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
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.
103 @section Controlling the Compilation Driver, @file{gcc}
105 @cindex controlling the compilation driver
107 @c prevent bad page break with this line
108 You can control the compilation driver.
110 @defmac DRIVER_SELF_SPECS
111 A list of specs for the driver itself. It should be a suitable
112 initializer for an array of strings, with no surrounding braces.
114 The driver applies these specs to its own command line between loading
115 default @file{specs} files (but not command-line specified ones) and
116 choosing the multilib directory or running any subcommands. It
117 applies them in the order given, so each spec can depend on the
118 options added by earlier ones. It is also possible to remove options
119 using @samp{%<@var{option}} in the usual way.
121 This macro can be useful when a port has several interdependent target
122 options. It provides a way of standardizing the command line so
123 that the other specs are easier to write.
125 Do not define this macro if it does not need to do anything.
128 @defmac OPTION_DEFAULT_SPECS
129 A list of specs used to support configure-time default options (i.e.@:
130 @option{--with} options) in the driver. It should be a suitable initializer
131 for an array of structures, each containing two strings, without the
132 outermost pair of surrounding braces.
134 The first item in the pair is the name of the default. This must match
135 the code in @file{config.gcc} for the target. The second item is a spec
136 to apply if a default with this name was specified. The string
137 @samp{%(VALUE)} in the spec will be replaced by the value of the default
138 everywhere it occurs.
140 The driver will apply these specs to its own command line between loading
141 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
142 the same mechanism as @code{DRIVER_SELF_SPECS}.
144 Do not define this macro if it does not need to do anything.
148 A C string constant that tells the GCC driver program options to
149 pass to CPP@. It can also specify how to translate options you
150 give to GCC into options for GCC to pass to the CPP@.
152 Do not define this macro if it does not need to do anything.
155 @defmac CPLUSPLUS_CPP_SPEC
156 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
157 than C@. If you do not define this macro, then the value of
158 @code{CPP_SPEC} (if any) will be used instead.
162 A C string constant that tells the GCC driver program options to
163 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
165 It can also specify how to translate options you give to GCC into options
166 for GCC to pass to front ends.
168 Do not define this macro if it does not need to do anything.
172 A C string constant that tells the GCC driver program options to
173 pass to @code{cc1plus}. It can also specify how to translate options you
174 give to GCC into options for GCC to pass to the @code{cc1plus}.
176 Do not define this macro if it does not need to do anything.
177 Note that everything defined in CC1_SPEC is already passed to
178 @code{cc1plus} so there is no need to duplicate the contents of
179 CC1_SPEC in CC1PLUS_SPEC@.
183 A C string constant that tells the GCC driver program options to
184 pass to the assembler. It can also specify how to translate options
185 you give to GCC into options for GCC to pass to the assembler.
186 See the file @file{sun3.h} for an example of this.
188 Do not define this macro if it does not need to do anything.
191 @defmac ASM_FINAL_SPEC
192 A C string constant that tells the GCC driver program how to
193 run any programs which cleanup after the normal assembler.
194 Normally, this is not needed. See the file @file{mips.h} for
197 Do not define this macro if it does not need to do anything.
200 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
201 Define this macro, with no value, if the driver should give the assembler
202 an argument consisting of a single dash, @option{-}, to instruct it to
203 read from its standard input (which will be a pipe connected to the
204 output of the compiler proper). This argument is given after any
205 @option{-o} option specifying the name of the output file.
207 If you do not define this macro, the assembler is assumed to read its
208 standard input if given no non-option arguments. If your assembler
209 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
210 see @file{mips.h} for instance.
214 A C string constant that tells the GCC driver program options to
215 pass to the linker. It can also specify how to translate options you
216 give to GCC into options for GCC to pass to the linker.
218 Do not define this macro if it does not need to do anything.
222 Another C string constant used much like @code{LINK_SPEC}. The difference
223 between the two is that @code{LIB_SPEC} is used at the end of the
224 command given to the linker.
226 If this macro is not defined, a default is provided that
227 loads the standard C library from the usual place. See @file{gcc.c}.
231 Another C string constant that tells the GCC driver program
232 how and when to place a reference to @file{libgcc.a} into the
233 linker command line. This constant is placed both before and after
234 the value of @code{LIB_SPEC}.
236 If this macro is not defined, the GCC driver provides a default that
237 passes the string @option{-lgcc} to the linker.
240 @defmac REAL_LIBGCC_SPEC
241 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
242 @code{LIBGCC_SPEC} is not directly used by the driver program but is
243 instead modified to refer to different versions of @file{libgcc.a}
244 depending on the values of the command line flags @option{-static},
245 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
246 targets where these modifications are inappropriate, define
247 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
248 driver how to place a reference to @file{libgcc} on the link command
249 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
252 @defmac USE_LD_AS_NEEDED
253 A macro that controls the modifications to @code{LIBGCC_SPEC}
254 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
255 generated that uses --as-needed and the shared libgcc in place of the
256 static exception handler library, when linking without any of
257 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
261 If defined, this C string constant is added to @code{LINK_SPEC}.
262 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
263 the modifications to @code{LIBGCC_SPEC} mentioned in
264 @code{REAL_LIBGCC_SPEC}.
267 @defmac STARTFILE_SPEC
268 Another C string constant used much like @code{LINK_SPEC}. The
269 difference between the two is that @code{STARTFILE_SPEC} is used at
270 the very beginning of the command given to the linker.
272 If this macro is not defined, a default is provided that loads the
273 standard C startup file from the usual place. See @file{gcc.c}.
277 Another C string constant used much like @code{LINK_SPEC}. The
278 difference between the two is that @code{ENDFILE_SPEC} is used at
279 the very end of the command given to the linker.
281 Do not define this macro if it does not need to do anything.
284 @defmac THREAD_MODEL_SPEC
285 GCC @code{-v} will print the thread model GCC was configured to use.
286 However, this doesn't work on platforms that are multilibbed on thread
287 models, such as AIX 4.3. On such platforms, define
288 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
289 blanks that names one of the recognized thread models. @code{%*}, the
290 default value of this macro, will expand to the value of
291 @code{thread_file} set in @file{config.gcc}.
294 @defmac SYSROOT_SUFFIX_SPEC
295 Define this macro to add a suffix to the target sysroot when GCC is
296 configured with a sysroot. This will cause GCC to search for usr/lib,
297 et al, within sysroot+suffix.
300 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
301 Define this macro to add a headers_suffix to the target sysroot when
302 GCC is configured with a sysroot. This will cause GCC to pass the
303 updated sysroot+headers_suffix to CPP, causing it to search for
304 usr/include, et al, within sysroot+headers_suffix.
308 Define this macro to provide additional specifications to put in the
309 @file{specs} file that can be used in various specifications like
312 The definition should be an initializer for an array of structures,
313 containing a string constant, that defines the specification name, and a
314 string constant that provides the specification.
316 Do not define this macro if it does not need to do anything.
318 @code{EXTRA_SPECS} is useful when an architecture contains several
319 related targets, which have various @code{@dots{}_SPECS} which are similar
320 to each other, and the maintainer would like one central place to keep
323 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
324 define either @code{_CALL_SYSV} when the System V calling sequence is
325 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
328 The @file{config/rs6000/rs6000.h} target file defines:
331 #define EXTRA_SPECS \
332 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
334 #define CPP_SYS_DEFAULT ""
337 The @file{config/rs6000/sysv.h} target file defines:
341 "%@{posix: -D_POSIX_SOURCE @} \
342 %@{mcall-sysv: -D_CALL_SYSV @} \
343 %@{!mcall-sysv: %(cpp_sysv_default) @} \
344 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
346 #undef CPP_SYSV_DEFAULT
347 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
350 while the @file{config/rs6000/eabiaix.h} target file defines
351 @code{CPP_SYSV_DEFAULT} as:
354 #undef CPP_SYSV_DEFAULT
355 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
359 @defmac LINK_LIBGCC_SPECIAL_1
360 Define this macro if the driver program should find the library
361 @file{libgcc.a}. If you do not define this macro, the driver program will pass
362 the argument @option{-lgcc} to tell the linker to do the search.
365 @defmac LINK_GCC_C_SEQUENCE_SPEC
366 The sequence in which libgcc and libc are specified to the linker.
367 By default this is @code{%G %L %G}.
370 @defmac LINK_COMMAND_SPEC
371 A C string constant giving the complete command line need to execute the
372 linker. When you do this, you will need to update your port each time a
373 change is made to the link command line within @file{gcc.c}. Therefore,
374 define this macro only if you need to completely redefine the command
375 line for invoking the linker and there is no other way to accomplish
376 the effect you need. Overriding this macro may be avoidable by overriding
377 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
380 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
381 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
382 directories from linking commands. Do not give it a nonzero value if
383 removing duplicate search directories changes the linker's semantics.
386 @defmac MULTILIB_DEFAULTS
387 Define this macro as a C expression for the initializer of an array of
388 string to tell the driver program which options are defaults for this
389 target and thus do not need to be handled specially when using
390 @code{MULTILIB_OPTIONS}.
392 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
393 the target makefile fragment or if none of the options listed in
394 @code{MULTILIB_OPTIONS} are set by default.
395 @xref{Target Fragment}.
398 @defmac RELATIVE_PREFIX_NOT_LINKDIR
399 Define this macro to tell @command{gcc} that it should only translate
400 a @option{-B} prefix into a @option{-L} linker option if the prefix
401 indicates an absolute file name.
404 @defmac MD_EXEC_PREFIX
405 If defined, this macro is an additional prefix to try after
406 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
407 when the compiler is built as a cross
408 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
409 to the list of directories used to find the assembler in @file{configure.in}.
412 @defmac STANDARD_STARTFILE_PREFIX
413 Define this macro as a C string constant if you wish to override the
414 standard choice of @code{libdir} as the default prefix to
415 try when searching for startup files such as @file{crt0.o}.
416 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
417 is built as a cross compiler.
420 @defmac STANDARD_STARTFILE_PREFIX_1
421 Define this macro as a C string constant if you wish to override the
422 standard choice of @code{/lib} as a prefix to try after the default prefix
423 when searching for startup files such as @file{crt0.o}.
424 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
425 is built as a cross compiler.
428 @defmac STANDARD_STARTFILE_PREFIX_2
429 Define this macro as a C string constant if you wish to override the
430 standard choice of @code{/lib} as yet another prefix to try after the
431 default prefix when searching for startup files such as @file{crt0.o}.
432 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
433 is built as a cross compiler.
