the assembler operation to identify the following data as
6907 part of the @code{.fini_array} (or equivalent) section. If not
6908 defined, GCC will assume such a section does not exist. Do not define
6909 both this macro and @code{FINI_SECTION_ASM_OP}.
6912 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6913 If defined, an ASM statement that switches to a different section
6914 via @var{section_op}, calls @var{function}, and switches back to
6915 the text section. This is used in @file{crtstuff.c} if
6916 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6917 to initialization and finalization functions from the init and fini
6918 sections. By default, this macro uses a simple function call. Some
6919 ports need hand-crafted assembly code to avoid dependencies on
6920 registers initialized in the function prologue or to ensure that
6921 constant pools don't end up too far way in the text section.
6924 @defmac TARGET_LIBGCC_SDATA_SECTION
6925 If defined, a string which names the section into which small
6926 variables defined in crtstuff and libgcc should go. This is useful
6927 when the target has options for optimizing access to small data, and
6928 you want the crtstuff and libgcc routines to be conservative in what
6929 they expect of your application yet liberal in what your application
6930 expects. For example, for targets with a @code{.sdata} section (like
6931 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6932 require small data support from your application, but use this macro
6933 to put small data into @code{.sdata} so that your application can
6934 access these variables whether it uses small data or not.
6937 @defmac FORCE_CODE_SECTION_ALIGN
6938 If defined, an ASM statement that aligns a code section to some
6939 arbitrary boundary. This is used to force all fragments of the
6940 @code{.init} and @code{.fini} sections to have to same alignment
6941 and thus prevent the linker from having to add any padding.
6944 @defmac JUMP_TABLES_IN_TEXT_SECTION
6945 Define this macro to be an expression with a nonzero value if jump
6946 tables (for @code{tablejump} insns) should be output in the text
6947 section, along with the assembler instructions. Otherwise, the
6948 readonly data section is used.
6950 This macro is irrelevant if there is no separate readonly data section.
6953 @hook TARGET_ASM_INIT_SECTIONS
6954 Define this hook if you need to do something special to set up the
6955 @file{varasm.c} sections, or if your target has some special sections
6956 of its own that you need to create.
6958 GCC calls this hook after processing the command line, but before writing
6959 any assembly code, and before calling any of the section-returning hooks
6963 @hook TARGET_ASM_RELOC_RW_MASK
6964 Return a mask describing how relocations should be treated when
6965 selecting sections. Bit 1 should be set if global relocations
6966 should be placed in a read-write section; bit 0 should be set if
6967 local relocations should be placed in a read-write section.
6969 The default version of this function returns 3 when @option{-fpic}
6970 is in effect, and 0 otherwise. The hook is typically redefined
6971 when the target cannot support (some kinds of) dynamic relocations
6972 in read-only sections even in executables.
6975 @hook TARGET_ASM_SELECT_SECTION
6976 Return the section into which @var{exp} should be placed. You can
6977 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6978 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6979 requires link-time relocations. Bit 0 is set when variable contains
6980 local relocations only, while bit 1 is set for global relocations.
6981 @var{align} is the constant alignment in bits.
6983 The default version of this function takes care of putting read-only
6984 variables in @code{readonly_data_section}.
6986 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6989 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6990 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6991 for @code{FUNCTION_DECL}s as well as for variables and constants.
6993 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6994 function has been determined to be likely to be called, and nonzero if
6995 it is unlikely to be called.
6998 @hook TARGET_ASM_UNIQUE_SECTION
6999 Build up a unique section name, expressed as a @code{STRING_CST} node,
7000 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7001 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7002 the initial value of @var{exp} requires link-time relocations.
7004 The default version of this function appends the symbol name to the
7005 ELF section name that would normally be used for the symbol. For
7006 example, the function @code{foo} would be placed in @code{.text.foo}.
7007 Whatever the actual target object format, this is often good enough.
7010 @hook TARGET_ASM_FUNCTION_RODATA_SECTION
7011 Return the readonly data section associated with
7012 @samp{DECL_SECTION_NAME (@var{decl})}.
7013 The default version of this function selects @code{.gnu.linkonce.r.name} if
7014 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7015 if function is in @code{.text.name}, and the normal readonly-data section
7019 @hook TARGET_ASM_MERGEABLE_RODATA_PREFIX
7021 @hook TARGET_ASM_TM_CLONE_TABLE_SECTION
7023 @hook TARGET_ASM_SELECT_RTX_SECTION
7024 Return the section into which a constant @var{x}, of mode @var{mode},
7025 should be placed. You can assume that @var{x} is some kind of
7026 constant in RTL@. The argument @var{mode} is redundant except in the
7027 case of a @code{const_int} rtx. @var{align} is the constant alignment
7030 The default version of this function takes care of putting symbolic
7031 constants in @code{flag_pic} mode in @code{data_section} and everything
7032 else in @code{readonly_data_section}.
7035 @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
7036 Define this hook if you need to postprocess the assembler name generated
7037 by target-independent code. The @var{id} provided to this hook will be
7038 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7039 or the mangled name of the @var{decl} in C++). The return value of the
7040 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7041 your target system. The default implementation of this hook just
7042 returns the @var{id} provided.
7045 @hook TARGET_ENCODE_SECTION_INFO
7046 Define this hook if references to a symbol or a constant must be
7047 treated differently depending on something about the variable or
7048 function named by the symbol (such as what section it is in).
7050 The hook is executed immediately after rtl has been created for
7051 @var{decl}, which may be a variable or function declaration or
7052 an entry in the constant pool. In either case, @var{rtl} is the
7053 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7054 in this hook; that field may not have been initialized yet.
7056 In the case of a constant, it is safe to assume that the rtl is
7057 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7058 will also have this form, but that is not guaranteed. Global
7059 register variables, for instance, will have a @code{reg} for their
7060 rtl. (Normally the right thing to do with such unusual rtl is
7063 The @var{new_decl_p} argument will be true if this is the first time
7064 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7065 be false for subsequent invocations, which will happen for duplicate
7066 declarations. Whether or not anything must be done for the duplicate
7067 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7068 @var{new_decl_p} is always true when the hook is called for a constant.
7070 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7071 The usual thing for this hook to do is to record flags in the
7072 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7073 Historically, the name string was modified if it was necessary to
7074 encode more than one bit of information, but this practice is now
7075 discouraged; use @code{SYMBOL_REF_FLAGS}.
7077 The default definition of this hook, @code{default_encode_section_info}
7078 in @file{varasm.c}, sets a number of commonly-useful bits in
7079 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7080 before overriding it.
7083 @hook TARGET_STRIP_NAME_ENCODING
7084 Decode @var{name} and return the real name part, sans
7085 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7089 @hook TARGET_IN_SMALL_DATA_P
7090 Returns true if @var{exp} should be placed into a ``small data'' section.
7091 The default version of this hook always returns false.
7094 @hook TARGET_HAVE_SRODATA_SECTION
7095 Contains the value true if the target places read-only
7096 ``small data'' into a separate section. The default value is false.
7099 @hook TARGET_PROFILE_BEFORE_PROLOGUE
7101 @hook TARGET_BINDS_LOCAL_P
7102 Returns true if @var{exp} names an object for which name resolution
7103 rules must resolve to the current ``module'' (dynamic shared library
7104 or executable image).
7106 The default version of this hook implements the name resolution rules
7107 for ELF, which has a looser model of global name binding than other
7108 currently supported object file formats.
7111 @hook TARGET_HAVE_TLS
7112 Contains the value true if the target supports thread-local storage.
7113 The default value is false.
7118 @section Position Independent Code
7119 @cindex position independent code
7122 This section describes macros that help implement generation of position
7123 independent code. Simply defining these macros is not enough to
7124 generate valid PIC; you must also add support to the hook
7125 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7126 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7127 must modify the definition of @samp{movsi} to do something appropriate
7128 when the source operand contains a symbolic address. You may also
7129 need to alter the handling of switch statements so that they use
7131 @c i rearranged the order of the macros above to try to force one of
7132 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7134 @defmac PIC_OFFSET_TABLE_REGNUM
7135 The register number of the register used to address a table of static
7136 data addresses in memory. In some cases this register is defined by a
7137 processor's ``application binary interface'' (ABI)@. When this macro
7138 is defined, RTL is generated for this register once, as with the stack
7139 pointer and frame pointer registers. If this macro is not defined, it
7140 is up to the machine-dependent files to allocate such a register (if
7141 necessary). Note that this register must be fixed when in use (e.g.@:
7142 when @code{flag_pic} is true).
7145 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7146 A C expression that is nonzero if the register defined by
7147 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7148 the default is zero. Do not define
7149 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7152 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7153 A C expression that is nonzero if @var{x} is a legitimate immediate
7154 operand on the target machine when generating position independent code.
7155 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7156 check this. You can also assume @var{flag_pic} is true, so you need not
7157 check it either. You need not define this macro if all constants
7158 (including @code{SYMBOL_REF}) can be immediate operands when generating
7159 position independent code.
7162 @node Assembler Format
7163 @section Defining the Output Assembler Language
7165 This section describes macros whose principal purpose is to describe how
7166 to write instructions in assembler language---rather than what the
7170 * File Framework:: Structural information for the assembler file.
7171 * Data Output:: Output of constants (numbers, strings, addresses).
7172 * Uninitialized Data:: Output of uninitialized variables.
7173 * Label Output:: Output and generation of labels.
7174 * Initialization:: General principles of initialization
7175 and termination routines.
7176 * Macros for Initialization::
7177 Specific macros that control the handling of
7178 initialization and termination routines.
7179 * Instruction Output:: Output of actual instructions.
7180 * Dispatch Tables:: Output of jump tables.
7181 * Exception Region Output:: Output of exception region code.
7182 * Alignment Output:: Pseudo ops for alignment and skipping data.
7185 @node File Framework
7186 @subsection The Overall Framework of an Assembler File
7187 @cindex assembler format
7188 @cindex output of assembler code
7190 @c prevent bad page break with this line
7191 This describes the overall framework of an assembly file.
7193 @findex default_file_start
7194 @hook TARGET_ASM_FILE_START
7195 Output to @code{asm_out_file} any text which the assembler expects to
7196 find at the beginning of a file. The default behavior is controlled
7197 by two flags, documented below. Unless your target's assembler is
7198 quite unusual, if you override the default, you should call
7199 @code{default_file_start} at some point in your target hook. This
7200 lets other target files rely on these variables.
7203 @hook TARGET_ASM_FILE_START_APP_OFF
7204 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7205 printed as the very first line in the assembly file, unless
7206 @option{-fverbose-asm} is in effect. (If that macro has been defined
7207 to the empty string, this variable has no effect.) With the normal
7208 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7209 assembler that it need not bother stripping comments or extra
7210 whitespace from its input. This allows it to work a bit faster.
7212 The default is false. You should not set it to true unless you have
7213 verified that your port does not generate any extra whitespace or
7214 comments that will cause GAS to issue errors in NO_APP mode.
7217 @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
7218 If this flag is true, @code{output_file_directive} will be called
7219 for the primary source file, immediately after printing
7220 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7221 this to be done. The default is false.
7224 @hook TARGET_ASM_FILE_END
7225 Output to @code{asm_out_file} any text which the assembler expects
7226 to find at the end of a file. The default is to output nothing.
7229 @deftypefun void file_end_indicate_exec_stack ()
7230 Some systems use a common convention, the @samp{.note.GNU-stack}
7231 special section, to indicate whether or not an object file relies on
7232 the stack being executable. If your system uses this convention, you
7233 should define @code{TARGET_ASM_FILE_END} to this function. If you
7234 need to do other things in that hook, have your hook function call
7238 @hook TARGET_ASM_LTO_START
7239 Output to @code{asm_out_file} any text which the assembler expects
7240 to find at the start of an LTO section. The default is to output
7244 @hook TARGET_ASM_LTO_END
7245 Output to @code{asm_out_file} any text which the assembler expects
7246 to find at the end of an LTO section. The default is to output
7250 @hook TARGET_ASM_CODE_END
7251 Output to @code{asm_out_file} any text which is needed before emitting
7252 unwind info and debug info at the end of a file. Some targets emit
7253 here PIC setup thunks that cannot be emitted at the end of file,
7254 because they couldn't have unwind info then. The default is to output
7258 @defmac ASM_COMMENT_START
7259 A C string constant describing how to begin a comment in the target
7260 assembler language. The compiler assumes that the comment will end at
7261 the end of the line.
7265 A C string constant for text to be output before each @code{asm}
7266 statement or group of consecutive ones. Normally this is
7267 @code{"#APP"}, which is a comment that has no effect on most
7268 assemblers but tells the GNU assembler that it must check the lines
7269 that follow for all valid assembler constructs.
7273 A C string constant for text to be output after each @code{asm}
7274 statement or group of consecutive ones. Normally this is
7275 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7276 time-saving assumptions that are valid for ordinary compiler output.
7279 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7280 A C statement to output COFF information or DWARF debugging information
7281 which indicates that filename @var{name} is the current source file to
7282 the stdio stream @var{stream}.
7284 This macro need not be defined if the standard form of output
7285 for the file format in use is appropriate.
7288 @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
7290 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7291 A C statement to output the string @var{string} to the stdio stream
7292 @var{stream}. If you do not call the function @code{output_quoted_string}
7293 in your config files, GCC will only call it to output filenames to
7294 the assembler source. So you can use it to canonicalize the format
7295 of the filename using this macro.
7298 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7299 A C statement to output something to the assembler file to handle a
7300 @samp{#ident} directive containing the text @var{string}. If this
7301 macro is not defined, nothing is output for a @samp{#ident} directive.
7304 @hook TARGET_ASM_NAMED_SECTION
7305 Output assembly directives to switch to section @var{name}. The section
7306 should have attributes as specified by @var{flags}, which is a bit mask
7307 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7308 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7309 this section is associated.
7312 @hook TARGET_ASM_FUNCTION_SECTION
7313 Return preferred text (sub)section for function @var{decl}.
7314 Main purpose of this function is to separate cold, normal and hot
7315 functions. @var{startup} is true when function is known to be used only
7316 at startup (from static constructors or it is @code{main()}).
7317 @var{exit} is true when function is known to be used only at exit
7318 (from static destructors).
7319 Return NULL if function should go to default text section.
7322 @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
7324 @hook TARGET_HAVE_NAMED_SECTIONS
7325 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7326 It must not be modified by command-line option processing.
7329 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7330 @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7331 This flag is true if we can create zeroed data by switching to a BSS
7332 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7333 This is true on most ELF targets.
7336 @hook TARGET_SECTION_TYPE_FLAGS
7337 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7338 based on a variable or function decl, a section name, and whether or not the
7339 declaration's initializer may contain runtime relocations. @var{decl} may be
7340 null, in which case read-write data should be assumed.
7342 The default version of this function handles choosing code vs data,
7343 read-only vs read-write data, and @code{flag_pic}. You should only
7344 need to override this if your target has special flags that might be
7345 set via @code{__attribute__}.
7348 @hook TARGET_ASM_RECORD_GCC_SWITCHES
7349 Provides the target with the ability to record the gcc command line
7350 switches that have been passed to the compiler, and options that are
7351 enabled. The @var{type} argument specifies what is being recorded.
7352 It can take the following values:
7355 @item SWITCH_TYPE_PASSED
7356 @var{text} is a command line switch that has been set by the user.
7358 @item SWITCH_TYPE_ENABLED
7359 @var{text} is an option which has been enabled. This might be as a
7360 direct result of a command line switch, or because it is enabled by
7361 default or because it has been enabled as a side effect of a different
7362 command line switch. For example, the @option{-O2} switch enables
7363 various different individual optimization passes.
7365 @item SWITCH_TYPE_DESCRIPTIVE
7366 @var{text} is either NULL or some descriptive text which should be
7367 ignored. If @var{text} is NULL then it is being used to warn the
7368 target hook that either recording is starting or ending. The first
7369 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7370 warning is for start up and the second time the warning is for
7371 wind down. This feature is to allow the target hook to make any
7372 necessary preparations before it starts to record switches and to
7373 perform any necessary tidying up after it has finished recording
7376 @item SWITCH_TYPE_LINE_START
7377 This option can be ignored by this target hook.
7379 @item SWITCH_TYPE_LINE_END
7380 This option can be ignored by this target hook.
7383 The hook's return value must be zero. Other return values may be
7384 supported in the future.
7386 By default this hook is set to NULL, but an example implementation is
7387 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7388 it records the switches as ASCII text inside a new, string mergeable
7389 section in the assembler output file. The name of the new section is
7390 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7394 @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7395 This is the name of the section that will be created by the example
7396 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7402 @subsection Output of Data
7405 @hook TARGET_ASM_BYTE_OP
7406 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7407 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7408 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7409 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7410 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7411 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7412 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7413 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7414 These hooks specify assembly directives for creating certain kinds
7415 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7416 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7417 aligned two-byte object, and so on. Any of the hooks may be
7418 @code{NULL}, indicating that no suitable directive is available.
7420 The compiler will print these strings at the start of a new line,
7421 followed immediately by the object's initial value. In most cases,
7422 the string should contain a tab, a pseudo-op, and then another tab.
7425 @hook TARGET_ASM_INTEGER
7426 The @code{assemble_integer} function uses this hook to output an
7427 integer object. @var{x} is the object's value, @var{size} is its size
7428 in bytes and @var{aligned_p} indicates whether it is aligned. The
7429 function should return @code{true} if it was able to output the
7430 object. If it returns false, @code{assemble_integer} will try to
7431 split the object into smaller parts.
7433 The default implementation of this hook will use the
7434 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7435 when the relevant string is @code{NULL}.
7438 @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
7439 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7440 can't deal with, and output assembly code to @var{file} corresponding to
7441 the pattern @var{x}. This may be used to allow machine-dependent
7442 @code{UNSPEC}s to appear within constants.
7444 If target hook fails to recognize a pattern, it must return @code{false},
7445 so that a standard error message is printed. If it prints an error message
7446 itself, by calling, for example, @code{output_operand_lossage}, it may just
7450 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7451 A C statement to output to the stdio stream @var{stream} an assembler
7452 instruction to assemble a string constant containing the @var{len}
7453 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7454 @code{char *} and @var{len} a C expression of type @code{int}.
7456 If the assembler has a @code{.ascii} pseudo-op as found in the
7457 Berkeley Unix assembler, do not define the macro
7458 @code{ASM_OUTPUT_ASCII}.
7461 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7462 A C statement to output word @var{n} of a function descriptor for
7463 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7464 is defined, and is otherwise unused.
7467 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7468 You may define this macro as a C expression. You should define the
7469 expression to have a nonzero value if GCC should output the constant
7470 pool for a function before the code for the function, or a zero value if
7471 GCC should output the constant pool after the function. If you do
7472 not define this macro, the usual case, GCC will output the constant
7473 pool before the function.
7476 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7477 A C statement to output assembler commands to define the start of the
7478 constant pool for a function. @var{funname} is a string giving
7479 the name of the function. Should the return type of the function
7480 be required, it can be obtained via @var{fundecl}. @var{size}
7481 is the size, in bytes, of the constant pool that will be written
7482 immediately after this call.
7484 If no constant-pool prefix is required, the usual case, this macro need
7488 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7489 A C statement (with or without semicolon) to output a constant in the
7490 constant pool, if it needs special treatment. (This macro need not do
7491 anything for RTL expressions that can be output normally.)
7493 The argument @var{file} is the standard I/O stream to output the
7494 assembler code on. @var{x} is the RTL expression for the constant to
7495 output, and @var{mode} is the machine mode (in case @var{x} is a
7496 @samp{const_int}). @var{align} is the required alignment for the value
7497 @var{x}; you should output an assembler directive to force this much
7500 The argument @var{labelno} is a number to use in an internal label for
7501 the address of this pool entry. The definition of this macro is
7502 responsible for outputting the label definition at the proper place.
7503 Here is how to do this:
7506 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7509 When you output a pool entry specially, you should end with a
7510 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7511 entry from being output a second time in the usual manner.
7513 You need not define this macro if it would do nothing.
7516 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7517 A C statement to output assembler commands to at the end of the constant
7518 pool for a function. @var{funname} is a string giving the name of the
7519 function. Should the return type of the function be required, you can
7520 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7521 constant pool that GCC wrote immediately before this call.
7523 If no constant-pool epilogue is required, the usual case, you need not
7527 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7528 Define this macro as a C expression which is nonzero if @var{C} is
7529 used as a logical line separator by the assembler. @var{STR} points
7530 to the position in the string where @var{C} was found; this can be used if
7531 a line separator uses multiple characters.
7533 If you do not define this macro, the default is that only
7534 the character @samp{;} is treated as a logical line separator.
7537 @hook TARGET_ASM_OPEN_PAREN
7538 These target hooks are C string constants, describing the syntax in the
7539 assembler for grouping arithmetic expressions. If not overridden, they
7540 default to normal parentheses, which is correct for most assemblers.
7543 These macros are provided by @file{real.h} for writing the definitions
7544 of @code{ASM_OUTPUT_DOUBLE} and the like:
7546 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7547 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7548 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7549 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7550 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7551 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7552 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7553 target's floating point representation, and store its bit pattern in
7554 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7555 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7556 simple @code{long int}. For the others, it should be an array of
7557 @code{long int}. The number of elements in this array is determined
7558 by the size of the desired target floating point data type: 32 bits of
7559 it go in each @code{long int} array element. Each array element holds
7560 32 bits of the result, even if @code{long int} is wider than 32 bits
7561 on the host machine.
7563 The array element values are designed so that you can print them out
7564 using @code{fprintf} in the order they should appear in the target
7568 @node Uninitialized Data
7569 @subsection Output of Uninitialized Variables
7571 Each of the macros in this section is used to do the whole job of
7572 outputting a single uninitialized variable.
7574 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7575 A C statement (sans semicolon) to output to the stdio stream
7576 @var{stream} the assembler definition of a common-label named
7577 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7578 is the size rounded up to whatever alignment the caller wants. It is
7579 possible that @var{size} may be zero, for instance if a struct with no
7580 other member than a zero-length array is defined. In this case, the
7581 backend must output a symbol definition that allocates at least one
7582 byte, both so that the address of the resulting object does not compare
7583 equal to any other, and because some object formats cannot even express
7584 the concept of a zero-sized common symbol, as that is how they represent
7585 an ordinary undefined external.
7587 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7588 output the name itself; before and after that, output the additional
7589 assembler syntax for defining the name, and a newline.
7591 This macro controls how the assembler definitions of uninitialized
7592 common global variables are output.
7595 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7596 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7597 separate, explicit argument. If you define this macro, it is used in
7598 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7599 handling the required alignment of the variable. The alignment is specified
7600 as the number of bits.
7603 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7604 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7605 variable to be output, if there is one, or @code{NULL_TREE} if there
7606 is no corresponding variable. If you define this macro, GCC will use it
7607 in place of both @code{ASM_OUTPUT_COMMON} and
7608 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7609 the variable's decl in order to chose what to output.
7612 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7613 A C statement (sans semicolon) to output to the stdio stream
7614 @var{stream} the assembler definition of uninitialized global @var{decl} named
7615 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
7616 is the alignment specified as the number of bits.
7618 Try to use function @code{asm_output_aligned_bss} defined in file
7619 @file{varasm.c} when defining this macro. If unable, use the expression
7620 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7621 before and after that, output the additional assembler syntax for defining
7622 the name, and a newline.
7624 There are two ways of handling global BSS@. One is to define this macro.
7625 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7626 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7627 You do not need to do both.
7629 Some languages do not have @code{common} data, and require a
7630 non-common form of global BSS in order to handle uninitialized globals
7631 efficiently. C++ is one example of this. However, if the target does
7632 not support global BSS, the front end may choose to make globals
7633 common in order to save space in the object file.
7636 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7637 A C statement (sans semicolon) to output to the stdio stream
7638 @var{stream} the assembler definition of a local-common-label named
7639 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7640 is the size rounded up to whatever alignment the caller wants.
7642 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7643 output the name itself; before and after that, output the additional
7644 assembler syntax for defining the name, and a newline.
7646 This macro controls how the assembler definitions of uninitialized
7647 static variables are output.
7650 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7651 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7652 separate, explicit argument. If you define this macro, it is used in
7653 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7654 handling the required alignment of the variable. The alignment is specified
7655 as the number of bits.
7658 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7659 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7660 variable to be output, if there is one, or @code{NULL_TREE} if there
7661 is no corresponding variable. If you define this macro, GCC will use it
7662 in place of both @code{ASM_OUTPUT_DECL} and
7663 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7664 the variable's decl in order to chose what to output.
7668 @subsection Output and Generation of Labels
7670 @c prevent bad page break with this line
7671 This is about outputting labels.
7673 @findex assemble_name
7674 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7675 A C statement (sans semicolon) to output to the stdio stream
7676 @var{stream} the assembler definition of a label named @var{name}.
7677 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7678 output the name itself; before and after that, output the additional
7679 assembler syntax for defining the name, and a newline. A default
7680 definition of this macro is provided which is correct for most systems.
7683 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7684 A C statement (sans semicolon) to output to the stdio stream
7685 @var{stream} the assembler definition of a label named @var{name} of
7687 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7688 output the name itself; before and after that, output the additional
7689 assembler syntax for defining the name, and a newline. A default
7690 definition of this macro is provided which is correct for most systems.
7692 If this macro is not defined, then the function name is defined in the
7693 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7696 @findex assemble_name_raw
7697 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7698 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7699 to refer to a compiler-generated label. The default definition uses
7700 @code{assemble_name_raw}, which is like @code{assemble_name} except
7701 that it is more efficient.
7705 A C string containing the appropriate assembler directive to specify the
7706 size of a symbol, without any arguments. On systems that use ELF, the
7707 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7708 systems, the default is not to define this macro.
7710 Define this macro only if it is correct to use the default definitions
7711 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7712 for your system. If you need your own custom definitions of those
7713 macros, or if you do not need explicit symbol sizes at all, do not
7717 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7718 A C statement (sans semicolon) to output to the stdio stream
7719 @var{stream} a directive telling the assembler that the size of the
7720 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7721 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7725 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7726 A C statement (sans semicolon) to output to the stdio stream
7727 @var{stream} a directive telling the assembler to calculate the size of
7728 the symbol @var{name} by subtracting its address from the current
7731 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7732 provided. The default assumes that the assembler recognizes a special
7733 @samp{.} symbol as referring to the current address, and can calculate
7734 the difference between this and another symbol. If your assembler does
7735 not recognize @samp{.} or cannot do calculations with it, you will need
7736 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7740 A C string containing the appropriate assembler directive to specify the
7741 type of a symbol, without any arguments. On systems that use ELF, the
7742 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7743 systems, the default is not to define this macro.
7745 Define this macro only if it is correct to use the default definition of
7746 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7747 custom definition of this macro, or if you do not need explicit symbol
7748 types at all, do not define this macro.
7751 @defmac TYPE_OPERAND_FMT
7752 A C string which specifies (using @code{printf} syntax) the format of
7753 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7754 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7755 the default is not to define this macro.
7757 Define this macro only if it is correct to use the default definition of
7758 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7759 custom definition of this macro, or if you do not need explicit symbol
7760 types at all, do not define this macro.
7763 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7764 A C statement (sans semicolon) to output to the stdio stream
7765 @var{stream} a directive telling the assembler that the type of the
7766 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7767 that string is always either @samp{"function"} or @samp{"object"}, but
7768 you should not count on this.
7770 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7771 definition of this macro is provided.
7774 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7775 A C statement (sans semicolon) to output to the stdio stream
7776 @var{stream} any text necessary for declaring the name @var{name} of a
7777 function which is being defined. This macro is responsible for
7778 outputting the label definition (perhaps using
7779 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7780 @code{FUNCTION_DECL} tree node representing the function.
7782 If this macro is not defined, then the function name is defined in the
7783 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7785 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7789 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7790 A C statement (sans semicolon) to output to the stdio stream
7791 @var{stream} any text necessary for declaring the size of a function
7792 which is being defined. The argument @var{name} is the name of the
7793 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7794 representing the function.
7796 If this macro is not defined, then the function size is not defined.
7798 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7802 @defmac ASM_DECLARE_OBJECT_NAME (@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 name @var{name} of an
7805 initialized variable which is being defined. This macro must output the
7806 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7807 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7809 If this macro is not defined, then the variable name is defined in the
7810 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7812 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7813 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7816 @hook TARGET_ASM_DECLARE_CONSTANT_NAME
7817 A target hook to output to the stdio stream @var{file} any text necessary
7818 for declaring the name @var{name} of a constant which is being defined. This
7819 target hook is responsible for outputting the label definition (perhaps using
7820 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7821 and @var{size} is the size of the constant in bytes. The @var{name}
7822 will be an internal label.
7824 The default version of this target hook, define the @var{name} in the
7825 usual manner as a label (by means of @code{assemble_label}).
7827 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7830 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7831 A C statement (sans semicolon) to output to the stdio stream
7832 @var{stream} any text necessary for claiming a register @var{regno}
7833 for a global variable @var{decl} with name @var{name}.
7835 If you don't define this macro, that is equivalent to defining it to do
7839 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7840 A C statement (sans semicolon) to finish up declaring a variable name
7841 once the compiler has processed its initializer fully and thus has had a
7842 chance to determine the size of an array when controlled by an
7843 initializer. This is used on systems where it's necessary to declare
7844 something about the size of the object.
7846 If you don't define this macro, that is equivalent to defining it to do
7849 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7850 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7853 @hook TARGET_ASM_GLOBALIZE_LABEL
7854 This target hook is a function to output to the stdio stream
7855 @var{stream} some commands that will make the label @var{name} global;
7856 that is, available for reference from other files.
7858 The default implementation relies on a proper definition of
7859 @code{GLOBAL_ASM_OP}.
7862 @hook TARGET_ASM_GLOBALIZE_DECL_NAME
7863 This target hook is a function to output to the stdio stream
7864 @var{stream} some commands that will make the name associated with @var{decl}
7865 global; that is, available for reference from other files.
7867 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7870 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7871 A C statement (sans semicolon) to output to the stdio stream
7872 @var{stream} some commands that will make the label @var{name} weak;
7873 that is, available for reference from other files but only used if
7874 no other definition is available. Use the expression
7875 @code{assemble_name (@var{stream}, @var{name})} to output the name
7876 itself; before and after that, output the additional assembler syntax
7877 for making that name weak, and a newline.
7879 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7880 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7884 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7885 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7886 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7887 or variable decl. If @var{value} is not @code{NULL}, this C statement
7888 should output to the stdio stream @var{stream} assembler code which
7889 defines (equates) the weak symbol @var{name} to have the value
7890 @var{value}. If @var{value} is @code{NULL}, it should output commands
7891 to make @var{name} weak.
7894 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7895 Outputs a directive that enables @var{name} to be used to refer to
7896 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7897 declaration of @code{name}.
7900 @defmac SUPPORTS_WEAK
7901 A preprocessor constant expression which evaluates to true if the target
7902 supports weak symbols.
7904 If you don't define this macro, @file{defaults.h} provides a default
7905 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7906 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
7909 @defmac TARGET_SUPPORTS_WEAK
7910 A C expression which evaluates to true if the target supports weak symbols.
7912 If you don't define this macro, @file{defaults.h} provides a default
7913 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
7914 this macro if you want to control weak symbol support with a compiler
7915 flag such as @option{-melf}.
7918 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7919 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7920 public symbol such that extra copies in multiple translation units will
7921 be discarded by the linker. Define this macro if your object file
7922 format provides support for this concept, such as the @samp{COMDAT}
7923 section flags in the Microsoft Windows PE/COFF format, and this support
7924 requires changes to @var{decl}, such as putting it in a separate section.
7927 @defmac SUPPORTS_ONE_ONLY
7928 A C expression which evaluates to true if the target supports one-only
7931 If you don't define this macro, @file{varasm.c} provides a default
7932 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7933 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7934 you want to control one-only symbol support with a compiler flag, or if
7935 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7936 be emitted as one-only.
7939 @hook TARGET_ASM_ASSEMBLE_VISIBILITY
7940 This target hook is a function to output to @var{asm_out_file} some
7941 commands that will make the symbol(s) associated with @var{decl} have
7942 hidden, protected or internal visibility as specified by @var{visibility}.
7945 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7946 A C expression that evaluates to true if the target's linker expects
7947 that weak symbols do not appear in a static archive's table of contents.
7948 The default is @code{0}.
7950 Leaving weak symbols out of an archive's table of contents means that,
7951 if a symbol will only have a definition in one translation unit and
7952 will have undefined references from other translation units, that
7953 symbol should not be weak. Defining this macro to be nonzero will
7954 thus have the effect that certain symbols that would normally be weak
7955 (explicit template instantiations, and vtables for polymorphic classes
7956 with noninline key methods) will instead be nonweak.
7958 The C++ ABI requires this macro to be zero. Define this macro for
7959 targets where full C++ ABI compliance is impossible and where linker
7960 restrictions require weak symbols to be left out of a static archive's
7964 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7965 A C statement (sans semicolon) to output to the stdio stream
7966 @var{stream} any text necessary for declaring the name of an external
7967 symbol named @var{name} which is referenced in this compilation but
7968 not defined. The value of @var{decl} is the tree node for the
7971 This macro need not be defined if it does not need to output anything.
7972 The GNU assembler and most Unix assemblers don't require anything.
7975 @hook TARGET_ASM_EXTERNAL_LIBCALL
7976 This target hook is a function to output to @var{asm_out_file} an assembler
7977 pseudo-op to declare a library function name external. The name of the
7978 library function is given by @var{symref}, which is a @code{symbol_ref}.
7981 @hook TARGET_ASM_MARK_DECL_PRESERVED
7982 This target hook is a function to output to @var{asm_out_file} an assembler
7983 directive to annotate @var{symbol} as used. The Darwin target uses the
7984 .no_dead_code_strip directive.
7987 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
7988 A C statement (sans semicolon) to output to the stdio stream
7989 @var{stream} a reference in assembler syntax to a label named
7990 @var{name}. This should add @samp{_} to the front of the name, if that
7991 is customary on your operating system, as it is in most Berkeley Unix
7992 systems. This macro is used in @code{assemble_name}.
7995 @hook TARGET_MANGLE_ASSEMBLER_NAME
7997 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
7998 A C statement (sans semicolon) to output a reference to
7999 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8000 will be used to output the name of the symbol. This macro may be used
8001 to modify the way a symbol is referenced depending on information
8002 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8005 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8006 A C statement (sans semicolon) to output a reference to @var{buf}, the
8007 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8008 @code{assemble_name} will be used to output the name of the symbol.