436 @defmac MD_STARTFILE_PREFIX
437 If defined, this macro supplies an additional prefix to try after the
438 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
439 compiler is built as a cross compiler.
442 @defmac MD_STARTFILE_PREFIX_1
443 If defined, this macro supplies yet another prefix to try after the
444 standard prefixes. It is not searched when the compiler is built as a
448 @defmac INIT_ENVIRONMENT
449 Define this macro as a C string constant if you wish to set environment
450 variables for programs called by the driver, such as the assembler and
451 loader. The driver passes the value of this macro to @code{putenv} to
452 initialize the necessary environment variables.
455 @defmac LOCAL_INCLUDE_DIR
456 Define this macro as a C string constant if you wish to override the
457 standard choice of @file{/usr/local/include} as the default prefix to
458 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
459 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
461 Cross compilers do not search either @file{/usr/local/include} or its
465 @defmac SYSTEM_INCLUDE_DIR
466 Define this macro as a C string constant if you wish to specify a
467 system-specific directory to search for header files before the standard
468 directory. @code{SYSTEM_INCLUDE_DIR} comes before
469 @code{STANDARD_INCLUDE_DIR} in the search order.
471 Cross compilers do not use this macro and do not search the directory
475 @defmac STANDARD_INCLUDE_DIR
476 Define this macro as a C string constant if you wish to override the
477 standard choice of @file{/usr/include} as the default prefix to
478 try when searching for header files.
480 Cross compilers ignore this macro and do not search either
481 @file{/usr/include} or its replacement.
484 @defmac STANDARD_INCLUDE_COMPONENT
485 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
486 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
487 If you do not define this macro, no component is used.
490 @defmac INCLUDE_DEFAULTS
491 Define this macro if you wish to override the entire default search path
492 for include files. For a native compiler, the default search path
493 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
494 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
495 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
496 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
497 and specify private search areas for GCC@. The directory
498 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
500 The definition should be an initializer for an array of structures.
501 Each array element should have four elements: the directory name (a
502 string constant), the component name (also a string constant), a flag
503 for C++-only directories,
504 and a flag showing that the includes in the directory don't need to be
505 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
506 the array with a null element.
508 The component name denotes what GNU package the include file is part of,
509 if any, in all uppercase letters. For example, it might be @samp{GCC}
510 or @samp{BINUTILS}. If the package is part of a vendor-supplied
511 operating system, code the component name as @samp{0}.
513 For example, here is the definition used for VAX/VMS:
516 #define INCLUDE_DEFAULTS \
518 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
519 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
520 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
527 Here is the order of prefixes tried for exec files:
531 Any prefixes specified by the user with @option{-B}.
534 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
535 is not set and the compiler has not been installed in the configure-time
536 @var{prefix}, the location in which the compiler has actually been installed.
539 The directories specified by the environment variable @code{COMPILER_PATH}.
542 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
543 in the configured-time @var{prefix}.
546 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
549 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
552 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
556 Here is the order of prefixes tried for startfiles:
560 Any prefixes specified by the user with @option{-B}.
563 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
564 value based on the installed toolchain location.
567 The directories specified by the environment variable @code{LIBRARY_PATH}
568 (or port-specific name; native only, cross compilers do not use this).
571 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
572 in the configured @var{prefix} or this is a native compiler.
575 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
578 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
582 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
583 native compiler, or we have a target system root.
586 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
587 native compiler, or we have a target system root.
590 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
591 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
592 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
595 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
596 compiler, or we have a target system root. The default for this macro is
600 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
601 compiler, or we have a target system root. The default for this macro is
605 @node Run-time Target
606 @section Run-time Target Specification
607 @cindex run-time target specification
608 @cindex predefined macros
609 @cindex target specifications
611 @c prevent bad page break with this line
612 Here are run-time target specifications.
614 @defmac TARGET_CPU_CPP_BUILTINS ()
615 This function-like macro expands to a block of code that defines
616 built-in preprocessor macros and assertions for the target CPU, using
617 the functions @code{builtin_define}, @code{builtin_define_std} and
618 @code{builtin_assert}. When the front end
619 calls this macro it provides a trailing semicolon, and since it has
620 finished command line option processing your code can use those
623 @code{builtin_assert} takes a string in the form you pass to the
624 command-line option @option{-A}, such as @code{cpu=mips}, and creates
625 the assertion. @code{builtin_define} takes a string in the form
626 accepted by option @option{-D} and unconditionally defines the macro.
628 @code{builtin_define_std} takes a string representing the name of an
629 object-like macro. If it doesn't lie in the user's namespace,
630 @code{builtin_define_std} defines it unconditionally. Otherwise, it
631 defines a version with two leading underscores, and another version
632 with two leading and trailing underscores, and defines the original
633 only if an ISO standard was not requested on the command line. For
634 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
635 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
636 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
637 defines only @code{_ABI64}.
639 You can also test for the C dialect being compiled. The variable
640 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
641 or @code{clk_objective_c}. Note that if we are preprocessing
642 assembler, this variable will be @code{clk_c} but the function-like
643 macro @code{preprocessing_asm_p()} will return true, so you might want
644 to check for that first. If you need to check for strict ANSI, the
645 variable @code{flag_iso} can be used. The function-like macro
646 @code{preprocessing_trad_p()} can be used to check for traditional
650 @defmac TARGET_OS_CPP_BUILTINS ()
651 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
652 and is used for the target operating system instead.
655 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
656 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
657 and is used for the target object format. @file{elfos.h} uses this
658 macro to define @code{__ELF__}, so you probably do not need to define
662 @deftypevar {extern int} target_flags
663 This variable is declared in @file{options.h}, which is included before
664 any target-specific headers.
667 @deftypevr {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
668 This variable specifies the initial value of @code{target_flags}.
669 Its default setting is 0.
672 @cindex optional hardware or system features
673 @cindex features, optional, in system conventions
675 @deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (struct gcc_options *@var{opts}, struct gcc_options *@var{opts_set}, const struct cl_decoded_option *@var{decoded}, unsigned int @var{loc})
676 This hook is called whenever the user specifies one of the
677 target-specific options described by the @file{.opt} definition files
678 (@pxref{Options}). It has the opportunity to do some option-specific
679 processing and should return true if the option is valid. The default
680 definition does nothing but return true.
682 @var{decoded} specifies the option and its arguments. @var{opts} and
683 @var{opts_set} are the @code{gcc_options} structures to be used for
684 storing option state, and @var{loc} is the location at which the
685 option was passed (@code{UNKNOWN_LOCATION} except for options passed
689 @deftypefn {C Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
690 This target hook is called whenever the user specifies one of the
691 target-specific C language family options described by the @file{.opt}
692 definition files(@pxref{Options}). It has the opportunity to do some
693 option-specific processing and should return true if the option is
694 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
695 default definition does nothing but return false.
697 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
698 options. However, if processing an option requires routines that are
699 only available in the C (and related language) front ends, then you
700 should use @code{TARGET_HANDLE_C_OPTION} instead.
703 @deftypefn {C Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING_OBJECT (tree @var{string})
704 Targets may provide a string object type that can be used within and between C, C++ and their respective Objective-C dialects. A string object might, for example, embed encoding and length information. These objects are considered opaque to the compiler and handled as references. An ideal implementation makes the composition of the string object match that of the Objective-C @code{NSString} (@code{NXString} for GNUStep), allowing efficient interworking between C-only and Objective-C code. If a target implements string objects then this hook should return a reference to such an object constructed from the normal `C' string representation provided in @var{string}. At present, the hook is used by Objective-C only, to obtain a common-format string object when the target provides one.
707 @deftypefn {C Target Hook} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref})
708 If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object.
711 @deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list})
712 If a target implements string objects then this hook should should provide a facility to check the function arguments in @var{args_list} against the format specifiers in @var{format_arg} where the type of @var{format_arg} is one recognized as a valid string reference type.
715 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
716 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
717 but is called when the optimize level is changed via an attribute or
718 pragma or when it is reset at the end of the code affected by the
719 attribute or pragma. It is not called at the beginning of compilation
720 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
721 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
722 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
725 @defmac C_COMMON_OVERRIDE_OPTIONS
726 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
727 but is only used in the C
728 language frontends (C, Objective-C, C++, Objective-C++) and so can be
729 used to alter option flag variables which only exist in those
733 @deftypevr {Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
734 Some machines may desire to change what optimizations are performed for
735 various optimization levels. This variable, if defined, describes
736 options to enable at particular sets of optimization levels. These
737 options are processed once
738 just after the optimization level is determined and before the remainder
739 of the command options have been parsed, so may be overridden by other
740 options passed explicitly.
742 This processing is run once at program startup and when the optimization
743 options are changed via @code{#pragma GCC optimize} or by using the
744 @code{optimize} attribute.
747 @deftypefn {Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
748 Set target-dependent initial values of fields in @var{opts}.
751 @deftypefn {Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
752 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
755 @deftypefn {Target Hook} void TARGET_HELP (void)
756 This hook is called in response to the user invoking
757 @option{--target-help} on the command line. It gives the target a
758 chance to display extra information on the target specific command
759 line options found in its @file{.opt} file.
762 @defmac SWITCHABLE_TARGET
763 Some targets need to switch between substantially different subtargets
764 during compilation. For example, the MIPS target has one subtarget for
765 the traditional MIPS architecture and another for MIPS16. Source code
766 can switch between these two subarchitectures using the @code{mips16}
767 and @code{nomips16} attributes.
769 Such subtargets can differ in things like the set of available
770 registers, the set of available instructions, the costs of various
771 operations, and so on. GCC caches a lot of this type of information
772 in global variables, and recomputing them for each subtarget takes a
773 significant amount of time. The compiler therefore provides a facility
774 for maintaining several versions of the global variables and quickly
775 switching between them; see @file{target-globals.h} for details.
777 Define this macro to 1 if your target needs this facility. The default
781 @node Per-Function Data
782 @section Defining data structures for per-function information.
783 @cindex per-function data
784 @cindex data structures
786 If the target needs to store information on a per-function basis, GCC
787 provides a macro and a couple of variables to allow this. Note, just
788 using statics to store the information is a bad idea, since GCC supports
789 nested functions, so you can be halfway through encoding one function
790 when another one comes along.
792 GCC defines a data structure called @code{struct function} which
793 contains all of the data specific to an individual function. This
794 structure contains a field called @code{machine} whose type is
795 @code{struct machine_function *}, which can be used by targets to point
796 to their own specific data.
798 If a target needs per-function specific data it should define the type
799 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
800 This macro should be used to initialize the function pointer
801 @code{init_machine_status}. This pointer is explained below.