8009 This macro is not used by @code{output_asm_label}, or the @code{%l}
8010 specifier that calls it; the intention is that this macro should be set
8011 when it is necessary to output a label differently when its address is
8015 @hook TARGET_ASM_INTERNAL_LABEL
8016 A function to output to the stdio stream @var{stream} a label whose
8017 name is made from the string @var{prefix} and the number @var{labelno}.
8019 It is absolutely essential that these labels be distinct from the labels
8020 used for user-level functions and variables. Otherwise, certain programs
8021 will have name conflicts with internal labels.
8023 It is desirable to exclude internal labels from the symbol table of the
8024 object file. Most assemblers have a naming convention for labels that
8025 should be excluded; on many systems, the letter @samp{L} at the
8026 beginning of a label has this effect. You should find out what
8027 convention your system uses, and follow it.
8029 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8032 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8033 A C statement to output to the stdio stream @var{stream} a debug info
8034 label whose name is made from the string @var{prefix} and the number
8035 @var{num}. This is useful for VLIW targets, where debug info labels
8036 may need to be treated differently than branch target labels. On some
8037 systems, branch target labels must be at the beginning of instruction
8038 bundles, but debug info labels can occur in the middle of instruction
8041 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8045 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8046 A C statement to store into the string @var{string} a label whose name
8047 is made from the string @var{prefix} and the number @var{num}.
8049 This string, when output subsequently by @code{assemble_name}, should
8050 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8051 with the same @var{prefix} and @var{num}.
8053 If the string begins with @samp{*}, then @code{assemble_name} will
8054 output the rest of the string unchanged. It is often convenient for
8055 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8056 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8057 to output the string, and may change it. (Of course,
8058 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8059 you should know what it does on your machine.)
8062 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8063 A C expression to assign to @var{outvar} (which is a variable of type
8064 @code{char *}) a newly allocated string made from the string
8065 @var{name} and the number @var{number}, with some suitable punctuation
8066 added. Use @code{alloca} to get space for the string.
8068 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8069 produce an assembler label for an internal static variable whose name is
8070 @var{name}. Therefore, the string must be such as to result in valid
8071 assembler code. The argument @var{number} is different each time this
8072 macro is executed; it prevents conflicts between similarly-named
8073 internal static variables in different scopes.
8075 Ideally this string should not be a valid C identifier, to prevent any
8076 conflict with the user's own symbols. Most assemblers allow periods
8077 or percent signs in assembler symbols; putting at least one of these
8078 between the name and the number will suffice.
8080 If this macro is not defined, a default definition will be provided
8081 which is correct for most systems.
8084 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8085 A C statement to output to the stdio stream @var{stream} assembler code
8086 which defines (equates) the symbol @var{name} to have the value @var{value}.
8089 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8090 correct for most systems.
8093 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8094 A C statement to output to the stdio stream @var{stream} assembler code
8095 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8096 to have the value of the tree node @var{decl_of_value}. This macro will
8097 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8098 the tree nodes are available.
8101 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8102 correct for most systems.
8105 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8106 A C statement that evaluates to true if the assembler code which defines
8107 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8108 of the tree node @var{decl_of_value} should be emitted near the end of the
8109 current compilation unit. The default is to not defer output of defines.
8110 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8111 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8114 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8115 A C statement to output to the stdio stream @var{stream} assembler code
8116 which defines (equates) the weak symbol @var{name} to have the value
8117 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8118 an undefined weak symbol.
8120 Define this macro if the target only supports weak aliases; define
8121 @code{ASM_OUTPUT_DEF} instead if possible.
8124 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8125 Define this macro to override the default assembler names used for
8126 Objective-C methods.
8128 The default name is a unique method number followed by the name of the
8129 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8130 the category is also included in the assembler name (e.g.@:
8133 These names are safe on most systems, but make debugging difficult since
8134 the method's selector is not present in the name. Therefore, particular
8135 systems define other ways of computing names.
8137 @var{buf} is an expression of type @code{char *} which gives you a
8138 buffer in which to store the name; its length is as long as
8139 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8140 50 characters extra.
8142 The argument @var{is_inst} specifies whether the method is an instance
8143 method or a class method; @var{class_name} is the name of the class;
8144 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8145 in a category); and @var{sel_name} is the name of the selector.
8147 On systems where the assembler can handle quoted names, you can use this
8148 macro to provide more human-readable names.
8151 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
8152 A C statement (sans semicolon) to output to the stdio stream
8153 @var{stream} commands to declare that the label @var{name} is an
8154 Objective-C class reference. This is only needed for targets whose
8155 linkers have special support for NeXT-style runtimes.
8158 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
8159 A C statement (sans semicolon) to output to the stdio stream
8160 @var{stream} commands to declare that the label @var{name} is an
8161 unresolved Objective-C class reference. This is only needed for targets
8162 whose linkers have special support for NeXT-style runtimes.
8165 @node Initialization
8166 @subsection How Initialization Functions Are Handled
8167 @cindex initialization routines
8168 @cindex termination routines
8169 @cindex constructors, output of
8170 @cindex destructors, output of
8172 The compiled code for certain languages includes @dfn{constructors}
8173 (also called @dfn{initialization routines})---functions to initialize
8174 data in the program when the program is started. These functions need
8175 to be called before the program is ``started''---that is to say, before
8176 @code{main} is called.
8178 Compiling some languages generates @dfn{destructors} (also called
8179 @dfn{termination routines}) that should be called when the program
8182 To make the initialization and termination functions work, the compiler
8183 must output something in the assembler code to cause those functions to
8184 be called at the appropriate time. When you port the compiler to a new
8185 system, you need to specify how to do this.
8187 There are two major ways that GCC currently supports the execution of
8188 initialization and termination functions. Each way has two variants.
8189 Much of the structure is common to all four variations.
8191 @findex __CTOR_LIST__
8192 @findex __DTOR_LIST__
8193 The linker must build two lists of these functions---a list of
8194 initialization functions, called @code{__CTOR_LIST__}, and a list of
8195 termination functions, called @code{__DTOR_LIST__}.
8197 Each list always begins with an ignored function pointer (which may hold
8198 0, @minus{}1, or a count of the function pointers after it, depending on
8199 the environment). This is followed by a series of zero or more function
8200 pointers to constructors (or destructors), followed by a function
8201 pointer containing zero.
8203 Depending on the operating system and its executable file format, either
8204 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8205 time and exit time. Constructors are called in reverse order of the
8206 list; destructors in forward order.
8208 The best way to handle static constructors works only for object file
8209 formats which provide arbitrarily-named sections. A section is set
8210 aside for a list of constructors, and another for a list of destructors.
8211 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8212 object file that defines an initialization function also puts a word in
8213 the constructor section to point to that function. The linker
8214 accumulates all these words into one contiguous @samp{.ctors} section.
8215 Termination functions are handled similarly.
8217 This method will be chosen as the default by @file{target-def.h} if
8218 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8219 support arbitrary sections, but does support special designated
8220 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8221 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8223 When arbitrary sections are available, there are two variants, depending
8224 upon how the code in @file{crtstuff.c} is called. On systems that
8225 support a @dfn{.init} section which is executed at program startup,
8226 parts of @file{crtstuff.c} are compiled into that section. The
8227 program is linked by the @command{gcc} driver like this:
8230 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8233 The prologue of a function (@code{__init}) appears in the @code{.init}
8234 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8235 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8236 files are provided by the operating system or by the GNU C library, but
8237 are provided by GCC for a few targets.
8239 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8240 compiled from @file{crtstuff.c}. They contain, among other things, code
8241 fragments within the @code{.init} and @code{.fini} sections that branch
8242 to routines in the @code{.text} section. The linker will pull all parts
8243 of a section together, which results in a complete @code{__init} function
8244 that invokes the routines we need at startup.
8246 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8249 If no init section is available, when GCC compiles any function called
8250 @code{main} (or more accurately, any function designated as a program
8251 entry point by the language front end calling @code{expand_main_function}),
8252 it inserts a procedure call to @code{__main} as the first executable code
8253 after the function prologue. The @code{__main} function is defined
8254 in @file{libgcc2.c} and runs the global constructors.
8256 In file formats that don't support arbitrary sections, there are again
8257 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8258 and an `a.out' format must be used. In this case,
8259 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8260 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8261 and with the address of the void function containing the initialization
8262 code as its value. The GNU linker recognizes this as a request to add
8263 the value to a @dfn{set}; the values are accumulated, and are eventually
8264 placed in the executable as a vector in the format described above, with
8265 a leading (ignored) count and a trailing zero element.
8266 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8267 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8268 the compilation of @code{main} to call @code{__main} as above, starting
8269 the initialization process.
8271 The last variant uses neither arbitrary sections nor the GNU linker.
8272 This is preferable when you want to do dynamic linking and when using
8273 file formats which the GNU linker does not support, such as `ECOFF'@. In
8274 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8275 termination functions are recognized simply by their names. This requires
8276 an extra program in the linkage step, called @command{collect2}. This program
8277 pretends to be the linker, for use with GCC; it does its job by running
8278 the ordinary linker, but also arranges to include the vectors of
8279 initialization and termination functions. These functions are called
8280 via @code{__main} as described above. In order to use this method,
8281 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8284 The following section describes the specific macros that control and
8285 customize the handling of initialization and termination functions.
8288 @node Macros for Initialization
8289 @subsection Macros Controlling Initialization Routines
8291 Here are the macros that control how the compiler handles initialization
8292 and termination functions:
8294 @defmac INIT_SECTION_ASM_OP
8295 If defined, a C string constant, including spacing, for the assembler
8296 operation to identify the following data as initialization code. If not
8297 defined, GCC will assume such a section does not exist. When you are
8298 using special sections for initialization and termination functions, this
8299 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8300 run the initialization functions.
8303 @defmac HAS_INIT_SECTION
8304 If defined, @code{main} will not call @code{__main} as described above.
8305 This macro should be defined for systems that control start-up code
8306 on a symbol-by-symbol basis, such as OSF/1, and should not
8307 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8310 @defmac LD_INIT_SWITCH
8311 If defined, a C string constant for a switch that tells the linker that
8312 the following symbol is an initialization routine.
8315 @defmac LD_FINI_SWITCH
8316 If defined, a C string constant for a switch that tells the linker that
8317 the following symbol is a finalization routine.
8320 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8321 If defined, a C statement that will write a function that can be
8322 automatically called when a shared library is loaded. The function
8323 should call @var{func}, which takes no arguments. If not defined, and
8324 the object format requires an explicit initialization function, then a
8325 function called @code{_GLOBAL__DI} will be generated.
8327 This function and the following one are used by collect2 when linking a
8328 shared library that needs constructors or destructors, or has DWARF2
8329 exception tables embedded in the code.
8332 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8333 If defined, a C statement that will write a function that can be
8334 automatically called when a shared library is unloaded. The function
8335 should call @var{func}, which takes no arguments. If not defined, and
8336 the object format requires an explicit finalization function, then a
8337 function called @code{_GLOBAL__DD} will be generated.
8340 @defmac INVOKE__main
8341 If defined, @code{main} will call @code{__main} despite the presence of
8342 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8343 where the init section is not actually run automatically, but is still
8344 useful for collecting the lists of constructors and destructors.
8347 @defmac SUPPORTS_INIT_PRIORITY
8348 If nonzero, the C++ @code{init_priority} attribute is supported and the
8349 compiler should emit instructions to control the order of initialization
8350 of objects. If zero, the compiler will issue an error message upon
8351 encountering an @code{init_priority} attribute.
8354 @hook TARGET_HAVE_CTORS_DTORS
8355 This value is true if the target supports some ``native'' method of
8356 collecting constructors and destructors to be run at startup and exit.
8357 It is false if we must use @command{collect2}.
8360 @hook TARGET_ASM_CONSTRUCTOR
8361 If defined, a function that outputs assembler code to arrange to call
8362 the function referenced by @var{symbol} at initialization time.
8364 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8365 no arguments and with no return value. If the target supports initialization
8366 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8367 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8369 If this macro is not defined by the target, a suitable default will
8370 be chosen if (1) the target supports arbitrary section names, (2) the
8371 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8375 @hook TARGET_ASM_DESTRUCTOR
8376 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8377 functions rather than initialization functions.
8380 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8381 generated for the generated object file will have static linkage.
8383 If your system uses @command{collect2} as the means of processing
8384 constructors, then that program normally uses @command{nm} to scan
8385 an object file for constructor functions to be called.
8387 On certain kinds of systems, you can define this macro to make
8388 @command{collect2} work faster (and, in some cases, make it work at all):
8390 @defmac OBJECT_FORMAT_COFF
8391 Define this macro if the system uses COFF (Common Object File Format)
8392 object files, so that @command{collect2} can assume this format and scan
8393 object files directly for dynamic constructor/destructor functions.
8395 This macro is effective only in a native compiler; @command{collect2} as
8396 part of a cross compiler always uses @command{nm} for the target machine.
8399 @defmac REAL_NM_FILE_NAME
8400 Define this macro as a C string constant containing the file name to use
8401 to execute @command{nm}. The default is to search the path normally for
8406 @command{collect2} calls @command{nm} to scan object files for static
8407 constructors and destructors and LTO info. By default, @option{-n} is
8408 passed. Define @code{NM_FLAGS} to a C string constant if other options
8409 are needed to get the same output format as GNU @command{nm -n}
8413 If your system supports shared libraries and has a program to list the
8414 dynamic dependencies of a given library or executable, you can define
8415 these macros to enable support for running initialization and
8416 termination functions in shared libraries:
8419 Define this macro to a C string constant containing the name of the program
8420 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8423 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8424 Define this macro to be C code that extracts filenames from the output
8425 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8426 of type @code{char *} that points to the beginning of a line of output
8427 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8428 code must advance @var{ptr} to the beginning of the filename on that
8429 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8432 @defmac SHLIB_SUFFIX
8433 Define this macro to a C string constant containing the default shared
8434 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8435 strips version information after this suffix when generating global
8436 constructor and destructor names. This define is only needed on targets
8437 that use @command{collect2} to process constructors and destructors.
8440 @node Instruction Output
8441 @subsection Output of Assembler Instructions
8443 @c prevent bad page break with this line
8444 This describes assembler instruction output.
8446 @defmac REGISTER_NAMES
8447 A C initializer containing the assembler's names for the machine
8448 registers, each one as a C string constant. This is what translates
8449 register numbers in the compiler into assembler language.
8452 @defmac ADDITIONAL_REGISTER_NAMES
8453 If defined, a C initializer for an array of structures containing a name
8454 and a register number. This macro defines additional names for hard
8455 registers, thus allowing the @code{asm} option in declarations to refer
8456 to registers using alternate names.
8459 @defmac OVERLAPPING_REGISTER_NAMES
8460 If defined, a C initializer for an array of structures containing a
8461 name, a register number and a count of the number of consecutive
8462 machine registers the name overlaps. This macro defines additional
8463 names for hard registers, thus allowing the @code{asm} option in
8464 declarations to refer to registers using alternate names. Unlike
8465 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8466 register name implies multiple underlying registers.
8468 This macro should be used when it is important that a clobber in an
8469 @code{asm} statement clobbers all the underlying values implied by the
8470 register name. For example, on ARM, clobbering the double-precision
8471 VFP register ``d0'' implies clobbering both single-precision registers
8475 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8476 Define this macro if you are using an unusual assembler that
8477 requires different names for the machine instructions.
8479 The definition is a C statement or statements which output an
8480 assembler instruction opcode to the stdio stream @var{stream}. The
8481 macro-operand @var{ptr} is a variable of type @code{char *} which
8482 points to the opcode name in its ``internal'' form---the form that is
8483 written in the machine description. The definition should output the
8484 opcode name to @var{stream}, performing any translation you desire, and
8485 increment the variable @var{ptr} to point at the end of the opcode
8486 so that it will not be output twice.
8488 In fact, your macro definition may process less than the entire opcode
8489 name, or more than the opcode name; but if you want to process text
8490 that includes @samp{%}-sequences to substitute operands, you must take
8491 care of the substitution yourself. Just be sure to increment
8492 @var{ptr} over whatever text should not be output normally.
8494 @findex recog_data.operand
8495 If you need to look at the operand values, they can be found as the
8496 elements of @code{recog_data.operand}.
8498 If the macro definition does nothing, the instruction is output
8502 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8503 If defined, a C statement to be executed just prior to the output of
8504 assembler code for @var{insn}, to modify the extracted operands so
8505 they will be output differently.
8507 Here the argument @var{opvec} is the vector containing the operands
8508 extracted from @var{insn}, and @var{noperands} is the number of
8509 elements of the vector which contain meaningful data for this insn.
8510 The contents of this vector are what will be used to convert the insn
8511 template into assembler code, so you can change the assembler output
8512 by changing the contents of the vector.
8514 This macro is useful when various assembler syntaxes share a single
8515 file of instruction patterns; by defining this macro differently, you
8516 can cause a large class of instructions to be output differently (such
8517 as with rearranged operands). Naturally, variations in assembler
8518 syntax affecting individual insn patterns ought to be handled by
8519 writing conditional output routines in those patterns.
8521 If this macro is not defined, it is equivalent to a null statement.
8524 @hook TARGET_ASM_FINAL_POSTSCAN_INSN
8525 If defined, this target hook is a function which is executed just after the
8526 output of assembler code for @var{insn}, to change the mode of the assembler
8529 Here the argument @var{opvec} is the vector containing the operands
8530 extracted from @var{insn}, and @var{noperands} is the number of
8531 elements of the vector which contain meaningful data for this insn.
8532 The contents of this vector are what was used to convert the insn
8533 template into assembler code, so you can change the assembler mode
8534 by checking the contents of the vector.
8537 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8538 A C compound statement to output to stdio stream @var{stream} the
8539 assembler syntax for an instruction operand @var{x}. @var{x} is an
8542 @var{code} is a value that can be used to specify one of several ways
8543 of printing the operand. It is used when identical operands must be
8544 printed differently depending on the context. @var{code} comes from
8545 the @samp{%} specification that was used to request printing of the
8546 operand. If the specification was just @samp{%@var{digit}} then
8547 @var{code} is 0; if the specification was @samp{%@var{ltr}
8548 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8551 If @var{x} is a register, this macro should print the register's name.
8552 The names can be found in an array @code{reg_names} whose type is
8553 @code{char *[]}. @code{reg_names} is initialized from
8554 @code{REGISTER_NAMES}.
8556 When the machine description has a specification @samp{%@var{punct}}
8557 (a @samp{%} followed by a punctuation character), this macro is called
8558 with a null pointer for @var{x} and the punctuation character for
8562 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8563 A C expression which evaluates to true if @var{code} is a valid
8564 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8565 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8566 punctuation characters (except for the standard one, @samp{%}) are used
8570 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8571 A C compound statement to output to stdio stream @var{stream} the
8572 assembler syntax for an instruction operand that is a memory reference
8573 whose address is @var{x}. @var{x} is an RTL expression.
8575 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8576 On some machines, the syntax for a symbolic address depends on the
8577 section that the address refers to. On these machines, define the hook
8578 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8579 @code{symbol_ref}, and then check for it here. @xref{Assembler
8583 @findex dbr_sequence_length
8584 @defmac DBR_OUTPUT_SEQEND (@var{file})
8585 A C statement, to be executed after all slot-filler instructions have
8586 been output. If necessary, call @code{dbr_sequence_length} to
8587 determine the number of slots filled in a sequence (zero if not
8588 currently outputting a sequence), to decide how many no-ops to output,
8591 Don't define this macro if it has nothing to do, but it is helpful in
8592 reading assembly output if the extent of the delay sequence is made
8593 explicit (e.g.@: with white space).
8596 @findex final_sequence
8597 Note that output routines for instructions with delay slots must be
8598 prepared to deal with not being output as part of a sequence
8599 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8600 found.) The variable @code{final_sequence} is null when not
8601 processing a sequence, otherwise it contains the @code{sequence} rtx
8605 @defmac REGISTER_PREFIX
8606 @defmacx LOCAL_LABEL_PREFIX
8607 @defmacx USER_LABEL_PREFIX
8608 @defmacx IMMEDIATE_PREFIX
8609 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8610 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8611 @file{final.c}). These are useful when a single @file{md} file must
8612 support multiple assembler formats. In that case, the various @file{tm.h}
8613 files can define these macros differently.
8616 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8617 If defined this macro should expand to a series of @code{case}
8618 statements which will be parsed inside the @code{switch} statement of
8619 the @code{asm_fprintf} function. This allows targets to define extra
8620 printf formats which may useful when generating their assembler
8621 statements. Note that uppercase letters are reserved for future
8622 generic extensions to asm_fprintf, and so are not available to target
8623 specific code. The output file is given by the parameter @var{file}.
8624 The varargs input pointer is @var{argptr} and the rest of the format
8625 string, starting the character after the one that is being switched
8626 upon, is pointed to by @var{format}.
8629 @defmac ASSEMBLER_DIALECT
8630 If your target supports multiple dialects of assembler language (such as
8631 different opcodes), define this macro as a C expression that gives the
8632 numeric index of the assembler language dialect to use, with zero as the
8635 If this macro is defined, you may use constructs of the form
8637 @samp{@{option0|option1|option2@dots{}@}}
8640 in the output templates of patterns (@pxref{Output Template}) or in the
8641 first argument of @code{asm_fprintf}. This construct outputs
8642 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8643 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8644 within these strings retain their usual meaning. If there are fewer
8645 alternatives within the braces than the value of
8646 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8648 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8649 @samp{@}} do not have any special meaning when used in templates or
8650 operands to @code{asm_fprintf}.
8652 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8653 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8654 the variations in assembler language syntax with that mechanism. Define
8655 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8656 if the syntax variant are larger and involve such things as different
8657 opcodes or operand order.
8660 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8661 A C expression to output to @var{stream} some assembler code
8662 which will push hard register number @var{regno} onto the stack.
8663 The code need not be optimal, since this macro is used only when
8667 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8668 A C expression to output to @var{stream} some assembler code
8669 which will pop hard register number @var{regno} off of the stack.
8670 The code need not be optimal, since this macro is used only when
8674 @node Dispatch Tables
8675 @subsection Output of Dispatch Tables
8677 @c prevent bad page break with this line
8678 This concerns dispatch tables.
8680 @cindex dispatch table
8681 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8682 A C statement to output to the stdio stream @var{stream} an assembler
8683 pseudo-instruction to generate a difference between two labels.
8684 @var{value} and @var{rel} are the numbers of two internal labels. The
8685 definitions of these labels are output using
8686 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8687 way here. For example,
8690 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8691 @var{value}, @var{rel})
8694 You must provide this macro on machines where the addresses in a
8695 dispatch table are relative to the table's own address. If defined, GCC
8696 will also use this macro on all machines when producing PIC@.
8697 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8698 mode and flags can be read.
8701 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8702 This macro should be provided on machines where the addresses
8703 in a dispatch table are absolute.
8705 The definition should be a C statement to output to the stdio stream
8706 @var{stream} an assembler pseudo-instruction to generate a reference to
8707 a label. @var{value} is the number of an internal label whose
8708 definition is output using @code{(*targetm.asm_out.internal_label)}.
8712 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8716 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8717 Define this if the label before a jump-table needs to be output
8718 specially. The first three arguments are the same as for
8719 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8720 jump-table which follows (a @code{jump_insn} containing an
8721 @code{addr_vec} or @code{addr_diff_vec}).
8723 This feature is used on system V to output a @code{swbeg} statement
8726 If this macro is not defined, these labels are output with
8727 @code{(*targetm.asm_out.internal_label)}.
8730 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8731 Define this if something special must be output at the end of a
8732 jump-table. The definition should be a C statement to be executed
8733 after the assembler code for the table is written. It should write
8734 the appropriate code to stdio stream @var{stream}. The argument
8735 @var{table} is the jump-table insn, and @var{num} is the label-number
8736 of the preceding label.
8738 If this macro is not defined, nothing special is output at the end of
8742 @hook TARGET_ASM_EMIT_UNWIND_LABEL
8743 This target hook emits a label at the beginning of each FDE@. It
8744 should be defined on targets where FDEs need special labels, and it
8745 should write the appropriate label, for the FDE associated with the
8746 function declaration @var{decl}, to the stdio stream @var{stream}.
8747 The third argument, @var{for_eh}, is a boolean: true if this is for an
8748 exception table. The fourth argument, @var{empty}, is a boolean:
8749 true if this is a placeholder label for an omitted FDE@.
8751 The default is that FDEs are not given nonlocal labels.
8754 @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
8755 This target hook emits a label at the beginning of the exception table.
8756 It should be defined on targets where it is desirable for the table
8757 to be broken up according to function.
8759 The default is that no label is emitted.
8762 @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
8764 @hook TARGET_ASM_UNWIND_EMIT
8765 This target hook emits assembly directives required to unwind the
8766 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8767 returns @code{UI_TARGET}.
8770 @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8772 @node Exception Region Output
8773 @subsection Assembler Commands for Exception Regions
8775 @c prevent bad page break with this line
8777 This describes commands marking the start and the end of an exception
8780 @defmac EH_FRAME_SECTION_NAME
8781 If defined, a C string constant for the name of the section containing
8782 exception handling frame unwind information. If not defined, GCC will
8783 provide a default definition if the target supports named sections.
8784 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8786 You should define this symbol if your target supports DWARF 2 frame
8787 unwind information and the default definition does not work.
8790 @defmac EH_FRAME_IN_DATA_SECTION
8791 If defined, DWARF 2 frame unwind information will be placed in the
8792 data section even though the target supports named sections. This
8793 might be necessary, for instance, if the system linker does garbage
8794 collection and sections cannot be marked as not to be collected.
8796 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8800 @defmac EH_TABLES_CAN_BE_READ_ONLY
8801 Define this macro to 1 if your target is such that no frame unwind
8802 information encoding used with non-PIC code will ever require a
8803 runtime relocation, but the linker may not support merging read-only
8804 and read-write sections into a single read-write section.
8807 @defmac MASK_RETURN_ADDR
8808 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8809 that it does not contain any extraneous set bits in it.
8812 @defmac DWARF2_UNWIND_INFO
8813 Define this macro to 0 if your target supports DWARF 2 frame unwind
8814 information, but it does not yet work with exception handling.
8815 Otherwise, if your target supports this information (if it defines
8816 @code{INCOMING_RETURN_ADDR_RTX} and either @code{UNALIGNED_INT_ASM_OP}
8817 or @code{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8820 @hook TARGET_EXCEPT_UNWIND_INFO
8821 This hook defines the mechanism that will be used for exception handling
8822 by the target. If the target has ABI specified unwind tables, the hook
8823 should return @code{UI_TARGET}. If the target is to use the
8824 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8825 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8826 information, the hook should return @code{UI_DWARF2}.
8828 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8829 This may end up simplifying other parts of target-specific code. The
8830 default implementation of this hook never returns @code{UI_NONE}.
8832 Note that the value returned by this hook should be constant. It should
8833 not depend on anything except the command-line switches described by
8834 @var{opts}. In particular, the
8835 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8836 macros and builtin functions related to exception handling are set up
8837 depending on this setting.
8839 The default implementation of the hook first honors the
8840 @option{--enable-sjlj-exceptions} configure option, then
8841 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
8842 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
8843 must define this hook so that @var{opts} is used correctly.
8846 @hook TARGET_UNWIND_TABLES_DEFAULT
8847 This variable should be set to @code{true} if the target ABI requires unwinding
8848 tables even when exceptions are not used. It must not be modified by
8849 command-line option processing.
8852 @defmac DONT_USE_BUILTIN_SETJMP
8853 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8854 should use the @code{setjmp}/@code{longjmp} functions from the C library
8855 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8858 @defmac DWARF_CIE_DATA_ALIGNMENT
8859 This macro need only be defined if the target might save registers in the
8860 function prologue at an offset to the stack pointer that is not aligned to
8861 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8862 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8863 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8864 the target supports DWARF 2 frame unwind information.
8867 @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
8868 Contains the value true if the target should add a zero word onto the
8869 end of a Dwarf-2 frame info section when used for exception handling.
8870 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8874 @hook TARGET_DWARF_REGISTER_SPAN
8875 Given a register, this hook should return a parallel of registers to
8876 represent where to find the register pieces. Define this hook if the
8877 register and its mode are represented in Dwarf in non-contiguous
8878 locations, or if the register should be represented in more than one
8879 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8880 If not defined, the default is to return @code{NULL_RTX}.
8883 @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
8884 If some registers are represented in Dwarf-2 unwind information in
8885 multiple pieces, define this hook to fill in information about the
8886 sizes of those pieces in the table used by the unwinder at runtime.
8887 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8888 filling in a single size corresponding to each hard register;
8889 @var{address} is the address of the table.
8892 @hook TARGET_ASM_TTYPE
8893 This hook is used to output a reference from a frame unwinding table to
8894 the type_info object identified by @var{sym}. It should return @code{true}
8895 if the reference was output. Returning @code{false} will cause the
8896 reference to be output using the normal Dwarf2 routines.
8899 @hook TARGET_ARM_EABI_UNWINDER
8900 This flag should be set to @code{true} on targets that use an ARM EABI
8901 based unwinding library, and @code{false} on other targets. This effects
8902 the format of unwinding tables, and how the unwinder in entered after
8903 running a cleanup. The default is @code{false}.
8906 @node Alignment Output
8907 @subsection Assembler Commands for Alignment
8909 @c prevent bad page break with this line
8910 This describes commands for alignment.
8912 @defmac JUMP_ALIGN (@var{label})
8913 The alignment (log base 2) to put in front of @var{label}, which is
8914 a common destination of jumps and has no fallthru incoming edge.
8916 This macro need not be defined if you don't want any special alignment
8917 to be done at such a time. Most machine descriptions do not currently
8920 Unless it's necessary to inspect the @var{label} parameter, it is better
8921 to set the variable @var{align_jumps} in the target's
8922 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8923 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8926 @hook TARGET_ASM_JUMP_ALIGN_MAX_SKIP
8927 The maximum number of bytes to skip before @var{label} when applying
8928 @code{JUMP_ALIGN}. This works only if
8929 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8932 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8933 The alignment (log base 2) to put in front of @var{label}, which follows
8936 This macro need not be defined if you don't want any special alignment
8937 to be done at such a time. Most machine descriptions do not currently
8941 @hook TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8942 The maximum number of bytes to skip before @var{label} when applying
8943 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8944 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8947 @defmac LOOP_ALIGN (@var{label})
8948 The alignment (log base 2) to put in front of @var{label}, which follows
8949 a @code{NOTE_INSN_LOOP_BEG} note.
8951 This macro need not be defined if you don't want any special alignment
8952 to be done at such a time. Most machine descriptions do not currently
8955 Unless it's necessary to inspect the @var{label} parameter, it is better
8956 to set the variable @code{align_loops} in the target's
8957 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8958 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8961 @hook TARGET_ASM_LOOP_ALIGN_MAX_SKIP
8962 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
8963 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
8967 @defmac LABEL_ALIGN (@var{label})
8968 The alignment (log base 2) to put in front of @var{label}.
8969 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8970 the maximum of the specified values is used.
8972 Unless it's necessary to inspect the @var{label} parameter, it is better
8973 to set the variable @code{align_labels} in the target's
8974 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8975 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8978 @hook TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8979 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
8980 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
8984 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8985 A C statement to output to the stdio stream @var{stream} an assembler
8986 instruction to advance the location counter by @var{nbytes} bytes.
8987 Those bytes should be zero when loaded. @var{nbytes} will be a C
8988 expression of type @code{unsigned HOST_WIDE_INT}.
8991 @defmac ASM_NO_SKIP_IN_TEXT
8992 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
8993 text section because it fails to put zeros in the bytes that are skipped.
8994 This is true on many Unix systems, where the pseudo--op to skip bytes
8995 produces no-op instructions rather than zeros when used in the text
8999 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9000 A C statement to output to the stdio stream @var{stream} an assembler
9001 command to advance the location counter to a multiple of 2 to the
9002 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9005 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9006 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9007 for padding, if necessary.
9010 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9011 A C statement to output to the stdio stream @var{stream} an assembler
9012 command to advance the location counter to a multiple of 2 to the
9013 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9014 satisfy the alignment request. @var{power} and @var{max_skip} will be
9015 a C expression of type @code{int}.
9019 @node Debugging Info
9020 @section Controlling Debugging Information Format
9022 @c prevent bad page break with this line
9023 This describes how to specify debugging information.
9026 * All Debuggers:: Macros that affect all debugging formats uniformly.
9027 * DBX Options:: Macros enabling specific options in DBX format.
9028 * DBX Hooks:: Hook macros for varying DBX format.
9029 * File Names and DBX:: Macros controlling output of file names in DBX format.
9030 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9031 * VMS Debug:: Macros for VMS debug format.
9035 @subsection Macros Affecting All Debugging Formats
9037 @c prevent bad page break with this line
9038 These macros affect all debugging formats.
9040 @defmac DBX_REGISTER_NUMBER (@var{regno})
9041 A C expression that returns the DBX register number for the compiler
9042 register number @var{regno}. In the default macro provided, the value
9043 of this expression will be @var{regno} itself. But sometimes there are
9044 some registers that the compiler knows about and DBX does not, or vice
9045 versa. In such cases, some register may need to have one number in the
9046 compiler and another for DBX@.
9048 If two registers have consecutive numbers inside GCC, and they can be
9049 used as a pair to hold a multiword value, then they @emph{must} have
9050 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9051 Otherwise, debuggers will be unable to access such a pair, because they
9052 expect register pairs to be consecutive in their own numbering scheme.
9054 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9055 does not preserve register pairs, then what you must do instead is
9056 redefine the actual register numbering scheme.
9059 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9060 A C expression that returns the integer offset value for an automatic
9061 variable having address @var{x} (an RTL expression). The default
9062 computation assumes that @var{x} is based on the frame-pointer and
9063 gives the offset from the frame-pointer. This is required for targets
9064 that produce debugging output for DBX or COFF-style debugging output
9065 for SDB and allow the frame-pointer to be eliminated when the
9066 @option{-g} options is used.
9069 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9070 A C expression that returns the integer offset value for an argument
9071 having address @var{x} (an RTL expression). The nominal offset is
9075 @defmac PREFERRED_DEBUGGING_TYPE
9076 A C expression that returns the type of debugging output GCC should
9077 produce when the user specifies just @option{-g}. Define
9078 this if you have arranged for GCC to support more than one format of
9079 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9080 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9081 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9083 When the user specifies @option{-ggdb}, GCC normally also uses the
9084 value of this macro to select the debugging output format, but with two
9085 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9086 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9087 defined, GCC uses @code{DBX_DEBUG}.