803 One typical use of per-function, target specific data is to create an
804 RTX to hold the register containing the function's return address. This
805 RTX can then be used to implement the @code{__builtin_return_address}
806 function, for level 0.
808 Note---earlier implementations of GCC used a single data area to hold
809 all of the per-function information. Thus when processing of a nested
810 function began the old per-function data had to be pushed onto a
811 stack, and when the processing was finished, it had to be popped off the
812 stack. GCC used to provide function pointers called
813 @code{save_machine_status} and @code{restore_machine_status} to handle
814 the saving and restoring of the target specific information. Since the
815 single data area approach is no longer used, these pointers are no
818 @defmac INIT_EXPANDERS
819 Macro called to initialize any target specific information. This macro
820 is called once per function, before generation of any RTL has begun.
821 The intention of this macro is to allow the initialization of the
822 function pointer @code{init_machine_status}.
825 @deftypevar {void (*)(struct function *)} init_machine_status
826 If this function pointer is non-@code{NULL} it will be called once per
827 function, before function compilation starts, in order to allow the
828 target to perform any target specific initialization of the
829 @code{struct function} structure. It is intended that this would be
830 used to initialize the @code{machine} of that structure.
832 @code{struct machine_function} structures are expected to be freed by GC@.
833 Generally, any memory that they reference must be allocated by using
834 GC allocation, including the structure itself.
838 @section Storage Layout
839 @cindex storage layout
841 Note that the definitions of the macros in this table which are sizes or
842 alignments measured in bits do not need to be constant. They can be C
843 expressions that refer to static variables, such as the @code{target_flags}.
844 @xref{Run-time Target}.
846 @defmac BITS_BIG_ENDIAN
847 Define this macro to have the value 1 if the most significant bit in a
848 byte has the lowest number; otherwise define it to have the value zero.
849 This means that bit-field instructions count from the most significant
850 bit. If the machine has no bit-field instructions, then this must still
851 be defined, but it doesn't matter which value it is defined to. This
852 macro need not be a constant.
854 This macro does not affect the way structure fields are packed into
855 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
858 @defmac BYTES_BIG_ENDIAN
859 Define this macro to have the value 1 if the most significant byte in a
860 word has the lowest number. This macro need not be a constant.
863 @defmac WORDS_BIG_ENDIAN
864 Define this macro to have the value 1 if, in a multiword object, the
865 most significant word has the lowest number. This applies to both
866 memory locations and registers; GCC fundamentally assumes that the
867 order of words in memory is the same as the order in registers. This
868 macro need not be a constant.
871 @defmac FLOAT_WORDS_BIG_ENDIAN
872 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
873 @code{TFmode} floating point numbers are stored in memory with the word
874 containing the sign bit at the lowest address; otherwise define it to
875 have the value 0. This macro need not be a constant.
877 You need not define this macro if the ordering is the same as for
881 @defmac BITS_PER_UNIT
882 Define this macro to be the number of bits in an addressable storage
883 unit (byte). If you do not define this macro the default is 8.
886 @defmac BITS_PER_WORD
887 Number of bits in a word. If you do not define this macro, the default
888 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
891 @defmac MAX_BITS_PER_WORD
892 Maximum number of bits in a word. If this is undefined, the default is
893 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
894 largest value that @code{BITS_PER_WORD} can have at run-time.
897 @defmac UNITS_PER_WORD
898 Number of storage units in a word; normally the size of a general-purpose
899 register, a power of two from 1 or 8.
902 @defmac MIN_UNITS_PER_WORD
903 Minimum number of units in a word. If this is undefined, the default is
904 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
905 smallest value that @code{UNITS_PER_WORD} can have at run-time.
909 Width of a pointer, in bits. You must specify a value no wider than the
910 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
911 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
912 a value the default is @code{BITS_PER_WORD}.
915 @defmac POINTERS_EXTEND_UNSIGNED
916 A C expression that determines how pointers should be extended from
917 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
918 greater than zero if pointers should be zero-extended, zero if they
919 should be sign-extended, and negative if some other sort of conversion
920 is needed. In the last case, the extension is done by the target's
921 @code{ptr_extend} instruction.
923 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
924 and @code{word_mode} are all the same width.
927 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
928 A macro to update @var{m} and @var{unsignedp} when an object whose type
929 is @var{type} and which has the specified mode and signedness is to be
930 stored in a register. This macro is only called when @var{type} is a
933 On most RISC machines, which only have operations that operate on a full
934 register, define this macro to set @var{m} to @code{word_mode} if
935 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
936 cases, only integer modes should be widened because wider-precision
937 floating-point operations are usually more expensive than their narrower
940 For most machines, the macro definition does not change @var{unsignedp}.
941 However, some machines, have instructions that preferentially handle
942 either signed or unsigned quantities of certain modes. For example, on
943 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
944 sign-extend the result to 64 bits. On such machines, set
945 @var{unsignedp} according to which kind of extension is more efficient.
947 Do not define this macro if it would never modify @var{m}.
950 @deftypefn {Target Hook} {enum machine_mode} TARGET_PROMOTE_FUNCTION_MODE (const_tree @var{type}, enum machine_mode @var{mode}, int *@var{punsignedp}, const_tree @var{funtype}, int @var{for_return})
951 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
952 function return values. The target hook should return the new mode
953 and possibly change @code{*@var{punsignedp}} if the promotion should
954 change signedness. This function is called only for scalar @emph{or
957 @var{for_return} allows to distinguish the promotion of arguments and
958 return values. If it is @code{1}, a return value is being promoted and
959 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
960 If it is @code{2}, the returned mode should be that of the register in
961 which an incoming parameter is copied, or the outgoing result is computed;
962 then the hook should return the same mode as @code{promote_mode}, though
963 the signedness may be different.
965 @var{type} can be NULL when promoting function arguments of libcalls.
967 The default is to not promote arguments and return values. You can
968 also define the hook to @code{default_promote_function_mode_always_promote}
969 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
972 @defmac PARM_BOUNDARY
973 Normal alignment required for function parameters on the stack, in
974 bits. All stack parameters receive at least this much alignment
975 regardless of data type. On most machines, this is the same as the
979 @defmac STACK_BOUNDARY
980 Define this macro to the minimum alignment enforced by hardware for the
981 stack pointer on this machine. The definition is a C expression for the
982 desired alignment (measured in bits). This value is used as a default
983 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
984 this should be the same as @code{PARM_BOUNDARY}.
987 @defmac PREFERRED_STACK_BOUNDARY
988 Define this macro if you wish to preserve a certain alignment for the
989 stack pointer, greater than what the hardware enforces. The definition
990 is a C expression for the desired alignment (measured in bits). This
991 macro must evaluate to a value equal to or larger than
992 @code{STACK_BOUNDARY}.
995 @defmac INCOMING_STACK_BOUNDARY
996 Define this macro if the incoming stack boundary may be different
997 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
998 to a value equal to or larger than @code{STACK_BOUNDARY}.
1001 @defmac FUNCTION_BOUNDARY
1002 Alignment required for a function entry point, in bits.
1005 @defmac BIGGEST_ALIGNMENT
1006 Biggest alignment that any data type can require on this machine, in
1007 bits. Note that this is not the biggest alignment that is supported,
1008 just the biggest alignment that, when violated, may cause a fault.
1011 @defmac MALLOC_ABI_ALIGNMENT
1012 Alignment, in bits, a C conformant malloc implementation has to
1013 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1016 @defmac ATTRIBUTE_ALIGNED_VALUE
1017 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1018 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1021 @defmac MINIMUM_ATOMIC_ALIGNMENT
1022 If defined, the smallest alignment, in bits, that can be given to an
1023 object that can be referenced in one operation, without disturbing any
1024 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1025 on machines that don't have byte or half-word store operations.
1028 @defmac BIGGEST_FIELD_ALIGNMENT
1029 Biggest alignment that any structure or union field can require on this
1030 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1031 structure and union fields only, unless the field alignment has been set
1032 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1035 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1036 An expression for the alignment of a structure field @var{field} if the
1037 alignment computed in the usual way (including applying of
1038 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1039 alignment) is @var{computed}. It overrides alignment only if the
1040 field alignment has not been set by the
1041 @code{__attribute__ ((aligned (@var{n})))} construct.
1044 @defmac MAX_STACK_ALIGNMENT
1045 Biggest stack alignment guaranteed by the backend. Use this macro
1046 to specify the maximum alignment of a variable on stack.
1048 If not defined, the default value is @code{STACK_BOUNDARY}.
1050 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1051 @c But the fix for PR 32893 indicates that we can only guarantee
1052 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1053 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1056 @defmac MAX_OFILE_ALIGNMENT
1057 Biggest alignment supported by the object file format of this machine.
1058 Use this macro to limit the alignment which can be specified using the
1059 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1060 the default value is @code{BIGGEST_ALIGNMENT}.
1062 On systems that use ELF, the default (in @file{config/elfos.h}) is
1063 the largest supported 32-bit ELF section alignment representable on
1064 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1065 On 32-bit ELF the largest supported section alignment in bits is
1066 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1069 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1070 If defined, a C expression to compute the alignment for a variable in
1071 the static store. @var{type} is the data type, and @var{basic-align} is
1072 the alignment that the object would ordinarily have. The value of this
1073 macro is used instead of that alignment to align the object.
1075 If this macro is not defined, then @var{basic-align} is used.
1078 One use of this macro is to increase alignment of medium-size data to
1079 make it all fit in fewer cache lines. Another is to cause character
1080 arrays to be word-aligned so that @code{strcpy} calls that copy
1081 constants to character arrays can be done inline.
1084 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1085 If defined, a C expression to compute the alignment given to a constant
1086 that is being placed in memory. @var{constant} is the constant and
1087 @var{basic-align} is the alignment that the object would ordinarily
1088 have. The value of this macro is used instead of that alignment to
1091 If this macro is not defined, then @var{basic-align} is used.
1093 The typical use of this macro is to increase alignment for string
1094 constants to be word aligned so that @code{strcpy} calls that copy
1095 constants can be done inline.
1098 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1099 If defined, a C expression to compute the alignment for a variable in
1100 the local store. @var{type} is the data type, and @var{basic-align} is
1101 the alignment that the object would ordinarily have. The value of this
1102 macro is used instead of that alignment to align the object.
1104 If this macro is not defined, then @var{basic-align} is used.
1106 One use of this macro is to increase alignment of medium-size data to
1107 make it all fit in fewer cache lines.