9089 The value of this macro only affects the default debugging output; the
9090 user can always get a specific type of output by using @option{-gstabs},
9091 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9095 @subsection Specific Options for DBX Output
9097 @c prevent bad page break with this line
9098 These are specific options for DBX output.
9100 @defmac DBX_DEBUGGING_INFO
9101 Define this macro if GCC should produce debugging output for DBX
9102 in response to the @option{-g} option.
9105 @defmac XCOFF_DEBUGGING_INFO
9106 Define this macro if GCC should produce XCOFF format debugging output
9107 in response to the @option{-g} option. This is a variant of DBX format.
9110 @defmac DEFAULT_GDB_EXTENSIONS
9111 Define this macro to control whether GCC should by default generate
9112 GDB's extended version of DBX debugging information (assuming DBX-format
9113 debugging information is enabled at all). If you don't define the
9114 macro, the default is 1: always generate the extended information
9115 if there is any occasion to.
9118 @defmac DEBUG_SYMS_TEXT
9119 Define this macro if all @code{.stabs} commands should be output while
9120 in the text section.
9123 @defmac ASM_STABS_OP
9124 A C string constant, including spacing, naming the assembler pseudo op to
9125 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9126 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9127 applies only to DBX debugging information format.
9130 @defmac ASM_STABD_OP
9131 A C string constant, including spacing, naming the assembler pseudo op to
9132 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9133 value is the current location. If you don't define this macro,
9134 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9138 @defmac ASM_STABN_OP
9139 A C string constant, including spacing, naming the assembler pseudo op to
9140 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9141 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9142 macro applies only to DBX debugging information format.
9145 @defmac DBX_NO_XREFS
9146 Define this macro if DBX on your system does not support the construct
9147 @samp{xs@var{tagname}}. On some systems, this construct is used to
9148 describe a forward reference to a structure named @var{tagname}.
9149 On other systems, this construct is not supported at all.
9152 @defmac DBX_CONTIN_LENGTH
9153 A symbol name in DBX-format debugging information is normally
9154 continued (split into two separate @code{.stabs} directives) when it
9155 exceeds a certain length (by default, 80 characters). On some
9156 operating systems, DBX requires this splitting; on others, splitting
9157 must not be done. You can inhibit splitting by defining this macro
9158 with the value zero. You can override the default splitting-length by
9159 defining this macro as an expression for the length you desire.
9162 @defmac DBX_CONTIN_CHAR
9163 Normally continuation is indicated by adding a @samp{\} character to
9164 the end of a @code{.stabs} string when a continuation follows. To use
9165 a different character instead, define this macro as a character
9166 constant for the character you want to use. Do not define this macro
9167 if backslash is correct for your system.
9170 @defmac DBX_STATIC_STAB_DATA_SECTION
9171 Define this macro if it is necessary to go to the data section before
9172 outputting the @samp{.stabs} pseudo-op for a non-global static
9176 @defmac DBX_TYPE_DECL_STABS_CODE
9177 The value to use in the ``code'' field of the @code{.stabs} directive
9178 for a typedef. The default is @code{N_LSYM}.
9181 @defmac DBX_STATIC_CONST_VAR_CODE
9182 The value to use in the ``code'' field of the @code{.stabs} directive
9183 for a static variable located in the text section. DBX format does not
9184 provide any ``right'' way to do this. The default is @code{N_FUN}.
9187 @defmac DBX_REGPARM_STABS_CODE
9188 The value to use in the ``code'' field of the @code{.stabs} directive
9189 for a parameter passed in registers. DBX format does not provide any
9190 ``right'' way to do this. The default is @code{N_RSYM}.
9193 @defmac DBX_REGPARM_STABS_LETTER
9194 The letter to use in DBX symbol data to identify a symbol as a parameter
9195 passed in registers. DBX format does not customarily provide any way to
9196 do this. The default is @code{'P'}.
9199 @defmac DBX_FUNCTION_FIRST
9200 Define this macro if the DBX information for a function and its
9201 arguments should precede the assembler code for the function. Normally,
9202 in DBX format, the debugging information entirely follows the assembler
9206 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9207 Define this macro, with value 1, if the value of a symbol describing
9208 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9209 relative to the start of the enclosing function. Normally, GCC uses
9210 an absolute address.
9213 @defmac DBX_LINES_FUNCTION_RELATIVE
9214 Define this macro, with value 1, if the value of a symbol indicating
9215 the current line number (@code{N_SLINE}) should be relative to the
9216 start of the enclosing function. Normally, GCC uses an absolute address.
9219 @defmac DBX_USE_BINCL
9220 Define this macro if GCC should generate @code{N_BINCL} and
9221 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9222 macro also directs GCC to output a type number as a pair of a file
9223 number and a type number within the file. Normally, GCC does not
9224 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9225 number for a type number.
9229 @subsection Open-Ended Hooks for DBX Format
9231 @c prevent bad page break with this line
9232 These are hooks for DBX format.
9234 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9235 Define this macro to say how to output to @var{stream} the debugging
9236 information for the start of a scope level for variable names. The
9237 argument @var{name} is the name of an assembler symbol (for use with
9238 @code{assemble_name}) whose value is the address where the scope begins.
9241 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9242 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9245 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9246 Define this macro if the target machine requires special handling to
9247 output an @code{N_FUN} entry for the function @var{decl}.
9250 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9251 A C statement to output DBX debugging information before code for line
9252 number @var{line} of the current source file to the stdio stream
9253 @var{stream}. @var{counter} is the number of time the macro was
9254 invoked, including the current invocation; it is intended to generate
9255 unique labels in the assembly output.
9257 This macro should not be defined if the default output is correct, or
9258 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9261 @defmac NO_DBX_FUNCTION_END
9262 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9263 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9264 On those machines, define this macro to turn this feature off without
9265 disturbing the rest of the gdb extensions.
9268 @defmac NO_DBX_BNSYM_ENSYM
9269 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9270 extension construct. On those machines, define this macro to turn this
9271 feature off without disturbing the rest of the gdb extensions.
9274 @node File Names and DBX
9275 @subsection File Names in DBX Format
9277 @c prevent bad page break with this line
9278 This describes file names in DBX format.
9280 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9281 A C statement to output DBX debugging information to the stdio stream
9282 @var{stream}, which indicates that file @var{name} is the main source
9283 file---the file specified as the input file for compilation.
9284 This macro is called only once, at the beginning of compilation.
9286 This macro need not be defined if the standard form of output
9287 for DBX debugging information is appropriate.
9289 It may be necessary to refer to a label equal to the beginning of the
9290 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9291 to do so. If you do this, you must also set the variable
9292 @var{used_ltext_label_name} to @code{true}.
9295 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9296 Define this macro, with value 1, if GCC should not emit an indication
9297 of the current directory for compilation and current source language at
9298 the beginning of the file.
9301 @defmac NO_DBX_GCC_MARKER
9302 Define this macro, with value 1, if GCC should not emit an indication
9303 that this object file was compiled by GCC@. The default is to emit
9304 an @code{N_OPT} stab at the beginning of every source file, with
9305 @samp{gcc2_compiled.} for the string and value 0.
9308 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9309 A C statement to output DBX debugging information at the end of
9310 compilation of the main source file @var{name}. Output should be
9311 written to the stdio stream @var{stream}.
9313 If you don't define this macro, nothing special is output at the end
9314 of compilation, which is correct for most machines.
9317 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9318 Define this macro @emph{instead of} defining
9319 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9320 the end of compilation is an @code{N_SO} stab with an empty string,
9321 whose value is the highest absolute text address in the file.
9326 @subsection Macros for SDB and DWARF Output
9328 @c prevent bad page break with this line
9329 Here are macros for SDB and DWARF output.
9331 @defmac SDB_DEBUGGING_INFO
9332 Define this macro if GCC should produce COFF-style debugging output
9333 for SDB in response to the @option{-g} option.
9336 @defmac DWARF2_DEBUGGING_INFO
9337 Define this macro if GCC should produce dwarf version 2 format
9338 debugging output in response to the @option{-g} option.
9340 @hook TARGET_DWARF_CALLING_CONVENTION
9341 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9342 be emitted for each function. Instead of an integer return the enum
9343 value for the @code{DW_CC_} tag.
9346 To support optional call frame debugging information, you must also
9347 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9348 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9349 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9350 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9353 @defmac DWARF2_FRAME_INFO
9354 Define this macro to a nonzero value if GCC should always output
9355 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9356 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9357 exceptions are enabled, GCC will output this information not matter
9358 how you define @code{DWARF2_FRAME_INFO}.
9361 @hook TARGET_DEBUG_UNWIND_INFO
9362 This hook defines the mechanism that will be used for describing frame
9363 unwind information to the debugger. Normally the hook will return
9364 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9365 return @code{UI_NONE} otherwise.
9367 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9368 is disabled in order to always output DWARF 2 frame information.
9370 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9371 This will suppress generation of the normal debug frame unwind information.
9374 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9375 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9376 line debug info sections. This will result in much more compact line number
9377 tables, and hence is desirable if it works.
9380 @hook TARGET_WANT_DEBUG_PUB_SECTIONS
9382 @hook TARGET_DELAY_SCHED2
9384 @hook TARGET_DELAY_VARTRACK
9386 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9387 A C statement to issue assembly directives that create a difference
9388 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9391 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9392 A C statement to issue assembly directives that create a difference
9393 between the two given labels in system defined units, e.g. instruction
9394 slots on IA64 VMS, using an integer of the given size.
9397 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9398 A C statement to issue assembly directives that create a
9399 section-relative reference to the given @var{label}, using an integer of the
9400 given @var{size}. The label is known to be defined in the given @var{section}.
9403 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9404 A C statement to issue assembly directives that create a self-relative
9405 reference to the given @var{label}, using an integer of the given @var{size}.
9408 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9409 A C statement to issue assembly directives that create a reference to
9410 the DWARF table identifier @var{label} from the current section. This
9411 is used on some systems to avoid garbage collecting a DWARF table which
9412 is referenced by a function.
9415 @hook TARGET_ASM_OUTPUT_DWARF_DTPREL
9416 If defined, this target hook is a function which outputs a DTP-relative
9417 reference to the given TLS symbol of the specified size.
9420 @defmac PUT_SDB_@dots{}
9421 Define these macros to override the assembler syntax for the special
9422 SDB assembler directives. See @file{sdbout.c} for a list of these
9423 macros and their arguments. If the standard syntax is used, you need
9424 not define them yourself.
9428 Some assemblers do not support a semicolon as a delimiter, even between
9429 SDB assembler directives. In that case, define this macro to be the
9430 delimiter to use (usually @samp{\n}). It is not necessary to define
9431 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9435 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9436 Define this macro to allow references to unknown structure,
9437 union, or enumeration tags to be emitted. Standard COFF does not
9438 allow handling of unknown references, MIPS ECOFF has support for
9442 @defmac SDB_ALLOW_FORWARD_REFERENCES
9443 Define this macro to allow references to structure, union, or
9444 enumeration tags that have not yet been seen to be handled. Some
9445 assemblers choke if forward tags are used, while some require it.
9448 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9449 A C statement to output SDB debugging information before code for line
9450 number @var{line} of the current source file to the stdio stream
9451 @var{stream}. The default is to emit an @code{.ln} directive.
9456 @subsection Macros for VMS Debug Format
9458 @c prevent bad page break with this line
9459 Here are macros for VMS debug format.
9461 @defmac VMS_DEBUGGING_INFO
9462 Define this macro if GCC should produce debugging output for VMS
9463 in response to the @option{-g} option. The default behavior for VMS
9464 is to generate minimal debug info for a traceback in the absence of
9465 @option{-g} unless explicitly overridden with @option{-g0}. This
9466 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9467 @code{TARGET_OPTION_OVERRIDE}.
9470 @node Floating Point
9471 @section Cross Compilation and Floating Point
9472 @cindex cross compilation and floating point
9473 @cindex floating point and cross compilation
9475 While all modern machines use twos-complement representation for integers,
9476 there are a variety of representations for floating point numbers. This
9477 means that in a cross-compiler the representation of floating point numbers
9478 in the compiled program may be different from that used in the machine
9479 doing the compilation.
9481 Because different representation systems may offer different amounts of
9482 range and precision, all floating point constants must be represented in
9483 the target machine's format. Therefore, the cross compiler cannot
9484 safely use the host machine's floating point arithmetic; it must emulate
9485 the target's arithmetic. To ensure consistency, GCC always uses
9486 emulation to work with floating point values, even when the host and
9487 target floating point formats are identical.
9489 The following macros are provided by @file{real.h} for the compiler to
9490 use. All parts of the compiler which generate or optimize
9491 floating-point calculations must use these macros. They may evaluate
9492 their operands more than once, so operands must not have side effects.
9494 @defmac REAL_VALUE_TYPE
9495 The C data type to be used to hold a floating point value in the target
9496 machine's format. Typically this is a @code{struct} containing an
9497 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9501 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9502 Compares for equality the two values, @var{x} and @var{y}. If the target
9503 floating point format supports negative zeroes and/or NaNs,
9504 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9505 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9508 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9509 Tests whether @var{x} is less than @var{y}.
9512 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9513 Truncates @var{x} to a signed integer, rounding toward zero.
9516 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9517 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9518 @var{x} is negative, returns zero.
9521 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9522 Converts @var{string} into a floating point number in the target machine's
9523 representation for mode @var{mode}. This routine can handle both
9524 decimal and hexadecimal floating point constants, using the syntax
9525 defined by the C language for both.
9528 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9529 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9532 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9533 Determines whether @var{x} represents infinity (positive or negative).
9536 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9537 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9540 @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})
9541 Calculates an arithmetic operation on the two floating point values
9542 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9545 The operation to be performed is specified by @var{code}. Only the
9546 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9547 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9549 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9550 target's floating point format cannot represent infinity, it will call
9551 @code{abort}. Callers should check for this situation first, using
9552 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9555 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9556 Returns the negative of the floating point value @var{x}.
9559 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9560 Returns the absolute value of @var{x}.
9563 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9564 Truncates the floating point value @var{x} to fit in @var{mode}. The
9565 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9566 appropriate bit pattern to be output as a floating constant whose
9567 precision accords with mode @var{mode}.
9570 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9571 Converts a floating point value @var{x} into a double-precision integer
9572 which is then stored into @var{low} and @var{high}. If the value is not
9573 integral, it is truncated.
9576 @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})
9577 Converts a double-precision integer found in @var{low} and @var{high},
9578 into a floating point value which is then stored into @var{x}. The
9579 value is truncated to fit in mode @var{mode}.
9582 @node Mode Switching
9583 @section Mode Switching Instructions
9584 @cindex mode switching
9585 The following macros control mode switching optimizations:
9587 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9588 Define this macro if the port needs extra instructions inserted for mode
9589 switching in an optimizing compilation.
9591 For an example, the SH4 can perform both single and double precision
9592 floating point operations, but to perform a single precision operation,
9593 the FPSCR PR bit has to be cleared, while for a double precision
9594 operation, this bit has to be set. Changing the PR bit requires a general
9595 purpose register as a scratch register, hence these FPSCR sets have to
9596 be inserted before reload, i.e.@: you can't put this into instruction emitting
9597 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9599 You can have multiple entities that are mode-switched, and select at run time
9600 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9601 return nonzero for any @var{entity} that needs mode-switching.
9602 If you define this macro, you also have to define
9603 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9604 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9605 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9609 @defmac NUM_MODES_FOR_MODE_SWITCHING
9610 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9611 initializer for an array of integers. Each initializer element
9612 N refers to an entity that needs mode switching, and specifies the number
9613 of different modes that might need to be set for this entity.
9614 The position of the initializer in the initializer---starting counting at
9615 zero---determines the integer that is used to refer to the mode-switched
9617 In macros that take mode arguments / yield a mode result, modes are
9618 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9619 switch is needed / supplied.
9622 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9623 @var{entity} is an integer specifying a mode-switched entity. If
9624 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9625 return an integer value not larger than the corresponding element in
9626 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9627 be switched into prior to the execution of @var{insn}.
9630 @defmac MODE_AFTER (@var{mode}, @var{insn})
9631 If this macro is defined, it is evaluated for every @var{insn} during
9632 mode switching. It determines the mode that an insn results in (if
9633 different from the incoming mode).
9636 @defmac MODE_ENTRY (@var{entity})
9637 If this macro is defined, it is evaluated for every @var{entity} that needs
9638 mode switching. It should evaluate to an integer, which is a mode that
9639 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9640 is defined then @code{MODE_EXIT} must be defined.
9643 @defmac MODE_EXIT (@var{entity})
9644 If this macro is defined, it is evaluated for every @var{entity} that needs
9645 mode switching. It should evaluate to an integer, which is a mode that
9646 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9647 is defined then @code{MODE_ENTRY} must be defined.
9650 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9651 This macro specifies the order in which modes for @var{entity} are processed.
9652 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9653 lowest. The value of the macro should be an integer designating a mode
9654 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9655 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9656 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9659 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9660 Generate one or more insns to set @var{entity} to @var{mode}.
9661 @var{hard_reg_live} is the set of hard registers live at the point where
9662 the insn(s) are to be inserted.
9665 @node Target Attributes
9666 @section Defining target-specific uses of @code{__attribute__}
9667 @cindex target attributes
9668 @cindex machine attributes
9669 @cindex attributes, target-specific
9671 Target-specific attributes may be defined for functions, data and types.
9672 These are described using the following target hooks; they also need to
9673 be documented in @file{extend.texi}.
9675 @hook TARGET_ATTRIBUTE_TABLE
9676 If defined, this target hook points to an array of @samp{struct
9677 attribute_spec} (defined in @file{tree.h}) specifying the machine
9678 specific attributes for this target and some of the restrictions on the
9679 entities to which these attributes are applied and the arguments they
9683 @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
9684 If defined, this target hook is a function which returns true if the
9685 machine-specific attribute named @var{name} expects an identifier
9686 given as its first argument to be passed on as a plain identifier, not
9687 subjected to name lookup. If this is not defined, the default is
9688 false for all machine-specific attributes.
9691 @hook TARGET_COMP_TYPE_ATTRIBUTES
9692 If defined, this target hook is a function which returns zero if the attributes on
9693 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9694 and two if they are nearly compatible (which causes a warning to be
9695 generated). If this is not defined, machine-specific attributes are
9696 supposed always to be compatible.
9699 @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
9700 If defined, this target hook is a function which assigns default attributes to
9701 the newly defined @var{type}.
9704 @hook TARGET_MERGE_TYPE_ATTRIBUTES
9705 Define this target hook if the merging of type attributes needs special
9706 handling. If defined, the result is a list of the combined
9707 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9708 that @code{comptypes} has already been called and returned 1. This
9709 function may call @code{merge_attributes} to handle machine-independent
9713 @hook TARGET_MERGE_DECL_ATTRIBUTES
9714 Define this target hook if the merging of decl attributes needs special
9715 handling. If defined, the result is a list of the combined
9716 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9717 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9718 when this is needed are when one attribute overrides another, or when an
9719 attribute is nullified by a subsequent definition. This function may
9720 call @code{merge_attributes} to handle machine-independent merging.
9722 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9723 If the only target-specific handling you require is @samp{dllimport}
9724 for Microsoft Windows targets, you should define the macro
9725 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9726 will then define a function called
9727 @code{merge_dllimport_decl_attributes} which can then be defined as
9728 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9729 add @code{handle_dll_attribute} in the attribute table for your port
9730 to perform initial processing of the @samp{dllimport} and
9731 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9732 @file{i386/i386.c}, for example.
9735 @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
9737 @defmac TARGET_DECLSPEC
9738 Define this macro to a nonzero value if you want to treat
9739 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9740 default, this behavior is enabled only for targets that define
9741 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9742 of @code{__declspec} is via a built-in macro, but you should not rely
9743 on this implementation detail.
9746 @hook TARGET_INSERT_ATTRIBUTES
9747 Define this target hook if you want to be able to add attributes to a decl
9748 when it is being created. This is normally useful for back ends which
9749 wish to implement a pragma by using the attributes which correspond to
9750 the pragma's effect. The @var{node} argument is the decl which is being
9751 created. The @var{attr_ptr} argument is a pointer to the attribute list
9752 for this decl. The list itself should not be modified, since it may be
9753 shared with other decls, but attributes may be chained on the head of
9754 the list and @code{*@var{attr_ptr}} modified to point to the new
9755 attributes, or a copy of the list may be made if further changes are
9759 @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
9761 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9762 into the current function, despite its having target-specific
9763 attributes, @code{false} otherwise. By default, if a function has a
9764 target specific attribute attached to it, it will not be inlined.
9767 @hook TARGET_OPTION_VALID_ATTRIBUTE_P
9768 This hook is called to parse the @code{attribute(option("..."))}, and
9769 it allows the function to set different target machine compile time
9770 options for the current function that might be different than the
9771 options specified on the command line. The hook should return
9772 @code{true} if the options are valid.
9774 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9775 the function declaration to hold a pointer to a target specific
9776 @var{struct cl_target_option} structure.
9779 @hook TARGET_OPTION_SAVE
9780 This hook is called to save any additional target specific information
9781 in the @var{struct cl_target_option} structure for function specific
9783 @xref{Option file format}.
9786 @hook TARGET_OPTION_RESTORE
9787 This hook is called to restore any additional target specific
9788 information in the @var{struct cl_target_option} structure for
9789 function specific options.
9792 @hook TARGET_OPTION_PRINT
9793 This hook is called to print any additional target specific
9794 information in the @var{struct cl_target_option} structure for
9795 function specific options.
9798 @hook TARGET_OPTION_PRAGMA_PARSE
9799 This target hook parses the options for @code{#pragma GCC option} to
9800 set the machine specific options for functions that occur later in the
9801 input stream. The options should be the same as handled by the
9802 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9805 @hook TARGET_OPTION_OVERRIDE
9806 Sometimes certain combinations of command options do not make sense on
9807 a particular target machine. You can override the hook
9808 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9809 once just after all the command options have been parsed.
9811 Don't use this hook to turn on various extra optimizations for
9812 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9814 If you need to do something whenever the optimization level is
9815 changed via the optimize attribute or pragma, see
9816 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9819 @hook TARGET_CAN_INLINE_P
9820 This target hook returns @code{false} if the @var{caller} function
9821 cannot inline @var{callee}, based on target specific information. By
9822 default, inlining is not allowed if the callee function has function
9823 specific target options and the caller does not use the same options.
9827 @section Emulating TLS
9828 @cindex Emulated TLS
9830 For targets whose psABI does not provide Thread Local Storage via
9831 specific relocations and instruction sequences, an emulation layer is
9832 used. A set of target hooks allows this emulation layer to be
9833 configured for the requirements of a particular target. For instance
9834 the psABI may in fact specify TLS support in terms of an emulation
9837 The emulation layer works by creating a control object for every TLS
9838 object. To access the TLS object, a lookup function is provided
9839 which, when given the address of the control object, will return the
9840 address of the current thread's instance of the TLS object.
9842 @hook TARGET_EMUTLS_GET_ADDRESS
9843 Contains the name of the helper function that uses a TLS control
9844 object to locate a TLS instance. The default causes libgcc's
9845 emulated TLS helper function to be used.
9848 @hook TARGET_EMUTLS_REGISTER_COMMON
9849 Contains the name of the helper function that should be used at
9850 program startup to register TLS objects that are implicitly
9851 initialized to zero. If this is @code{NULL}, all TLS objects will
9852 have explicit initializers. The default causes libgcc's emulated TLS
9853 registration function to be used.
9856 @hook TARGET_EMUTLS_VAR_SECTION
9857 Contains the name of the section in which TLS control variables should
9858 be placed. The default of @code{NULL} allows these to be placed in
9862 @hook TARGET_EMUTLS_TMPL_SECTION
9863 Contains the name of the section in which TLS initializers should be
9864 placed. The default of @code{NULL} allows these to be placed in any
9868 @hook TARGET_EMUTLS_VAR_PREFIX
9869 Contains the prefix to be prepended to TLS control variable names.
9870 The default of @code{NULL} uses a target-specific prefix.
9873 @hook TARGET_EMUTLS_TMPL_PREFIX
9874 Contains the prefix to be prepended to TLS initializer objects. The
9875 default of @code{NULL} uses a target-specific prefix.
9878 @hook TARGET_EMUTLS_VAR_FIELDS
9879 Specifies a function that generates the FIELD_DECLs for a TLS control
9880 object type. @var{type} is the RECORD_TYPE the fields are for and
9881 @var{name} should be filled with the structure tag, if the default of
9882 @code{__emutls_object} is unsuitable. The default creates a type suitable
9883 for libgcc's emulated TLS function.
9886 @hook TARGET_EMUTLS_VAR_INIT
9887 Specifies a function that generates the CONSTRUCTOR to initialize a
9888 TLS control object. @var{var} is the TLS control object, @var{decl}
9889 is the TLS object and @var{tmpl_addr} is the address of the
9890 initializer. The default initializes libgcc's emulated TLS control object.
9893 @hook TARGET_EMUTLS_VAR_ALIGN_FIXED
9894 Specifies whether the alignment of TLS control variable objects is
9895 fixed and should not be increased as some backends may do to optimize
9896 single objects. The default is false.
9899 @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9900 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9901 may be used to describe emulated TLS control objects.
9904 @node MIPS Coprocessors
9905 @section Defining coprocessor specifics for MIPS targets.
9906 @cindex MIPS coprocessor-definition macros
9908 The MIPS specification allows MIPS implementations to have as many as 4
9909 coprocessors, each with as many as 32 private registers. GCC supports
9910 accessing these registers and transferring values between the registers
9911 and memory using asm-ized variables. For example:
9914 register unsigned int cp0count asm ("c0r1");
9920 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9921 names may be added as described below, or the default names may be
9922 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9924 Coprocessor registers are assumed to be epilogue-used; sets to them will
9925 be preserved even if it does not appear that the register is used again
9926 later in the function.
9928 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9929 the FPU@. One accesses COP1 registers through standard mips
9930 floating-point support; they are not included in this mechanism.
9932 There is one macro used in defining the MIPS coprocessor interface which
9933 you may want to override in subtargets; it is described below.
9935 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9936 A comma-separated list (with leading comma) of pairs describing the
9937 alternate names of coprocessor registers. The format of each entry should be
9939 @{ @var{alternatename}, @var{register_number}@}
9945 @section Parameters for Precompiled Header Validity Checking
9946 @cindex parameters, precompiled headers
9948 @hook TARGET_GET_PCH_VALIDITY
9949 This hook returns a pointer to the data needed by
9950 @code{TARGET_PCH_VALID_P} and sets
9951 @samp{*@var{sz}} to the size of the data in bytes.
9954 @hook TARGET_PCH_VALID_P
9955 This hook checks whether the options used to create a PCH file are
9956 compatible with the current settings. It returns @code{NULL}
9957 if so and a suitable error message if not. Error messages will
9958 be presented to the user and must be localized using @samp{_(@var{msg})}.
9960 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9961 when the PCH file was created and @var{sz} is the size of that data in bytes.
9962 It's safe to assume that the data was created by the same version of the
9963 compiler, so no format checking is needed.
9965 The default definition of @code{default_pch_valid_p} should be
9966 suitable for most targets.
9969 @hook TARGET_CHECK_PCH_TARGET_FLAGS
9970 If this hook is nonnull, the default implementation of
9971 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9972 of @code{target_flags}. @var{pch_flags} specifies the value that
9973 @code{target_flags} had when the PCH file was created. The return
9974 value is the same as for @code{TARGET_PCH_VALID_P}.
9978 @section C++ ABI parameters
9979 @cindex parameters, c++ abi
9981 @hook TARGET_CXX_GUARD_TYPE
9982 Define this hook to override the integer type used for guard variables.
9983 These are used to implement one-time construction of static objects. The
9984 default is long_long_integer_type_node.
9987 @hook TARGET_CXX_GUARD_MASK_BIT
9988 This hook determines how guard variables are used. It should return
9989 @code{false} (the default) if the first byte should be used. A return value of
9990 @code{true} indicates that only the least significant bit should be used.
9993 @hook TARGET_CXX_GET_COOKIE_SIZE
9994 This hook returns the size of the cookie to use when allocating an array
9995 whose elements have the indicated @var{type}. Assumes that it is already
9996 known that a cookie is needed. The default is
9997 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
9998 IA64/Generic C++ ABI@.
10001 @hook TARGET_CXX_COOKIE_HAS_SIZE
10002 This hook should return @code{true} if the element size should be stored in
10003 array cookies. The default is to return @code{false}.
10006 @hook TARGET_CXX_IMPORT_EXPORT_CLASS
10007 If defined by a backend this hook allows the decision made to export
10008 class @var{type} to be overruled. Upon entry @var{import_export}
10009 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10010 to be imported and 0 otherwise. This function should return the
10011 modified value and perform any other actions necessary to support the
10012 backend's targeted operating system.
10015 @hook TARGET_CXX_CDTOR_RETURNS_THIS
10016 This hook should return @code{true} if constructors and destructors return
10017 the address of the object created/destroyed. The default is to return
10021 @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
10022 This hook returns true if the key method for a class (i.e., the method
10023 which, if defined in the current translation unit, causes the virtual
10024 table to be emitted) may be an inline function. Under the standard
10025 Itanium C++ ABI the key method may be an inline function so long as
10026 the function is not declared inline in the class definition. Under
10027 some variants of the ABI, an inline function can never be the key
10028 method. The default is to return @code{true}.
10031 @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
10033 @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
10034 This hook returns true (the default) if virtual tables and other
10035 similar implicit class data objects are always COMDAT if they have
10036 external linkage. If this hook returns false, then class data for
10037 classes whose virtual table will be emitted in only one translation
10038 unit will not be COMDAT.
10041 @hook TARGET_CXX_LIBRARY_RTTI_COMDAT
10042 This hook returns true (the default) if the RTTI information for
10043 the basic types which is defined in the C++ runtime should always
10044 be COMDAT, false if it should not be COMDAT.
10047 @hook TARGET_CXX_USE_AEABI_ATEXIT
10048 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10049 should be used to register static destructors when @option{-fuse-cxa-atexit}
10050 is in effect. The default is to return false to use @code{__cxa_atexit}.
10053 @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
10054 This hook returns true if the target @code{atexit} function can be used
10055 in the same manner as @code{__cxa_atexit} to register C++ static
10056 destructors. This requires that @code{atexit}-registered functions in
10057 shared libraries are run in the correct order when the libraries are
10058 unloaded. The default is to return false.
10061 @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
10063 @node Named Address Spaces
10064 @section Adding support for named address spaces
10065 @cindex named address spaces
10067 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10068 standards committee, @cite{Programming Languages - C - Extensions to
10069 support embedded processors}, specifies a syntax for embedded
10070 processors to specify alternate address spaces. You can configure a
10071 GCC port to support section 5.1 of the draft report to add support for
10072 address spaces other than the default address space. These address
10073 spaces are new keywords that are similar to the @code{volatile} and
10074 @code{const} type attributes.
10076 Pointers to named address spaces can have a different size than
10077 pointers to the generic address space.
10079 For example, the SPU port uses the @code{__ea} address space to refer
10080 to memory in the host processor, rather than memory local to the SPU
10081 processor. Access to memory in the @code{__ea} address space involves
10082 issuing DMA operations to move data between the host processor and the
10083 local processor memory address space. Pointers in the @code{__ea}
10084 address space are either 32 bits or 64 bits based on the
10085 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10088 Internally, address spaces are represented as a small integer in the
10089 range 0 to 15 with address space 0 being reserved for the generic
10092 To register a named address space qualifier keyword with the C front end,
10093 the target may call the @code{c_register_addr_space} routine. For example,
10094 the SPU port uses the following to declare @code{__ea} as the keyword for
10095 named address space #1:
10097 #define ADDR_SPACE_EA 1
10098 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10101 @hook TARGET_ADDR_SPACE_POINTER_MODE
10102 Define this to return the machine mode to use for pointers to
10103 @var{address_space} if the target supports named address spaces.
10104 The default version of this hook returns @code{ptr_mode} for the
10105 generic address space only.
10108 @hook TARGET_ADDR_SPACE_ADDRESS_MODE
10109 Define this to return the machine mode to use for addresses in
10110 @var{address_space} if the target supports named address spaces.
10111 The default version of this hook returns @code{Pmode} for the
10112 generic address space only.
10115 @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
10116 Define this to return nonzero if the port can handle pointers
10117 with machine mode @var{mode} to address space @var{as}. This target
10118 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10119 except that it includes explicit named address space support. The default
10120 version of this hook returns true for the modes returned by either the
10121 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10122 target hooks for the given address space.
10125 @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
10126 Define this to return true if @var{exp} is a valid address for mode
10127 @var{mode} in the named address space @var{as}. The @var{strict}
10128 parameter says whether strict addressing is in effect after reload has
10129 finished. This target hook is the same as the
10130 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10131 explicit named address space support.
10134 @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
10135 Define this to modify an invalid address @var{x} to be a valid address
10136 with mode @var{mode} in the named address space @var{as}. This target
10137 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10138 except that it includes explicit named address space support.
10141 @hook TARGET_ADDR_SPACE_SUBSET_P
10142 Define this to return whether the @var{subset} named address space is
10143 contained within the @var{superset} named address space. Pointers to
10144 a named address space that is a subset of another named address space
10145 will be converted automatically without a cast if used together in
10146 arithmetic operations. Pointers to a superset address space can be
10147 converted to pointers to a subset address space via explicit casts.
10150 @hook TARGET_ADDR_SPACE_CONVERT
10151 Define this to convert the pointer expression represented by the RTL
10152 @var{op} with type @var{from_type} that points to a named address
10153 space to a new pointer expression with type @var{to_type} that points
10154 to a different named address space. When this hook it called, it is
10155 guaranteed that one of the two address spaces is a subset of the other,
10156 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10160 @section Miscellaneous Parameters
10161 @cindex parameters, miscellaneous
10163 @c prevent bad page break with this line
10164 Here are several miscellaneous parameters.
10166 @defmac HAS_LONG_COND_BRANCH
10167 Define this boolean macro to indicate whether or not your architecture
10168 has conditional branches that can span all of memory. It is used in
10169 conjunction with an optimization that partitions hot and cold basic
10170 blocks into separate sections of the executable. If this macro is
10171 set to false, gcc will convert any conditional branches that attempt
10172 to cross between sections into unconditional branches or indirect jumps.