1109 If the value of this macro has a type, it should be an unsigned type.
1112 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1113 If defined, a C expression to compute the alignment for stack slot.
1114 @var{type} is the data type, @var{mode} is the widest mode available,
1115 and @var{basic-align} is the alignment that the slot would ordinarily
1116 have. The value of this macro is used instead of that alignment to
1119 If this macro is not defined, then @var{basic-align} is used when
1120 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1123 This macro is to set alignment of stack slot to the maximum alignment
1124 of all possible modes which the slot may have.
1126 If the value of this macro has a type, it should be an unsigned type.
1129 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1130 If defined, a C expression to compute the alignment for a local
1131 variable @var{decl}.
1133 If this macro is not defined, then
1134 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1137 One use of this macro is to increase alignment of medium-size data to
1138 make it all fit in fewer cache lines.
1140 If the value of this macro has a type, it should be an unsigned type.
1143 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1144 If defined, a C expression to compute the minimum required alignment
1145 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1146 @var{mode}, assuming normal alignment @var{align}.
1148 If this macro is not defined, then @var{align} will be used.
1151 @defmac EMPTY_FIELD_BOUNDARY
1152 Alignment in bits to be given to a structure bit-field that follows an
1153 empty field such as @code{int : 0;}.
1155 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1158 @defmac STRUCTURE_SIZE_BOUNDARY
1159 Number of bits which any structure or union's size must be a multiple of.
1160 Each structure or union's size is rounded up to a multiple of this.
1162 If you do not define this macro, the default is the same as
1163 @code{BITS_PER_UNIT}.
1166 @defmac STRICT_ALIGNMENT
1167 Define this macro to be the value 1 if instructions will fail to work
1168 if given data not on the nominal alignment. If instructions will merely
1169 go slower in that case, define this macro as 0.
1172 @defmac PCC_BITFIELD_TYPE_MATTERS
1173 Define this if you wish to imitate the way many other C compilers handle
1174 alignment of bit-fields and the structures that contain them.
1176 The behavior is that the type written for a named bit-field (@code{int},
1177 @code{short}, or other integer type) imposes an alignment for the entire
1178 structure, as if the structure really did contain an ordinary field of
1179 that type. In addition, the bit-field is placed within the structure so
1180 that it would fit within such a field, not crossing a boundary for it.
1182 Thus, on most machines, a named bit-field whose type is written as
1183 @code{int} would not cross a four-byte boundary, and would force
1184 four-byte alignment for the whole structure. (The alignment used may
1185 not be four bytes; it is controlled by the other alignment parameters.)
1187 An unnamed bit-field will not affect the alignment of the containing
1190 If the macro is defined, its definition should be a C expression;
1191 a nonzero value for the expression enables this behavior.
1193 Note that if this macro is not defined, or its value is zero, some
1194 bit-fields may cross more than one alignment boundary. The compiler can
1195 support such references if there are @samp{insv}, @samp{extv}, and
1196 @samp{extzv} insns that can directly reference memory.
1198 The other known way of making bit-fields work is to define
1199 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1200 Then every structure can be accessed with fullwords.
1202 Unless the machine has bit-field instructions or you define
1203 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1204 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1206 If your aim is to make GCC use the same conventions for laying out
1207 bit-fields as are used by another compiler, here is how to investigate
1208 what the other compiler does. Compile and run this program:
1227 printf ("Size of foo1 is %d\n",
1228 sizeof (struct foo1));
1229 printf ("Size of foo2 is %d\n",
1230 sizeof (struct foo2));
1235 If this prints 2 and 5, then the compiler's behavior is what you would
1236 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1239 @defmac BITFIELD_NBYTES_LIMITED
1240 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1241 to aligning a bit-field within the structure.
1244 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1245 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1246 whether unnamed bitfields affect the alignment of the containing
1247 structure. The hook should return true if the structure should inherit
1248 the alignment requirements of an unnamed bitfield's type.
1251 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1252 This target hook should return @code{true} if accesses to volatile bitfields
1253 should use the narrowest mode possible. It should return @code{false} if
1254 these accesses should use the bitfield container type.
1256 The default is @code{!TARGET_STRICT_ALIGN}.
1259 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1260 Return 1 if a structure or array containing @var{field} should be accessed using
1263 If @var{field} is the only field in the structure, @var{mode} is its
1264 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1265 case where structures of one field would require the structure's mode to
1266 retain the field's mode.
1268 Normally, this is not needed.
1271 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1272 Define this macro as an expression for the alignment of a type (given
1273 by @var{type} as a tree node) if the alignment computed in the usual
1274 way is @var{computed} and the alignment explicitly specified was
1277 The default is to use @var{specified} if it is larger; otherwise, use
1278 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1281 @defmac MAX_FIXED_MODE_SIZE
1282 An integer expression for the size in bits of the largest integer
1283 machine mode that should actually be used. All integer machine modes of
1284 this size or smaller can be used for structures and unions with the
1285 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1286 (DImode)} is assumed.
1289 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1290 If defined, an expression of type @code{enum machine_mode} that
1291 specifies the mode of the save area operand of a
1292 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1293 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1294 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1295 having its mode specified.
1297 You need not define this macro if it always returns @code{Pmode}. You
1298 would most commonly define this macro if the
1299 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1303 @defmac STACK_SIZE_MODE
1304 If defined, an expression of type @code{enum machine_mode} that
1305 specifies the mode of the size increment operand of an
1306 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1308 You need not define this macro if it always returns @code{word_mode}.
1309 You would most commonly define this macro if the @code{allocate_stack}
1310 pattern needs to support both a 32- and a 64-bit mode.
1313 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE (void)
1314 This target hook should return the mode to be used for the return value
1315 of compare instructions expanded to libgcc calls. If not defined
1316 @code{word_mode} is returned which is the right choice for a majority of
1320 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1321 This target hook should return the mode to be used for the shift count operand
1322 of shift instructions expanded to libgcc calls. If not defined
1323 @code{word_mode} is returned which is the right choice for a majority of
1327 @deftypefn {Target Hook} {enum machine_mode} TARGET_UNWIND_WORD_MODE (void)
1328 Return machine mode to be used for @code{_Unwind_Word} type.
1329 The default is to use @code{word_mode}.
1332 @defmac ROUND_TOWARDS_ZERO
1333 If defined, this macro should be true if the prevailing rounding
1334 mode is towards zero.
1336 Defining this macro only affects the way @file{libgcc.a} emulates
1337 floating-point arithmetic.
1339 Not defining this macro is equivalent to returning zero.
1342 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1343 This macro should return true if floats with @var{size}
1344 bits do not have a NaN or infinity representation, but use the largest
1345 exponent for normal numbers instead.
1347 Defining this macro only affects the way @file{libgcc.a} emulates
1348 floating-point arithmetic.
1350 The default definition of this macro returns false for all sizes.
1353 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1354 This target hook returns @code{true} if bit-fields in the given
1355 @var{record_type} are to be laid out following the rules of Microsoft
1356 Visual C/C++, namely: (i) a bit-field won't share the same storage
1357 unit with the previous bit-field if their underlying types have
1358 different sizes, and the bit-field will be aligned to the highest
1359 alignment of the underlying types of itself and of the previous
1360 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1361 the whole enclosing structure, even if it is unnamed; except that
1362 (iii) a zero-sized bit-field will be disregarded unless it follows
1363 another bit-field of nonzero size. If this hook returns @code{true},
1364 other macros that control bit-field layout are ignored.
1366 When a bit-field is inserted into a packed record, the whole size
1367 of the underlying type is used by one or more same-size adjacent
1368 bit-fields (that is, if its long:3, 32 bits is used in the record,
1369 and any additional adjacent long bit-fields are packed into the same
1370 chunk of 32 bits. However, if the size changes, a new field of that
1371 size is allocated). In an unpacked record, this is the same as using
1372 alignment, but not equivalent when packing.
1374 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1375 the latter will take precedence. If @samp{__attribute__((packed))} is
1376 used on a single field when MS bit-fields are in use, it will take
1377 precedence for that field, but the alignment of the rest of the structure
1378 may affect its placement.
1381 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1382 Returns true if the target supports decimal floating point.
1385 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1386 Returns true if the target supports fixed-point arithmetic.
1389 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1390 This hook is called just before expansion into rtl, allowing the target
1391 to perform additional initializations or analysis before the expansion.
1392 For example, the rs6000 port uses it to allocate a scratch stack slot
1393 for use in copying SDmode values between memory and floating point
1394 registers whenever the function being expanded has any SDmode
1398 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1399 This hook allows the backend to perform additional instantiations on rtl
1400 that are not actually in any insns yet, but will be later.
1403 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1404 If your target defines any fundamental types, or any types your target
1405 uses should be mangled differently from the default, define this hook
1406 to return the appropriate encoding for these types as part of a C++
1407 mangled name. The @var{type} argument is the tree structure representing
1408 the type to be mangled. The hook may be applied to trees which are
1409 not target-specific fundamental types; it should return @code{NULL}
1410 for all such types, as well as arguments it does not recognize. If the
1411 return value is not @code{NULL}, it must point to a statically-allocated
1414 Target-specific fundamental types might be new fundamental types or
1415 qualified versions of ordinary fundamental types. Encode new
1416 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1417 is the name used for the type in source code, and @var{n} is the
1418 length of @var{name} in decimal. Encode qualified versions of
1419 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1420 @var{name} is the name used for the type qualifier in source code,
1421 @var{n} is the length of @var{name} as above, and @var{code} is the
1422 code used to represent the unqualified version of this type. (See
1423 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1424 codes.) In both cases the spaces are for clarity; do not include any
1425 spaces in your string.
1427 This hook is applied to types prior to typedef resolution. If the mangled
1428 name for a particular type depends only on that type's main variant, you
1429 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1432 The default version of this hook always returns @code{NULL}, which is
1433 appropriate for a target that does not define any new fundamental
1438 @section Layout of Source Language Data Types
1440 These macros define the sizes and other characteristics of the standard
1441 basic data types used in programs being compiled. Unlike the macros in
1442 the previous section, these apply to specific features of C and related
1443 languages, rather than to fundamental aspects of storage layout.
1445 @defmac INT_TYPE_SIZE
1446 A C expression for the size in bits of the type @code{int} on the
1447 target machine. If you don't define this, the default is one word.
1450 @defmac SHORT_TYPE_SIZE
1451 A C expression for the size in bits of the type @code{short} on the
1452 target machine. If you don't define this, the default is half a word.