10175 @defmac HAS_LONG_UNCOND_BRANCH
10176 Define this boolean macro to indicate whether or not your architecture
10177 has unconditional branches that can span all of memory. It is used in
10178 conjunction with an optimization that partitions hot and cold basic
10179 blocks into separate sections of the executable. If this macro is
10180 set to false, gcc will convert any unconditional branches that attempt
10181 to cross between sections into indirect jumps.
10184 @defmac CASE_VECTOR_MODE
10185 An alias for a machine mode name. This is the machine mode that
10186 elements of a jump-table should have.
10189 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10190 Optional: return the preferred mode for an @code{addr_diff_vec}
10191 when the minimum and maximum offset are known. If you define this,
10192 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10193 To make this work, you also have to define @code{INSN_ALIGN} and
10194 make the alignment for @code{addr_diff_vec} explicit.
10195 The @var{body} argument is provided so that the offset_unsigned and scale
10196 flags can be updated.
10199 @defmac CASE_VECTOR_PC_RELATIVE
10200 Define this macro to be a C expression to indicate when jump-tables
10201 should contain relative addresses. You need not define this macro if
10202 jump-tables never contain relative addresses, or jump-tables should
10203 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10207 @hook TARGET_CASE_VALUES_THRESHOLD
10208 This function return the smallest number of different values for which it
10209 is best to use a jump-table instead of a tree of conditional branches.
10210 The default is four for machines with a @code{casesi} instruction and
10211 five otherwise. This is best for most machines.
10214 @defmac CASE_USE_BIT_TESTS
10215 Define this macro to be a C expression to indicate whether C switch
10216 statements may be implemented by a sequence of bit tests. This is
10217 advantageous on processors that can efficiently implement left shift
10218 of 1 by the number of bits held in a register, but inappropriate on
10219 targets that would require a loop. By default, this macro returns
10220 @code{true} if the target defines an @code{ashlsi3} pattern, and
10221 @code{false} otherwise.
10224 @defmac WORD_REGISTER_OPERATIONS
10225 Define this macro if operations between registers with integral mode
10226 smaller than a word are always performed on the entire register.
10227 Most RISC machines have this property and most CISC machines do not.
10230 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10231 Define this macro to be a C expression indicating when insns that read
10232 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10233 bits outside of @var{mem_mode} to be either the sign-extension or the
10234 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10235 of @var{mem_mode} for which the
10236 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10237 @code{UNKNOWN} for other modes.
10239 This macro is not called with @var{mem_mode} non-integral or with a width
10240 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10241 value in this case. Do not define this macro if it would always return
10242 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10243 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10245 You may return a non-@code{UNKNOWN} value even if for some hard registers
10246 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10247 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10248 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10249 integral mode larger than this but not larger than @code{word_mode}.
10251 You must return @code{UNKNOWN} if for some hard registers that allow this
10252 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10253 @code{word_mode}, but that they can change to another integral mode that
10254 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10257 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10258 Define this macro if loading short immediate values into registers sign
10262 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
10263 Define this macro if the same instructions that convert a floating
10264 point number to a signed fixed point number also convert validly to an
10268 @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
10269 When @option{-ffast-math} is in effect, GCC tries to optimize
10270 divisions by the same divisor, by turning them into multiplications by
10271 the reciprocal. This target hook specifies the minimum number of divisions
10272 that should be there for GCC to perform the optimization for a variable
10273 of mode @var{mode}. The default implementation returns 3 if the machine
10274 has an instruction for the division, and 2 if it does not.
10278 The maximum number of bytes that a single instruction can move quickly
10279 between memory and registers or between two memory locations.
10282 @defmac MAX_MOVE_MAX
10283 The maximum number of bytes that a single instruction can move quickly
10284 between memory and registers or between two memory locations. If this
10285 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10286 constant value that is the largest value that @code{MOVE_MAX} can have
10290 @defmac SHIFT_COUNT_TRUNCATED
10291 A C expression that is nonzero if on this machine the number of bits
10292 actually used for the count of a shift operation is equal to the number
10293 of bits needed to represent the size of the object being shifted. When
10294 this macro is nonzero, the compiler will assume that it is safe to omit
10295 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10296 truncates the count of a shift operation. On machines that have
10297 instructions that act on bit-fields at variable positions, which may
10298 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10299 also enables deletion of truncations of the values that serve as
10300 arguments to bit-field instructions.
10302 If both types of instructions truncate the count (for shifts) and
10303 position (for bit-field operations), or if no variable-position bit-field
10304 instructions exist, you should define this macro.
10306 However, on some machines, such as the 80386 and the 680x0, truncation
10307 only applies to shift operations and not the (real or pretended)
10308 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10309 such machines. Instead, add patterns to the @file{md} file that include
10310 the implied truncation of the shift instructions.
10312 You need not define this macro if it would always have the value of zero.
10315 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10316 @hook TARGET_SHIFT_TRUNCATION_MASK
10317 This function describes how the standard shift patterns for @var{mode}
10318 deal with shifts by negative amounts or by more than the width of the mode.
10319 @xref{shift patterns}.
10321 On many machines, the shift patterns will apply a mask @var{m} to the
10322 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10323 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10324 this is true for mode @var{mode}, the function should return @var{m},
10325 otherwise it should return 0. A return value of 0 indicates that no
10326 particular behavior is guaranteed.
10328 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10329 @emph{not} apply to general shift rtxes; it applies only to instructions
10330 that are generated by the named shift patterns.
10332 The default implementation of this function returns
10333 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10334 and 0 otherwise. This definition is always safe, but if
10335 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10336 nevertheless truncate the shift count, you may get better code
10340 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10341 A C expression which is nonzero if on this machine it is safe to
10342 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10343 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10344 operating on it as if it had only @var{outprec} bits.
10346 On many machines, this expression can be 1.
10348 @c rearranged this, removed the phrase "it is reported that". this was
10349 @c to fix an overfull hbox. --mew 10feb93
10350 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10351 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10352 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10353 such cases may improve things.
10356 @hook TARGET_MODE_REP_EXTENDED
10357 The representation of an integral mode can be such that the values
10358 are always extended to a wider integral mode. Return
10359 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10360 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10361 otherwise. (Currently, none of the targets use zero-extended
10362 representation this way so unlike @code{LOAD_EXTEND_OP},
10363 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10364 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10365 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10366 widest integral mode and currently we take advantage of this fact.)
10368 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10369 value even if the extension is not performed on certain hard registers
10370 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10371 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10373 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10374 describe two related properties. If you define
10375 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10376 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10379 In order to enforce the representation of @code{mode},
10380 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10384 @defmac STORE_FLAG_VALUE
10385 A C expression describing the value returned by a comparison operator
10386 with an integral mode and stored by a store-flag instruction
10387 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10388 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10389 comparison operators whose results have a @code{MODE_INT} mode.
10391 A value of 1 or @minus{}1 means that the instruction implementing the
10392 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10393 and 0 when the comparison is false. Otherwise, the value indicates
10394 which bits of the result are guaranteed to be 1 when the comparison is
10395 true. This value is interpreted in the mode of the comparison
10396 operation, which is given by the mode of the first operand in the
10397 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10398 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10401 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10402 generate code that depends only on the specified bits. It can also
10403 replace comparison operators with equivalent operations if they cause
10404 the required bits to be set, even if the remaining bits are undefined.
10405 For example, on a machine whose comparison operators return an
10406 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10407 @samp{0x80000000}, saying that just the sign bit is relevant, the
10411 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10415 can be converted to
10418 (ashift:SI @var{x} (const_int @var{n}))
10422 where @var{n} is the appropriate shift count to move the bit being
10423 tested into the sign bit.
10425 There is no way to describe a machine that always sets the low-order bit
10426 for a true value, but does not guarantee the value of any other bits,
10427 but we do not know of any machine that has such an instruction. If you
10428 are trying to port GCC to such a machine, include an instruction to
10429 perform a logical-and of the result with 1 in the pattern for the
10430 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10432 Often, a machine will have multiple instructions that obtain a value
10433 from a comparison (or the condition codes). Here are rules to guide the
10434 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10439 Use the shortest sequence that yields a valid definition for
10440 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10441 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10442 comparison operators to do so because there may be opportunities to
10443 combine the normalization with other operations.
10446 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10447 slightly preferred on machines with expensive jumps and 1 preferred on
10451 As a second choice, choose a value of @samp{0x80000001} if instructions
10452 exist that set both the sign and low-order bits but do not define the
10456 Otherwise, use a value of @samp{0x80000000}.
10459 Many machines can produce both the value chosen for
10460 @code{STORE_FLAG_VALUE} and its negation in the same number of
10461 instructions. On those machines, you should also define a pattern for
10462 those cases, e.g., one matching
10465 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10468 Some machines can also perform @code{and} or @code{plus} operations on
10469 condition code values with less instructions than the corresponding
10470 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10471 machines, define the appropriate patterns. Use the names @code{incscc}
10472 and @code{decscc}, respectively, for the patterns which perform
10473 @code{plus} or @code{minus} operations on condition code values. See
10474 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10475 find such instruction sequences on other machines.
10477 If this macro is not defined, the default value, 1, is used. You need
10478 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10479 instructions, or if the value generated by these instructions is 1.
10482 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10483 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10484 returned when comparison operators with floating-point results are true.
10485 Define this macro on machines that have comparison operations that return
10486 floating-point values. If there are no such operations, do not define
10490 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10491 A C expression that gives a rtx representing the nonzero true element
10492 for vector comparisons. The returned rtx should be valid for the inner
10493 mode of @var{mode} which is guaranteed to be a vector mode. Define
10494 this macro on machines that have vector comparison operations that
10495 return a vector result. If there are no such operations, do not define
10496 this macro. Typically, this macro is defined as @code{const1_rtx} or
10497 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10498 the compiler optimizing such vector comparison operations for the
10502 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10503 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10504 A C expression that indicates whether the architecture defines a value
10505 for @code{clz} or @code{ctz} with a zero operand.
10506 A result of @code{0} indicates the value is undefined.
10507 If the value is defined for only the RTL expression, the macro should
10508 evaluate to @code{1}; if the value applies also to the corresponding optab
10509 entry (which is normally the case if it expands directly into
10510 the corresponding RTL), then the macro should evaluate to @code{2}.
10511 In the cases where the value is defined, @var{value} should be set to
10514 If this macro is not defined, the value of @code{clz} or
10515 @code{ctz} at zero is assumed to be undefined.
10517 This macro must be defined if the target's expansion for @code{ffs}
10518 relies on a particular value to get correct results. Otherwise it
10519 is not necessary, though it may be used to optimize some corner cases, and
10520 to provide a default expansion for the @code{ffs} optab.
10522 Note that regardless of this macro the ``definedness'' of @code{clz}
10523 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10524 visible to the user. Thus one may be free to adjust the value at will
10525 to match the target expansion of these operations without fear of
10530 An alias for the machine mode for pointers. On most machines, define
10531 this to be the integer mode corresponding to the width of a hardware
10532 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10533 On some machines you must define this to be one of the partial integer
10534 modes, such as @code{PSImode}.
10536 The width of @code{Pmode} must be at least as large as the value of
10537 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10538 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10542 @defmac FUNCTION_MODE
10543 An alias for the machine mode used for memory references to functions
10544 being called, in @code{call} RTL expressions. On most CISC machines,
10545 where an instruction can begin at any byte address, this should be
10546 @code{QImode}. On most RISC machines, where all instructions have fixed
10547 size and alignment, this should be a mode with the same size and alignment
10548 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10551 @defmac STDC_0_IN_SYSTEM_HEADERS
10552 In normal operation, the preprocessor expands @code{__STDC__} to the
10553 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10554 hosts, like Solaris, the system compiler uses a different convention,
10555 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10556 strict conformance to the C Standard.
10558 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10559 convention when processing system header files, but when processing user
10560 files @code{__STDC__} will always expand to 1.
10563 @defmac NO_IMPLICIT_EXTERN_C
10564 Define this macro if the system header files support C++ as well as C@.
10565 This macro inhibits the usual method of using system header files in
10566 C++, which is to pretend that the file's contents are enclosed in
10567 @samp{extern "C" @{@dots{}@}}.
10572 @defmac REGISTER_TARGET_PRAGMAS ()
10573 Define this macro if you want to implement any target-specific pragmas.
10574 If defined, it is a C expression which makes a series of calls to
10575 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10576 for each pragma. The macro may also do any
10577 setup required for the pragmas.
10579 The primary reason to define this macro is to provide compatibility with
10580 other compilers for the same target. In general, we discourage
10581 definition of target-specific pragmas for GCC@.
10583 If the pragma can be implemented by attributes then you should consider
10584 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10586 Preprocessor macros that appear on pragma lines are not expanded. All
10587 @samp{#pragma} directives that do not match any registered pragma are
10588 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10591 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10592 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10594 Each call to @code{c_register_pragma} or
10595 @code{c_register_pragma_with_expansion} establishes one pragma. The
10596 @var{callback} routine will be called when the preprocessor encounters a
10600 #pragma [@var{space}] @var{name} @dots{}
10603 @var{space} is the case-sensitive namespace of the pragma, or
10604 @code{NULL} to put the pragma in the global namespace. The callback
10605 routine receives @var{pfile} as its first argument, which can be passed
10606 on to cpplib's functions if necessary. You can lex tokens after the
10607 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10608 callback will be silently ignored. The end of the line is indicated by
10609 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10610 arguments of pragmas registered with
10611 @code{c_register_pragma_with_expansion} but not on the arguments of
10612 pragmas registered with @code{c_register_pragma}.
10614 Note that the use of @code{pragma_lex} is specific to the C and C++
10615 compilers. It will not work in the Java or Fortran compilers, or any
10616 other language compilers for that matter. Thus if @code{pragma_lex} is going
10617 to be called from target-specific code, it must only be done so when
10618 building the C and C++ compilers. This can be done by defining the
10619 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10620 target entry in the @file{config.gcc} file. These variables should name
10621 the target-specific, language-specific object file which contains the
10622 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10623 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10624 how to build this object file.
10627 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10628 Define this macro if macros should be expanded in the
10629 arguments of @samp{#pragma pack}.
10632 @hook TARGET_HANDLE_PRAGMA_EXTERN_PREFIX
10634 @defmac TARGET_DEFAULT_PACK_STRUCT
10635 If your target requires a structure packing default other than 0 (meaning
10636 the machine default), define this macro to the necessary value (in bytes).
10637 This must be a value that would also be valid to use with
10638 @samp{#pragma pack()} (that is, a small power of two).
10641 @defmac DOLLARS_IN_IDENTIFIERS
10642 Define this macro to control use of the character @samp{$} in
10643 identifier names for the C family of languages. 0 means @samp{$} is
10644 not allowed by default; 1 means it is allowed. 1 is the default;
10645 there is no need to define this macro in that case.
10648 @defmac NO_DOLLAR_IN_LABEL
10649 Define this macro if the assembler does not accept the character
10650 @samp{$} in label names. By default constructors and destructors in
10651 G++ have @samp{$} in the identifiers. If this macro is defined,
10652 @samp{.} is used instead.
10655 @defmac NO_DOT_IN_LABEL
10656 Define this macro if the assembler does not accept the character
10657 @samp{.} in label names. By default constructors and destructors in G++
10658 have names that use @samp{.}. If this macro is defined, these names
10659 are rewritten to avoid @samp{.}.
10662 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10663 Define this macro as a C expression that is nonzero if it is safe for the
10664 delay slot scheduler to place instructions in the delay slot of @var{insn},
10665 even if they appear to use a resource set or clobbered in @var{insn}.
10666 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10667 every @code{call_insn} has this behavior. On machines where some @code{insn}
10668 or @code{jump_insn} is really a function call and hence has this behavior,
10669 you should define this macro.
10671 You need not define this macro if it would always return zero.
10674 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10675 Define this macro as a C expression that is nonzero if it is safe for the
10676 delay slot scheduler to place instructions in the delay slot of @var{insn},
10677 even if they appear to set or clobber a resource referenced in @var{insn}.
10678 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10679 some @code{insn} or @code{jump_insn} is really a function call and its operands
10680 are registers whose use is actually in the subroutine it calls, you should
10681 define this macro. Doing so allows the delay slot scheduler to move
10682 instructions which copy arguments into the argument registers into the delay
10683 slot of @var{insn}.
10685 You need not define this macro if it would always return zero.
10688 @defmac MULTIPLE_SYMBOL_SPACES
10689 Define this macro as a C expression that is nonzero if, in some cases,
10690 global symbols from one translation unit may not be bound to undefined
10691 symbols in another translation unit without user intervention. For
10692 instance, under Microsoft Windows symbols must be explicitly imported
10693 from shared libraries (DLLs).
10695 You need not define this macro if it would always evaluate to zero.
10698 @hook TARGET_MD_ASM_CLOBBERS
10699 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10700 any hard regs the port wishes to automatically clobber for an asm.
10701 It should return the result of the last @code{tree_cons} used to add a
10702 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10703 corresponding parameters to the asm and may be inspected to avoid
10704 clobbering a register that is an input or output of the asm. You can use
10705 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10706 for overlap with regards to asm-declared registers.
10709 @defmac MATH_LIBRARY
10710 Define this macro as a C string constant for the linker argument to link
10711 in the system math library, minus the initial @samp{"-l"}, or
10712 @samp{""} if the target does not have a
10713 separate math library.