1453 (If this would be less than one storage unit, it is rounded up to one
1457 @defmac LONG_TYPE_SIZE
1458 A C expression for the size in bits of the type @code{long} on the
1459 target machine. If you don't define this, the default is one word.
1462 @defmac ADA_LONG_TYPE_SIZE
1463 On some machines, the size used for the Ada equivalent of the type
1464 @code{long} by a native Ada compiler differs from that used by C@. In
1465 that situation, define this macro to be a C expression to be used for
1466 the size of that type. If you don't define this, the default is the
1467 value of @code{LONG_TYPE_SIZE}.
1470 @defmac LONG_LONG_TYPE_SIZE
1471 A C expression for the size in bits of the type @code{long long} on the
1472 target machine. If you don't define this, the default is two
1473 words. If you want to support GNU Ada on your machine, the value of this
1474 macro must be at least 64.
1477 @defmac CHAR_TYPE_SIZE
1478 A C expression for the size in bits of the type @code{char} on the
1479 target machine. If you don't define this, the default is
1480 @code{BITS_PER_UNIT}.
1483 @defmac BOOL_TYPE_SIZE
1484 A C expression for the size in bits of the C++ type @code{bool} and
1485 C99 type @code{_Bool} on the target machine. If you don't define
1486 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1489 @defmac FLOAT_TYPE_SIZE
1490 A C expression for the size in bits of the type @code{float} on the
1491 target machine. If you don't define this, the default is one word.
1494 @defmac DOUBLE_TYPE_SIZE
1495 A C expression for the size in bits of the type @code{double} on the
1496 target machine. If you don't define this, the default is two
1500 @defmac LONG_DOUBLE_TYPE_SIZE
1501 A C expression for the size in bits of the type @code{long double} on
1502 the target machine. If you don't define this, the default is two
1506 @defmac SHORT_FRACT_TYPE_SIZE
1507 A C expression for the size in bits of the type @code{short _Fract} on
1508 the target machine. If you don't define this, the default is
1509 @code{BITS_PER_UNIT}.
1512 @defmac FRACT_TYPE_SIZE
1513 A C expression for the size in bits of the type @code{_Fract} on
1514 the target machine. If you don't define this, the default is
1515 @code{BITS_PER_UNIT * 2}.
1518 @defmac LONG_FRACT_TYPE_SIZE
1519 A C expression for the size in bits of the type @code{long _Fract} on
1520 the target machine. If you don't define this, the default is
1521 @code{BITS_PER_UNIT * 4}.
1524 @defmac LONG_LONG_FRACT_TYPE_SIZE
1525 A C expression for the size in bits of the type @code{long long _Fract} on
1526 the target machine. If you don't define this, the default is
1527 @code{BITS_PER_UNIT * 8}.
1530 @defmac SHORT_ACCUM_TYPE_SIZE
1531 A C expression for the size in bits of the type @code{short _Accum} on
1532 the target machine. If you don't define this, the default is
1533 @code{BITS_PER_UNIT * 2}.
1536 @defmac ACCUM_TYPE_SIZE
1537 A C expression for the size in bits of the type @code{_Accum} on
1538 the target machine. If you don't define this, the default is
1539 @code{BITS_PER_UNIT * 4}.
1542 @defmac LONG_ACCUM_TYPE_SIZE
1543 A C expression for the size in bits of the type @code{long _Accum} on
1544 the target machine. If you don't define this, the default is
1545 @code{BITS_PER_UNIT * 8}.
1548 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1549 A C expression for the size in bits of the type @code{long long _Accum} on
1550 the target machine. If you don't define this, the default is
1551 @code{BITS_PER_UNIT * 16}.
1554 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1555 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1556 if you want routines in @file{libgcc2.a} for a size other than
1557 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1558 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1561 @defmac LIBGCC2_HAS_DF_MODE
1562 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1563 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1564 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1565 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1566 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1570 @defmac LIBGCC2_HAS_XF_MODE
1571 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1572 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1573 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1574 is 80 then the default is 1, otherwise it is 0.
1577 @defmac LIBGCC2_HAS_TF_MODE
1578 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1579 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1580 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1581 is 128 then the default is 1, otherwise it is 0.
1588 Define these macros to be the size in bits of the mantissa of
1589 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1590 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1591 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1592 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1593 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1594 @code{DOUBLE_TYPE_SIZE} or
1595 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1598 @defmac TARGET_FLT_EVAL_METHOD
1599 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1600 assuming, if applicable, that the floating-point control word is in its
1601 default state. If you do not define this macro the value of
1602 @code{FLT_EVAL_METHOD} will be zero.
1605 @defmac WIDEST_HARDWARE_FP_SIZE
1606 A C expression for the size in bits of the widest floating-point format
1607 supported by the hardware. If you define this macro, you must specify a
1608 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1609 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1613 @defmac DEFAULT_SIGNED_CHAR
1614 An expression whose value is 1 or 0, according to whether the type
1615 @code{char} should be signed or unsigned by default. The user can
1616 always override this default with the options @option{-fsigned-char}
1617 and @option{-funsigned-char}.
1620 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1621 This target hook should return true if the compiler should give an
1622 @code{enum} type only as many bytes as it takes to represent the range
1623 of possible values of that type. It should return false if all
1624 @code{enum} types should be allocated like @code{int}.
1626 The default is to return false.
1630 A C expression for a string describing the name of the data type to use
1631 for size values. The typedef name @code{size_t} is defined using the
1632 contents of the string.
1634 The string can contain more than one keyword. If so, separate them with
1635 spaces, and write first any length keyword, then @code{unsigned} if
1636 appropriate, and finally @code{int}. The string must exactly match one
1637 of the data type names defined in the function
1638 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1639 omit @code{int} or change the order---that would cause the compiler to
1642 If you don't define this macro, the default is @code{"long unsigned
1646 @defmac PTRDIFF_TYPE
1647 A C expression for a string describing the name of the data type to use
1648 for the result of subtracting two pointers. The typedef name
1649 @code{ptrdiff_t} is defined using the contents of the string. See
1650 @code{SIZE_TYPE} above for more information.
1652 If you don't define this macro, the default is @code{"long int"}.
1656 A C expression for a string describing the name of the data type to use
1657 for wide characters. The typedef name @code{wchar_t} is defined using
1658 the contents of the string. See @code{SIZE_TYPE} above for more
1661 If you don't define this macro, the default is @code{"int"}.
1664 @defmac WCHAR_TYPE_SIZE
1665 A C expression for the size in bits of the data type for wide
1666 characters. This is used in @code{cpp}, which cannot make use of
1671 A C expression for a string describing the name of the data type to
1672 use for wide characters passed to @code{printf} and returned from
1673 @code{getwc}. The typedef name @code{wint_t} is defined using the
1674 contents of the string. See @code{SIZE_TYPE} above for more
1677 If you don't define this macro, the default is @code{"unsigned int"}.
1681 A C expression for a string describing the name of the data type that
1682 can represent any value of any standard or extended signed integer type.
1683 The typedef name @code{intmax_t} is defined using the contents of the
1684 string. See @code{SIZE_TYPE} above for more information.
1686 If you don't define this macro, the default is the first of
1687 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1688 much precision as @code{long long int}.
1691 @defmac UINTMAX_TYPE
1692 A C expression for a string describing the name of the data type that
1693 can represent any value of any standard or extended unsigned integer
1694 type. The typedef name @code{uintmax_t} is defined using the contents
1695 of the string. See @code{SIZE_TYPE} above for more information.
1697 If you don't define this macro, the default is the first of
1698 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1699 unsigned int"} that has as much precision as @code{long long unsigned
1703 @defmac SIG_ATOMIC_TYPE
1709 @defmacx UINT16_TYPE
1710 @defmacx UINT32_TYPE
1711 @defmacx UINT64_TYPE
1712 @defmacx INT_LEAST8_TYPE
1713 @defmacx INT_LEAST16_TYPE
1714 @defmacx INT_LEAST32_TYPE
1715 @defmacx INT_LEAST64_TYPE
1716 @defmacx UINT_LEAST8_TYPE
1717 @defmacx UINT_LEAST16_TYPE
1718 @defmacx UINT_LEAST32_TYPE
1719 @defmacx UINT_LEAST64_TYPE
1720 @defmacx INT_FAST8_TYPE
1721 @defmacx INT_FAST16_TYPE
1722 @defmacx INT_FAST32_TYPE
1723 @defmacx INT_FAST64_TYPE
1724 @defmacx UINT_FAST8_TYPE
1725 @defmacx UINT_FAST16_TYPE
1726 @defmacx UINT_FAST32_TYPE
1727 @defmacx UINT_FAST64_TYPE
1728 @defmacx INTPTR_TYPE
1729 @defmacx UINTPTR_TYPE
1730 C expressions for the standard types @code{sig_atomic_t},
1731 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1732 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1733 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1734 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1735 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1736 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1737 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1738 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1739 @code{SIZE_TYPE} above for more information.
1741 If any of these macros evaluates to a null pointer, the corresponding
1742 type is not supported; if GCC is configured to provide
1743 @code{<stdint.h>} in such a case, the header provided may not conform
1744 to C99, depending on the type in question. The defaults for all of
1745 these macros are null pointers.
1748 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1749 The C++ compiler represents a pointer-to-member-function with a struct
1756 ptrdiff_t vtable_index;
1763 The C++ compiler must use one bit to indicate whether the function that
1764 will be called through a pointer-to-member-function is virtual.
1765 Normally, we assume that the low-order bit of a function pointer must
1766 always be zero. Then, by ensuring that the vtable_index is odd, we can
1767 distinguish which variant of the union is in use. But, on some
1768 platforms function pointers can be odd, and so this doesn't work. In
1769 that case, we use the low-order bit of the @code{delta} field, and shift
1770 the remainder of the @code{delta} field to the left.
1772 GCC will automatically make the right selection about where to store
1773 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1774 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1775 set such that functions always start at even addresses, but the lowest
1776 bit of pointers to functions indicate whether the function at that
1777 address is in ARM or Thumb mode. If this is the case of your
1778 architecture, you should define this macro to
1779 @code{ptrmemfunc_vbit_in_delta}.
1781 In general, you should not have to define this macro. On architectures
1782 in which function addresses are always even, according to
1783 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1784 @code{ptrmemfunc_vbit_in_pfn}.
1787 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1788 Normally, the C++ compiler uses function pointers in vtables. This
1789 macro allows the target to change to use ``function descriptors''
1790 instead. Function descriptors are found on targets for whom a
1791 function pointer is actually a small data structure. Normally the
1792 data structure consists of the actual code address plus a data
1793 pointer to which the function's data is relative.
1795 If vtables are used, the value of this macro should be the number
1796 of words that the function descriptor occupies.
1799 @defmac TARGET_VTABLE_ENTRY_ALIGN
1800 By default, the vtable entries are void pointers, the so the alignment
1801 is the same as pointer alignment. The value of this macro specifies
1802 the alignment of the vtable entry in bits. It should be defined only
1803 when special alignment is necessary. */
1806 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1807 There are a few non-descriptor entries in the vtable at offsets below
1808 zero. If these entries must be padded (say, to preserve the alignment
1809 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1810 of words in each data entry.
1814 @section Register Usage
1815 @cindex register usage
1817 This section explains how to describe what registers the target machine
1818 has, and how (in general) they can be used.
1820 The description of which registers a specific instruction can use is
1821 done with register classes; see @ref{Register Classes}. For information
1822 on using registers to access a stack frame, see @ref{Frame Registers}.
1823 For passing values in registers, see @ref{Register Arguments}.
1824 For returning values in registers, see @ref{Scalar Return}.
1827 * Register Basics:: Number and kinds of registers.
1828 * Allocation Order:: Order in which registers are allocated.
1829 * Values in Registers:: What kinds of values each reg can hold.
1830 * Leaf Functions:: Renumbering registers for leaf functions.
1831 * Stack Registers:: Handling a register stack such as 80387.
1834 @node Register Basics
1835 @subsection Basic Characteristics of Registers
1837 @c prevent bad page break with this line
1838 Registers have various characteristics.
1840 @defmac FIRST_PSEUDO_REGISTER
1841 Number of hardware registers known to the compiler. They receive
1842 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1843 pseudo register's number really is assigned the number
1844 @code{FIRST_PSEUDO_REGISTER}.
1847 @defmac FIXED_REGISTERS
1848 @cindex fixed register
1849 An initializer that says which registers are used for fixed purposes
1850 all throughout the compiled code and are therefore not available for
1851 general allocation. These would include the stack pointer, the frame
1852 pointer (except on machines where that can be used as a general
1853 register when no frame pointer is needed), the program counter on
1854 machines where that is considered one of the addressable registers,
1855 and any other numbered register with a standard use.
1857 This information is expressed as a sequence of numbers, separated by
1858 commas and surrounded by braces. The @var{n}th number is 1 if
1859 register @var{n} is fixed, 0 otherwise.
1861 The table initialized from this macro, and the table initialized by
1862 the following one, may be overridden at run time either automatically,
1863 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1864 the user with the command options @option{-ffixed-@var{reg}},
1865 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1868 @defmac CALL_USED_REGISTERS
1869 @cindex call-used register
1870 @cindex call-clobbered register
1871 @cindex call-saved register
1872 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1873 clobbered (in general) by function calls as well as for fixed
1874 registers. This macro therefore identifies the registers that are not
1875 available for general allocation of values that must live across
1878 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1879 automatically saves it on function entry and restores it on function
1880 exit, if the register is used within the function.
1883 @defmac CALL_REALLY_USED_REGISTERS
1884 @cindex call-used register
1885 @cindex call-clobbered register
1886 @cindex call-saved register
1887 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1888 that the entire set of @code{FIXED_REGISTERS} be included.
1889 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1890 This macro is optional. If not specified, it defaults to the value
1891 of @code{CALL_USED_REGISTERS}.
1894 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1895 @cindex call-used register
1896 @cindex call-clobbered register
1897 @cindex call-saved register
1898 A C expression that is nonzero if it is not permissible to store a
1899 value of mode @var{mode} in hard register number @var{regno} across a
1900 call without some part of it being clobbered. For most machines this
1901 macro need not be defined. It is only required for machines that do not
1902 preserve the entire contents of a register across a call.
1906 @findex call_used_regs
1909 @findex reg_class_contents
1910 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1911 This hook may conditionally modify five variables
1912 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1913 @code{reg_names}, and @code{reg_class_contents}, to take into account
1914 any dependence of these register sets on target flags. The first three
1915 of these are of type @code{char []} (interpreted as Boolean vectors).
1916 @code{global_regs} is a @code{const char *[]}, and
1917 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1918 called, @code{fixed_regs}, @code{call_used_regs},
1919 @code{reg_class_contents}, and @code{reg_names} have been initialized
1920 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1921 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1922 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1923 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1924 command options have been applied.
1926 @cindex disabling certain registers
1927 @cindex controlling register usage
1928 If the usage of an entire class of registers depends on the target
1929 flags, you may indicate this to GCC by using this macro to modify
1930 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1931 registers in the classes which should not be used by GCC@. Also define
1932 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1933 to return @code{NO_REGS} if it
1934 is called with a letter for a class that shouldn't be used.
1936 (However, if this class is not included in @code{GENERAL_REGS} and all
1937 of the insn patterns whose constraints permit this class are
1938 controlled by target switches, then GCC will automatically avoid using
1939 these registers when the target switches are opposed to them.)
1942 @defmac INCOMING_REGNO (@var{out})
1943 Define this macro if the target machine has register windows. This C
1944 expression returns the register number as seen by the called function
1945 corresponding to the register number @var{out} as seen by the calling
1946 function. Return @var{out} if register number @var{out} is not an
1950 @defmac OUTGOING_REGNO (@var{in})
1951 Define this macro if the target machine has register windows. This C
1952 expression returns the register number as seen by the calling function
1953 corresponding to the register number @var{in} as seen by the called
1954 function. Return @var{in} if register number @var{in} is not an inbound
1958 @defmac LOCAL_REGNO (@var{regno})
1959 Define this macro if the target machine has register windows. This C
1960 expression returns true if the register is call-saved but is in the
1961 register window. Unlike most call-saved registers, such registers
1962 need not be explicitly restored on function exit or during non-local
1967 If the program counter has a register number, define this as that
1968 register number. Otherwise, do not define it.
1971 @node Allocation Order
1972 @subsection Order of Allocation of Registers
1973 @cindex order of register allocation
1974 @cindex register allocation order
1976 @c prevent bad page break with this line
1977 Registers are allocated in order.
1979 @defmac REG_ALLOC_ORDER
1980 If defined, an initializer for a vector of integers, containing the
1981 numbers of hard registers in the order in which GCC should prefer
1982 to use them (from most preferred to least).
1984 If this macro is not defined, registers are used lowest numbered first
1985 (all else being equal).
1987 One use of this macro is on machines where the highest numbered
1988 registers must always be saved and the save-multiple-registers
1989 instruction supports only sequences of consecutive registers. On such
1990 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1991 the highest numbered allocable register first.
1994 @defmac ADJUST_REG_ALLOC_ORDER
1995 A C statement (sans semicolon) to choose the order in which to allocate
1996 hard registers for pseudo-registers local to a basic block.
1998 Store the desired register order in the array @code{reg_alloc_order}.
1999 Element 0 should be the register to allocate first; element 1, the next
2000 register; and so on.
2002 The macro body should not assume anything about the contents of
2003 @code{reg_alloc_order} before execution of the macro.
2005 On most machines, it is not necessary to define this macro.
2008 @defmac HONOR_REG_ALLOC_ORDER
2009 Normally, IRA tries to estimate the costs for saving a register in the
2010 prologue and restoring it in the epilogue. This discourages it from
2011 using call-saved registers. If a machine wants to ensure that IRA
2012 allocates registers in the order given by REG_ALLOC_ORDER even if some
2013 call-saved registers appear earlier than call-used ones, this macro
2017 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2018 In some case register allocation order is not enough for the
2019 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2020 If this macro is defined, it should return a floating point value
2021 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2022 be increased by approximately the pseudo's usage frequency times the
2023 value returned by this macro. Not defining this macro is equivalent
2024 to having it always return @code{0.0}.
2026 On most machines, it is not necessary to define this macro.
2029 @node Values in Registers
2030 @subsection How Values Fit in Registers
2032 This section discusses the macros that describe which kinds of values
2033 (specifically, which machine modes) each register can hold, and how many
2034 consecutive registers are needed for a given mode.
2036 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2037 A C expression for the number of consecutive hard registers, starting
2038 at register number @var{regno}, required to hold a value of mode
2039 @var{mode}. This macro must never return zero, even if a register
2040 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2041 and/or CANNOT_CHANGE_MODE_CLASS instead.
2043 On a machine where all registers are exactly one word, a suitable
2044 definition of this macro is
2047 #define HARD_REGNO_NREGS(REGNO, MODE) \
2048 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2053 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2054 A C expression that is nonzero if a value of mode @var{mode}, stored
2055 in memory, ends with padding that causes it to take up more space than
2056 in registers starting at register number @var{regno} (as determined by
2057 multiplying GCC's notion of the size of the register when containing
2058 this mode by the number of registers returned by
2059 @code{HARD_REGNO_NREGS}). By default this is zero.
2061 For example, if a floating-point value is stored in three 32-bit
2062 registers but takes up 128 bits in memory, then this would be
2065 This macros only needs to be defined if there are cases where
2066 @code{subreg_get_info}
2067 would otherwise wrongly determine that a @code{subreg} can be
2068 represented by an offset to the register number, when in fact such a
2069 @code{subreg} would contain some of the padding not stored in
2070 registers and so not be representable.
2073 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2074 For values of @var{regno} and @var{mode} for which
2075 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2076 returning the greater number of registers required to hold the value
2077 including any padding. In the example above, the value would be four.
2080 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2081 Define this macro if the natural size of registers that hold values
2082 of mode @var{mode} is not the word size. It is a C expression that
2083 should give the natural size in bytes for the specified mode. It is
2084 used by the register allocator to try to optimize its results. This
2085 happens for example on SPARC 64-bit where the natural size of
2086 floating-point registers is still 32-bit.
2089 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2090 A C expression that is nonzero if it is permissible to store a value
2091 of mode @var{mode} in hard register number @var{regno} (or in several
2092 registers starting with that one). For a machine where all registers
2093 are equivalent, a suitable definition is
2096 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2099 You need not include code to check for the numbers of fixed registers,
2100 because the allocation mechanism considers them to be always occupied.
2102 @cindex register pairs
2103 On some machines, double-precision values must be kept in even/odd
2104 register pairs. You can implement that by defining this macro to reject
2105 odd register numbers for such modes.
2107 The minimum requirement for a mode to be OK in a register is that the
2108 @samp{mov@var{mode}} instruction pattern support moves between the
2109 register and other hard register in the same class and that moving a
2110 value into the register and back out not alter it.