10715 You need only define this macro if the default of @samp{"m"} is wrong.
10718 @defmac LIBRARY_PATH_ENV
10719 Define this macro as a C string constant for the environment variable that
10720 specifies where the linker should look for libraries.
10722 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10726 @defmac TARGET_POSIX_IO
10727 Define this macro if the target supports the following POSIX@ file
10728 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10729 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10730 to use file locking when exiting a program, which avoids race conditions
10731 if the program has forked. It will also create directories at run-time
10732 for cross-profiling.
10735 @defmac MAX_CONDITIONAL_EXECUTE
10737 A C expression for the maximum number of instructions to execute via
10738 conditional execution instructions instead of a branch. A value of
10739 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10740 1 if it does use cc0.
10743 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10744 Used if the target needs to perform machine-dependent modifications on the
10745 conditionals used for turning basic blocks into conditionally executed code.
10746 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10747 contains information about the currently processed blocks. @var{true_expr}
10748 and @var{false_expr} are the tests that are used for converting the
10749 then-block and the else-block, respectively. Set either @var{true_expr} or
10750 @var{false_expr} to a null pointer if the tests cannot be converted.
10753 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10754 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10755 if-statements into conditions combined by @code{and} and @code{or} operations.
10756 @var{bb} contains the basic block that contains the test that is currently
10757 being processed and about to be turned into a condition.
10760 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10761 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10762 be converted to conditional execution format. @var{ce_info} points to
10763 a data structure, @code{struct ce_if_block}, which contains information
10764 about the currently processed blocks.
10767 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10768 A C expression to perform any final machine dependent modifications in
10769 converting code to conditional execution. The involved basic blocks
10770 can be found in the @code{struct ce_if_block} structure that is pointed
10771 to by @var{ce_info}.
10774 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10775 A C expression to cancel any machine dependent modifications in
10776 converting code to conditional execution. The involved basic blocks
10777 can be found in the @code{struct ce_if_block} structure that is pointed
10778 to by @var{ce_info}.
10781 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10782 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10783 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10786 @defmac IFCVT_EXTRA_FIELDS
10787 If defined, it should expand to a set of field declarations that will be
10788 added to the @code{struct ce_if_block} structure. These should be initialized
10789 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10792 @hook TARGET_MACHINE_DEPENDENT_REORG
10793 If non-null, this hook performs a target-specific pass over the
10794 instruction stream. The compiler will run it at all optimization levels,
10795 just before the point at which it normally does delayed-branch scheduling.
10797 The exact purpose of the hook varies from target to target. Some use
10798 it to do transformations that are necessary for correctness, such as
10799 laying out in-function constant pools or avoiding hardware hazards.
10800 Others use it as an opportunity to do some machine-dependent optimizations.
10802 You need not implement the hook if it has nothing to do. The default
10803 definition is null.
10806 @hook TARGET_INIT_BUILTINS
10807 Define this hook if you have any machine-specific built-in functions
10808 that need to be defined. It should be a function that performs the
10811 Machine specific built-in functions can be useful to expand special machine
10812 instructions that would otherwise not normally be generated because
10813 they have no equivalent in the source language (for example, SIMD vector
10814 instructions or prefetch instructions).
10816 To create a built-in function, call the function
10817 @code{lang_hooks.builtin_function}
10818 which is defined by the language front end. You can use any type nodes set
10819 up by @code{build_common_tree_nodes};
10820 only language front ends that use those two functions will call
10821 @samp{TARGET_INIT_BUILTINS}.
10824 @hook TARGET_BUILTIN_DECL
10825 Define this hook if you have any machine-specific built-in functions
10826 that need to be defined. It should be a function that returns the
10827 builtin function declaration for the builtin function code @var{code}.
10828 If there is no such builtin and it cannot be initialized at this time
10829 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10830 If @var{code} is out of range the function should return
10831 @code{error_mark_node}.
10834 @hook TARGET_EXPAND_BUILTIN
10836 Expand a call to a machine specific built-in function that was set up by
10837 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10838 function call; the result should go to @var{target} if that is
10839 convenient, and have mode @var{mode} if that is convenient.
10840 @var{subtarget} may be used as the target for computing one of
10841 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10842 ignored. This function should return the result of the call to the
10846 @hook TARGET_RESOLVE_OVERLOADED_BUILTIN
10847 Select a replacement for a machine specific built-in function that
10848 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10849 @emph{before} regular type checking, and so allows the target to
10850 implement a crude form of function overloading. @var{fndecl} is the
10851 declaration of the built-in function. @var{arglist} is the list of
10852 arguments passed to the built-in function. The result is a
10853 complete expression that implements the operation, usually
10854 another @code{CALL_EXPR}.
10855 @var{arglist} really has type @samp{VEC(tree,gc)*}
10858 @hook TARGET_FOLD_BUILTIN
10859 Fold a call to a machine specific built-in function that was set up by
10860 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10861 built-in function. @var{n_args} is the number of arguments passed to
10862 the function; the arguments themselves are pointed to by @var{argp}.
10863 The result is another tree containing a simplified expression for the
10864 call's result. If @var{ignore} is true the value will be ignored.
10867 @hook TARGET_INVALID_WITHIN_DOLOOP
10869 Take an instruction in @var{insn} and return NULL if it is valid within a
10870 low-overhead loop, otherwise return a string explaining why doloop
10871 could not be applied.
10873 Many targets use special registers for low-overhead looping. For any
10874 instruction that clobbers these this function should return a string indicating
10875 the reason why the doloop could not be applied.
10876 By default, the RTL loop optimizer does not use a present doloop pattern for
10877 loops containing function calls or branch on table instructions.
10880 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10882 Take a branch insn in @var{branch1} and another in @var{branch2}.
10883 Return true if redirecting @var{branch1} to the destination of
10884 @var{branch2} is possible.
10886 On some targets, branches may have a limited range. Optimizing the
10887 filling of delay slots can result in branches being redirected, and this
10888 may in turn cause a branch offset to overflow.
10891 @hook TARGET_COMMUTATIVE_P
10892 This target hook returns @code{true} if @var{x} is considered to be commutative.
10893 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10894 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10895 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10898 @hook TARGET_ALLOCATE_INITIAL_VALUE
10900 When the initial value of a hard register has been copied in a pseudo
10901 register, it is often not necessary to actually allocate another register
10902 to this pseudo register, because the original hard register or a stack slot
10903 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10904 is called at the start of register allocation once for each hard register
10905 that had its initial value copied by using
10906 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10907 Possible values are @code{NULL_RTX}, if you don't want
10908 to do any special allocation, a @code{REG} rtx---that would typically be
10909 the hard register itself, if it is known not to be clobbered---or a
10911 If you are returning a @code{MEM}, this is only a hint for the allocator;
10912 it might decide to use another register anyways.
10913 You may use @code{current_function_leaf_function} in the hook, functions
10914 that use @code{REG_N_SETS}, to determine if the hard
10915 register in question will not be clobbered.
10916 The default value of this hook is @code{NULL}, which disables any special
10920 @hook TARGET_UNSPEC_MAY_TRAP_P
10921 This target hook returns nonzero if @var{x}, an @code{unspec} or
10922 @code{unspec_volatile} operation, might cause a trap. Targets can use
10923 this hook to enhance precision of analysis for @code{unspec} and
10924 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10925 to analyze inner elements of @var{x} in which case @var{flags} should be
10929 @hook TARGET_SET_CURRENT_FUNCTION
10930 The compiler invokes this hook whenever it changes its current function
10931 context (@code{cfun}). You can define this function if
10932 the back end needs to perform any initialization or reset actions on a
10933 per-function basis. For example, it may be used to implement function
10934 attributes that affect register usage or code generation patterns.
10935 The argument @var{decl} is the declaration for the new function context,
10936 and may be null to indicate that the compiler has left a function context
10937 and is returning to processing at the top level.
10938 The default hook function does nothing.
10940 GCC sets @code{cfun} to a dummy function context during initialization of
10941 some parts of the back end. The hook function is not invoked in this
10942 situation; you need not worry about the hook being invoked recursively,
10943 or when the back end is in a partially-initialized state.
10944 @code{cfun} might be @code{NULL} to indicate processing at top level,
10945 outside of any function scope.
10948 @defmac TARGET_OBJECT_SUFFIX
10949 Define this macro to be a C string representing the suffix for object
10950 files on your target machine. If you do not define this macro, GCC will
10951 use @samp{.o} as the suffix for object files.
10954 @defmac TARGET_EXECUTABLE_SUFFIX
10955 Define this macro to be a C string representing the suffix to be
10956 automatically added to executable files on your target machine. If you
10957 do not define this macro, GCC will use the null string as the suffix for
10961 @defmac COLLECT_EXPORT_LIST
10962 If defined, @code{collect2} will scan the individual object files
10963 specified on its command line and create an export list for the linker.
10964 Define this macro for systems like AIX, where the linker discards
10965 object files that are not referenced from @code{main} and uses export
10969 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
10970 Define this macro to a C expression representing a variant of the
10971 method call @var{mdecl}, if Java Native Interface (JNI) methods
10972 must be invoked differently from other methods on your target.
10973 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
10974 the @code{stdcall} calling convention and this macro is then
10975 defined as this expression:
10978 build_type_attribute_variant (@var{mdecl},
10980 (get_identifier ("stdcall"),
10985 @hook TARGET_CANNOT_MODIFY_JUMPS_P
10986 This target hook returns @code{true} past the point in which new jump
10987 instructions could be created. On machines that require a register for
10988 every jump such as the SHmedia ISA of SH5, this point would typically be
10989 reload, so this target hook should be defined to a function such as:
10993 cannot_modify_jumps_past_reload_p ()
10995 return (reload_completed || reload_in_progress);
11000 @hook TARGET_BRANCH_TARGET_REGISTER_CLASS
11001 This target hook returns a register class for which branch target register
11002 optimizations should be applied. All registers in this class should be
11003 usable interchangeably. After reload, registers in this class will be
11004 re-allocated and loads will be hoisted out of loops and be subjected
11005 to inter-block scheduling.
11008 @hook TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED
11009 Branch target register optimization will by default exclude callee-saved
11011 that are not already live during the current function; if this target hook
11012 returns true, they will be included. The target code must than make sure
11013 that all target registers in the class returned by
11014 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11015 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11016 epilogues have already been generated. Note, even if you only return
11017 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11018 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11019 to reserve space for caller-saved target registers.
11022 @hook TARGET_HAVE_CONDITIONAL_EXECUTION
11023 This target hook returns true if the target supports conditional execution.
11024 This target hook is required only when the target has several different
11025 modes and they have different conditional execution capability, such as ARM.
11028 @hook TARGET_LOOP_UNROLL_ADJUST
11029 This target hook returns a new value for the number of times @var{loop}
11030 should be unrolled. The parameter @var{nunroll} is the number of times
11031 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11032 the loop, which is going to be checked for unrolling. This target hook
11033 is required only when the target has special constraints like maximum
11034 number of memory accesses.
11037 @defmac POWI_MAX_MULTS
11038 If defined, this macro is interpreted as a signed integer C expression
11039 that specifies the maximum number of floating point multiplications
11040 that should be emitted when expanding exponentiation by an integer
11041 constant inline. When this value is defined, exponentiation requiring
11042 more than this number of multiplications is implemented by calling the
11043 system library's @code{pow}, @code{powf} or @code{powl} routines.
11044 The default value places no upper bound on the multiplication count.
11047 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11048 This target hook should register any extra include files for the
11049 target. The parameter @var{stdinc} indicates if normal include files
11050 are present. The parameter @var{sysroot} is the system root directory.
11051 The parameter @var{iprefix} is the prefix for the gcc directory.
11054 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11055 This target hook should register any extra include files for the
11056 target before any standard headers. The parameter @var{stdinc}
11057 indicates if normal include files are present. The parameter
11058 @var{sysroot} is the system root directory. The parameter
11059 @var{iprefix} is the prefix for the gcc directory.
11062 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11063 This target hook should register special include paths for the target.
11064 The parameter @var{path} is the include to register. On Darwin
11065 systems, this is used for Framework includes, which have semantics
11066 that are different from @option{-I}.
11069 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11070 This target macro returns @code{true} if it is safe to use a local alias
11071 for a virtual function @var{fndecl} when constructing thunks,
11072 @code{false} otherwise. By default, the macro returns @code{true} for all
11073 functions, if a target supports aliases (i.e.@: defines
11074 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11077 @defmac TARGET_FORMAT_TYPES
11078 If defined, this macro is the name of a global variable containing
11079 target-specific format checking information for the @option{-Wformat}
11080 option. The default is to have no target-specific format checks.
11083 @defmac TARGET_N_FORMAT_TYPES
11084 If defined, this macro is the number of entries in
11085 @code{TARGET_FORMAT_TYPES}.
11088 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11089 If defined, this macro is the name of a global variable containing
11090 target-specific format overrides for the @option{-Wformat} option. The
11091 default is to have no target-specific format overrides. If defined,
11092 @code{TARGET_FORMAT_TYPES} must be defined, too.
11095 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11096 If defined, this macro specifies the number of entries in
11097 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11100 @defmac TARGET_OVERRIDES_FORMAT_INIT
11101 If defined, this macro specifies the optional initialization
11102 routine for target specific customizations of the system printf
11103 and scanf formatter settings.
11106 @hook TARGET_RELAXED_ORDERING
11107 If set to @code{true}, means that the target's memory model does not
11108 guarantee that loads which do not depend on one another will access
11109 main memory in the order of the instruction stream; if ordering is
11110 important, an explicit memory barrier must be used. This is true of
11111 many recent processors which implement a policy of ``relaxed,''
11112 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11113 and ia64. The default is @code{false}.
11116 @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
11117 If defined, this macro returns the diagnostic message when it is
11118 illegal to pass argument @var{val} to function @var{funcdecl}
11119 with prototype @var{typelist}.
11122 @hook TARGET_INVALID_CONVERSION
11123 If defined, this macro returns the diagnostic message when it is
11124 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11125 if validity should be determined by the front end.
11128 @hook TARGET_INVALID_UNARY_OP
11129 If defined, this macro returns the diagnostic message when it is
11130 invalid to apply operation @var{op} (where unary plus is denoted by
11131 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11132 if validity should be determined by the front end.
11135 @hook TARGET_INVALID_BINARY_OP
11136 If defined, this macro returns the diagnostic message when it is
11137 invalid to apply operation @var{op} to operands of types @var{type1}
11138 and @var{type2}, or @code{NULL} if validity should be determined by
11142 @hook TARGET_INVALID_PARAMETER_TYPE
11143 If defined, this macro returns the diagnostic message when it is
11144 invalid for functions to include parameters of type @var{type},
11145 or @code{NULL} if validity should be determined by
11146 the front end. This is currently used only by the C and C++ front ends.
11149 @hook TARGET_INVALID_RETURN_TYPE
11150 If defined, this macro returns the diagnostic message when it is
11151 invalid for functions to have return type @var{type},
11152 or @code{NULL} if validity should be determined by
11153 the front end. This is currently used only by the C and C++ front ends.
11156 @hook TARGET_PROMOTED_TYPE
11157 If defined, this target hook returns the type to which values of
11158 @var{type} should be promoted when they appear in expressions,
11159 analogous to the integer promotions, or @code{NULL_TREE} to use the
11160 front end's normal promotion rules. This hook is useful when there are
11161 target-specific types with special promotion rules.
11162 This is currently used only by the C and C++ front ends.
11165 @hook TARGET_CONVERT_TO_TYPE
11166 If defined, this hook returns the result of converting @var{expr} to
11167 @var{type}. It should return the converted expression,
11168 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11169 This hook is useful when there are target-specific types with special
11171 This is currently used only by the C and C++ front ends.
11174 @defmac TARGET_USE_JCR_SECTION
11175 This macro determines whether to use the JCR section to register Java
11176 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11177 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11181 This macro determines the size of the objective C jump buffer for the
11182 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11185 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11186 Define this macro if any target-specific attributes need to be attached
11187 to the functions in @file{libgcc} that provide low-level support for
11188 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11189 and the associated definitions of those functions.
11192 @hook TARGET_UPDATE_STACK_BOUNDARY
11193 Define this macro to update the current function stack boundary if
11197 @hook TARGET_GET_DRAP_RTX
11198 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11199 different argument pointer register is needed to access the function's
11200 argument list due to stack realignment. Return @code{NULL} if no DRAP
11204 @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
11205 When optimization is disabled, this hook indicates whether or not
11206 arguments should be allocated to stack slots. Normally, GCC allocates
11207 stacks slots for arguments when not optimizing in order to make
11208 debugging easier. However, when a function is declared with
11209 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11210 cannot safely move arguments from the registers in which they are passed
11211 to the stack. Therefore, this hook should return true in general, but
11212 false for naked functions. The default implementation always returns true.
11215 @hook TARGET_CONST_ANCHOR
11216 On some architectures it can take multiple instructions to synthesize
11217 a constant. If there is another constant already in a register that
11218 is close enough in value then it is preferable that the new constant
11219 is computed from this register using immediate addition or
11220 subtraction. We accomplish this through CSE. Besides the value of
11221 the constant we also add a lower and an upper constant anchor to the
11222 available expressions. These are then queried when encountering new
11223 constants. The anchors are computed by rounding the constant up and
11224 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11225 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11226 accepted by immediate-add plus one. We currently assume that the
11227 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11228 MIPS, where add-immediate takes a 16-bit signed value,
11229 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11230 is zero, which disables this optimization. @end deftypevr