2112 Since the same instruction used to move @code{word_mode} will work for
2113 all narrower integer modes, it is not necessary on any machine for
2114 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2115 you define patterns @samp{movhi}, etc., to take advantage of this. This
2116 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2117 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2120 Many machines have special registers for floating point arithmetic.
2121 Often people assume that floating point machine modes are allowed only
2122 in floating point registers. This is not true. Any registers that
2123 can hold integers can safely @emph{hold} a floating point machine
2124 mode, whether or not floating arithmetic can be done on it in those
2125 registers. Integer move instructions can be used to move the values.
2127 On some machines, though, the converse is true: fixed-point machine
2128 modes may not go in floating registers. This is true if the floating
2129 registers normalize any value stored in them, because storing a
2130 non-floating value there would garble it. In this case,
2131 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2132 floating registers. But if the floating registers do not automatically
2133 normalize, if you can store any bit pattern in one and retrieve it
2134 unchanged without a trap, then any machine mode may go in a floating
2135 register, so you can define this macro to say so.
2137 The primary significance of special floating registers is rather that
2138 they are the registers acceptable in floating point arithmetic
2139 instructions. However, this is of no concern to
2140 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2141 constraints for those instructions.
2143 On some machines, the floating registers are especially slow to access,
2144 so that it is better to store a value in a stack frame than in such a
2145 register if floating point arithmetic is not being done. As long as the
2146 floating registers are not in class @code{GENERAL_REGS}, they will not
2147 be used unless some pattern's constraint asks for one.
2150 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2151 A C expression that is nonzero if it is OK to rename a hard register
2152 @var{from} to another hard register @var{to}.
2154 One common use of this macro is to prevent renaming of a register to
2155 another register that is not saved by a prologue in an interrupt
2158 The default is always nonzero.
2161 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2162 A C expression that is nonzero if a value of mode
2163 @var{mode1} is accessible in mode @var{mode2} without copying.
2165 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2166 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2167 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2168 should be nonzero. If they differ for any @var{r}, you should define
2169 this macro to return zero unless some other mechanism ensures the
2170 accessibility of the value in a narrower mode.
2172 You should define this macro to return nonzero in as many cases as
2173 possible since doing so will allow GCC to perform better register
2177 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2178 This target hook should return @code{true} if it is OK to use a hard register
2179 @var{regno} as scratch reg in peephole2.
2181 One common use of this macro is to prevent using of a register that
2182 is not saved by a prologue in an interrupt handler.
2184 The default version of this hook always returns @code{true}.
2187 @defmac AVOID_CCMODE_COPIES
2188 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2189 registers. You should only define this macro if support for copying to/from
2190 @code{CCmode} is incomplete.
2193 @node Leaf Functions
2194 @subsection Handling Leaf Functions
2196 @cindex leaf functions
2197 @cindex functions, leaf
2198 On some machines, a leaf function (i.e., one which makes no calls) can run
2199 more efficiently if it does not make its own register window. Often this
2200 means it is required to receive its arguments in the registers where they
2201 are passed by the caller, instead of the registers where they would
2204 The special treatment for leaf functions generally applies only when
2205 other conditions are met; for example, often they may use only those
2206 registers for its own variables and temporaries. We use the term ``leaf
2207 function'' to mean a function that is suitable for this special
2208 handling, so that functions with no calls are not necessarily ``leaf
2211 GCC assigns register numbers before it knows whether the function is
2212 suitable for leaf function treatment. So it needs to renumber the
2213 registers in order to output a leaf function. The following macros
2216 @defmac LEAF_REGISTERS
2217 Name of a char vector, indexed by hard register number, which
2218 contains 1 for a register that is allowable in a candidate for leaf
2221 If leaf function treatment involves renumbering the registers, then the
2222 registers marked here should be the ones before renumbering---those that
2223 GCC would ordinarily allocate. The registers which will actually be
2224 used in the assembler code, after renumbering, should not be marked with 1
2227 Define this macro only if the target machine offers a way to optimize
2228 the treatment of leaf functions.
2231 @defmac LEAF_REG_REMAP (@var{regno})
2232 A C expression whose value is the register number to which @var{regno}
2233 should be renumbered, when a function is treated as a leaf function.
2235 If @var{regno} is a register number which should not appear in a leaf
2236 function before renumbering, then the expression should yield @minus{}1, which
2237 will cause the compiler to abort.
2239 Define this macro only if the target machine offers a way to optimize the
2240 treatment of leaf functions, and registers need to be renumbered to do
2244 @findex current_function_is_leaf
2245 @findex current_function_uses_only_leaf_regs
2246 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2247 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2248 specially. They can test the C variable @code{current_function_is_leaf}
2249 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2250 set prior to local register allocation and is valid for the remaining
2251 compiler passes. They can also test the C variable
2252 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2253 functions which only use leaf registers.
2254 @code{current_function_uses_only_leaf_regs} is valid after all passes
2255 that modify the instructions have been run and is only useful if
2256 @code{LEAF_REGISTERS} is defined.
2257 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2258 @c of the next paragraph?! --mew 2feb93
2260 @node Stack Registers
2261 @subsection Registers That Form a Stack
2263 There are special features to handle computers where some of the
2264 ``registers'' form a stack. Stack registers are normally written by
2265 pushing onto the stack, and are numbered relative to the top of the
2268 Currently, GCC can only handle one group of stack-like registers, and
2269 they must be consecutively numbered. Furthermore, the existing
2270 support for stack-like registers is specific to the 80387 floating
2271 point coprocessor. If you have a new architecture that uses
2272 stack-like registers, you will need to do substantial work on
2273 @file{reg-stack.c} and write your machine description to cooperate
2274 with it, as well as defining these macros.
2277 Define this if the machine has any stack-like registers.
2280 @defmac STACK_REG_COVER_CLASS
2281 This is a cover class containing the stack registers. Define this if
2282 the machine has any stack-like registers.
2285 @defmac FIRST_STACK_REG
2286 The number of the first stack-like register. This one is the top
2290 @defmac LAST_STACK_REG
2291 The number of the last stack-like register. This one is the bottom of
2295 @node Register Classes
2296 @section Register Classes
2297 @cindex register class definitions
2298 @cindex class definitions, register
2300 On many machines, the numbered registers are not all equivalent.
2301 For example, certain registers may not be allowed for indexed addressing;
2302 certain registers may not be allowed in some instructions. These machine
2303 restrictions are described to the compiler using @dfn{register classes}.
2305 You define a number of register classes, giving each one a name and saying
2306 which of the registers belong to it. Then you can specify register classes
2307 that are allowed as operands to particular instruction patterns.
2311 In general, each register will belong to several classes. In fact, one
2312 class must be named @code{ALL_REGS} and contain all the registers. Another
2313 class must be named @code{NO_REGS} and contain no registers. Often the
2314 union of two classes will be another class; however, this is not required.
2316 @findex GENERAL_REGS
2317 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2318 terribly special about the name, but the operand constraint letters
2319 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2320 the same as @code{ALL_REGS}, just define it as a macro which expands
2323 Order the classes so that if class @var{x} is contained in class @var{y}
2324 then @var{x} has a lower class number than @var{y}.
2326 The way classes other than @code{GENERAL_REGS} are specified in operand
2327 constraints is through machine-dependent operand constraint letters.
2328 You can define such letters to correspond to various classes, then use
2329 them in operand constraints.
2331 You should define a class for the union of two classes whenever some
2332 instruction allows both classes. For example, if an instruction allows
2333 either a floating point (coprocessor) register or a general register for a
2334 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2335 which includes both of them. Otherwise you will get suboptimal code,
2336 or even internal compiler errors when reload cannot find a register in the
2337 class computed via @code{reg_class_subunion}.
2339 You must also specify certain redundant information about the register
2340 classes: for each class, which classes contain it and which ones are
2341 contained in it; for each pair of classes, the largest class contained
2344 When a value occupying several consecutive registers is expected in a
2345 certain class, all the registers used must belong to that class.
2346 Therefore, register classes cannot be used to enforce a requirement for
2347 a register pair to start with an even-numbered register. The way to
2348 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2350 Register classes used for input-operands of bitwise-and or shift
2351 instructions have a special requirement: each such class must have, for
2352 each fixed-point machine mode, a subclass whose registers can transfer that
2353 mode to or from memory. For example, on some machines, the operations for
2354 single-byte values (@code{QImode}) are limited to certain registers. When
2355 this is so, each register class that is used in a bitwise-and or shift
2356 instruction must have a subclass consisting of registers from which
2357 single-byte values can be loaded or stored. This is so that
2358 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2360 @deftp {Data type} {enum reg_class}
2361 An enumerated type that must be defined with all the register class names
2362 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2363 must be the last register class, followed by one more enumerated value,
2364 @code{LIM_REG_CLASSES}, which is not a register class but rather
2365 tells how many classes there are.
2367 Each register class has a number, which is the value of casting
2368 the class name to type @code{int}. The number serves as an index
2369 in many of the tables described below.
2372 @defmac N_REG_CLASSES
2373 The number of distinct register classes, defined as follows:
2376 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2380 @defmac REG_CLASS_NAMES
2381 An initializer containing the names of the register classes as C string
2382 constants. These names are used in writing some of the debugging dumps.
2385 @defmac REG_CLASS_CONTENTS
2386 An initializer containing the contents of the register classes, as integers
2387 which are bit masks. The @var{n}th integer specifies the contents of class
2388 @var{n}. The way the integer @var{mask} is interpreted is that
2389 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2391 When the machine has more than 32 registers, an integer does not suffice.
2392 Then the integers are replaced by sub-initializers, braced groupings containing
2393 several integers. Each sub-initializer must be suitable as an initializer
2394 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2395 In this situation, the first integer in each sub-initializer corresponds to
2396 registers 0 through 31, the second integer to registers 32 through 63, and
2400 @defmac REGNO_REG_CLASS (@var{regno})
2401 A C expression whose value is a register class containing hard register
2402 @var{regno}. In general there is more than one such class; choose a class
2403 which is @dfn{minimal}, meaning that no smaller class also contains the
2407 @defmac BASE_REG_CLASS
2408 A macro whose definition is the name of the class to which a valid
2409 base register must belong. A base register is one used in an address
2410 which is the register value plus a displacement.
2413 @defmac MODE_BASE_REG_CLASS (@var{mode})
2414 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2415 the selection of a base register in a mode dependent manner. If
2416 @var{mode} is VOIDmode then it should return the same value as
2417 @code{BASE_REG_CLASS}.
2420 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2421 A C expression whose value is the register class to which a valid
2422 base register must belong in order to be used in a base plus index
2423 register address. You should define this macro if base plus index
2424 addresses have different requirements than other base register uses.
2427 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2428 A C expression whose value is the register class to which a valid
2429 base register must belong. @var{outer_code} and @var{index_code} define the
2430 context in which the base register occurs. @var{outer_code} is the code of
2431 the immediately enclosing expression (@code{MEM} for the top level of an
2432 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{outer_code}, @var{index_code})
2473 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2474 that that expression may examine the context in which the register
2475 appears in the memory reference. @var{outer_code} is the code of the
2476 immediately enclosing expression (@code{MEM} if at the top level of the
2477 address, @code{ADDRESS} for something that occurs in an
2478 @code{address_operand}). @var{index_code} is the code of the
2479 corresponding index expression if @var{outer_code} is @code{PLUS};
2480 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2481 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2484 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2485 A C expression which is nonzero if register number @var{num} is
2486 suitable for use as an index register in operand addresses. It may be
2487 either a suitable hard register or a pseudo register that has been
2488 allocated such a hard register.
2490 The difference between an index register and a base register is that
2491 the index register may be scaled. If an address involves the sum of
2492 two registers, neither one of them scaled, then either one may be
2493 labeled the ``base'' and the other the ``index''; but whichever
2494 labeling is used must fit the machine's constraints of which registers
2495 may serve in each capacity. The compiler will try both labelings,
2496 looking for one that is valid, and will reload one or both registers
2497 only if neither labeling works.
2500 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2501 A target hook that places additional preference on the register class to use when it is necessary to rename a register in class @var{rclass} to another class, or perhaps @var{NO_REGS}, if no preferred register class is found or hook @code{preferred_rename_class} is not implemented. Sometimes returning a more restrictive class makes better code. For example, on ARM, thumb-2 instructions using @code{LO_REGS} may be smaller than instructions using @code{GENERIC_REGS}. By returning @code{LO_REGS} from @code{preferred_rename_class}, code size can be reduced.
2504 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2505 A target hook that places additional restrictions on the register class
2506 to use when it is necessary to copy value @var{x} into a register in class
2507 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2508 another, smaller class.
2510 The default version of this hook always returns value of @code{rclass} argument.
2512 Sometimes returning a more restrictive class makes better code. For
2513 example, on the 68000, when @var{x} is an integer constant that is in range
2514 for a @samp{moveq} instruction, the value of this macro is always
2515 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2516 Requiring a data register guarantees that a @samp{moveq} will be used.
2518 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2519 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2520 loaded into some register class. By returning @code{NO_REGS} you can
2521 force @var{x} into a memory location. For example, rs6000 can load
2522 immediate values into general-purpose registers, but does not have an
2523 instruction for loading an immediate value into a floating-point
2524 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2525 @var{x} is a floating-point constant. If the constant can't be loaded
2526 into any kind of register, code generation will be better if
2527 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2528 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2530 If an insn has pseudos in it after register allocation, reload will go
2531 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2532 to find the best one. Returning @code{NO_REGS}, in this case, makes
2533 reload add a @code{!} in front of the constraint: the x86 back-end uses
2534 this feature to discourage usage of 387 registers when math is done in
2535 the SSE registers (and vice versa).
2538 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2539 A C expression that places additional restrictions on the register class
2540 to use when it is necessary to copy value @var{x} into a register in class
2541 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2542 another, smaller class. On many machines, the following definition is
2546 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2549 Sometimes returning a more restrictive class makes better code. For
2550 example, on the 68000, when @var{x} is an integer constant that is in range
2551 for a @samp{moveq} instruction, the value of this macro is always
2552 @code{DATA_REGS} as long as @var{class} includes the data registers.
2553 Requiring a data register guarantees that a @samp{moveq} will be used.
2555 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2556 @var{class} is if @var{x} is a legitimate constant which cannot be
2557 loaded into some register class. By returning @code{NO_REGS} you can
2558 force @var{x} into a memory location. For example, rs6000 can load
2559 immediate values into general-purpose registers, but does not have an
2560 instruction for loading an immediate value into a floating-point
2561 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2562 @var{x} is a floating-point constant. If the constant can't be loaded
2563 into any kind of register, code generation will be better if
2564 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2565 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2567 If an insn has pseudos in it after register allocation, reload will go
2568 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2569 to find the best one. Returning @code{NO_REGS}, in this case, makes
2570 reload add a @code{!} in front of the constraint: the x86 back-end uses
2571 this feature to discourage usage of 387 registers when math is done in
2572 the SSE registers (and vice versa).
2575 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2576 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2577 input reloads. If you don't define this macro, the default is to use
2578 @var{class}, unchanged.
2580 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2581 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2584 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2585 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2588 The default version of this hook always returns value of @code{rclass}
2591 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2592 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2595 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2596 A C expression that places additional restrictions on the register class
2597 to use when it is necessary to be able to hold a value of mode
2598 @var{mode} in a reload register for which class @var{class} would
2601 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2602 there are certain modes that simply can't go in certain reload classes.
2604 The value is a register class; perhaps @var{class}, or perhaps another,
2607 Don't define this macro unless the target machine has limitations which
2608 require the macro to do something nontrivial.
2611 @deftypefn {Target Hook} reg_class_t TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, reg_class_t @var{reload_class}, enum machine_mode @var{reload_mode}, secondary_reload_info *@var{sri})
2612 Many machines have some registers that cannot be copied directly to or
2613 from memory or even from other types of registers. An example is the
2614 @samp{MQ} register, which on most machines, can only be copied to or
2615 from general registers, but not memory. Below, we shall be using the
2616 term 'intermediate register' when a move operation cannot be performed
2617 directly, but has to be done by copying the source into the intermediate
2618 register first, and then copying the intermediate register to the
2619 destination. An intermediate register always has the same mode as
2620 source and destination. Since it holds the actual value being copied,
2621 reload might apply optimizations to re-use an intermediate register
2622 and eliding the copy from the source when it can determine that the
2623 intermediate register still holds the required value.
2625 Another kind of secondary reload is required on some machines which
2626 allow copying all registers to and from memory, but require a scratch
2627 register for stores to some memory locations (e.g., those with symbolic
2628 address on the RT, and those with certain symbolic address on the SPARC
2629 when compiling PIC)@. Scratch registers need not have the same mode
2630 as the value being copied, and usually hold a different value than
2631 that being copied. Special patterns in the md file are needed to
2632 describe how the copy is performed with the help of the scratch register;
2633 these patterns also describe the number, register class(es) and mode(s)
2634 of the scratch register(s).
2636 In some cases, both an intermediate and a scratch register are required.
2638 For input reloads, this target hook is called with nonzero @var{in_p},
2639 and @var{x} is an rtx that needs to be copied to a register of class
2640 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2641 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2642 needs to be copied to rtx @var{x} in @var{reload_mode}.
2644 If copying a register of @var{reload_class} from/to @var{x} requires
2645 an intermediate register, the hook @code{secondary_reload} should
2646 return the register class required for this intermediate register.
2647 If no intermediate register is required, it should return NO_REGS.
2648 If more than one intermediate register is required, describe the one
2649 that is closest in the copy chain to the reload register.
2651 If scratch registers are needed, you also have to describe how to
2652 perform the copy from/to the reload register to/from this
2653 closest intermediate register. Or if no intermediate register is
2654 required, but still a scratch register is needed, describe the
2655 copy from/to the reload register to/from the reload operand @var{x}.
2657 You do this by setting @code{sri->icode} to the instruction code of a pattern
2658 in the md file which performs the move. Operands 0 and 1 are the output
2659 and input of this copy, respectively. Operands from operand 2 onward are
2660 for scratch operands. These scratch operands must have a mode, and a
2661 single-register-class
2662 @c [later: or memory]
2665 When an intermediate register is used, the @code{secondary_reload}
2666 hook will be called again to determine how to copy the intermediate
2667 register to/from the reload operand @var{x}, so your hook must also
2668 have code to handle the register class of the intermediate operand.
2670 @c [For later: maybe we'll allow multi-alternative reload patterns -
2671 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2672 @c and match the constraints of input and output to determine the required
2673 @c alternative. A restriction would be that constraints used to match
2674 @c against reloads registers would have to be written as register class
2675 @c constraints, or we need a new target macro / hook that tells us if an
2676 @c arbitrary constraint can match an unknown register of a given class.
2677 @c Such a macro / hook would also be useful in other places.]
2680 @var{x} might be a pseudo-register or a @code{subreg} of a
2681 pseudo-register, which could either be in a hard register or in memory.
2682 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2683 in memory and the hard register number if it is in a register.
2685 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2686 currently not supported. For the time being, you will have to continue
2687 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2689 @code{copy_cost} also uses this target hook to find out how values are
2690 copied. If you want it to include some extra cost for the need to allocate
2691 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2692 Or if two dependent moves are supposed to have a lower cost than the sum
2693 of the individual moves due to expected fortuitous scheduling and/or special
2694 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2697 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2698 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2699 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2700 These macros are obsolete, new ports should use the target hook
2701 @code{TARGET_SECONDARY_RELOAD} instead.
2703 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2704 target hook. Older ports still define these macros to indicate to the
2705 reload phase that it may
2706 need to allocate at least one register for a reload in addition to the
2707 register to contain the data. Specifically, if copying @var{x} to a
2708 register @var{class} in @var{mode} requires an intermediate register,
2709 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2710 largest register class all of whose registers can be used as
2711 intermediate registers or scratch registers.
2713 If copying a register @var{class} in @var{mode} to @var{x} requires an
2714 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2715 was supposed to be defined be defined to return the largest register
2716 class required. If the
2717 requirements for input and output reloads were the same, the macro
2718 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2721 The values returned by these macros are often @code{GENERAL_REGS}.
2722 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2723 can be directly copied to or from a register of @var{class} in
2724 @var{mode} without requiring a scratch register. Do not define this
2725 macro if it would always return @code{NO_REGS}.
2727 If a scratch register is required (either with or without an
2728 intermediate register), you were supposed to define patterns for
2729 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2730 (@pxref{Standard Names}. These patterns, which were normally
2731 implemented with a @code{define_expand}, should be similar to the
2732 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2735 These patterns need constraints for the reload register and scratch
2737 contain a single register class. If the original reload register (whose
2738 class is @var{class}) can meet the constraint given in the pattern, the
2739 value returned by these macros is used for the class of the scratch
2740 register. Otherwise, two additional reload registers are required.
2741 Their classes are obtained from the constraints in the insn pattern.
2743 @var{x} might be a pseudo-register or a @code{subreg} of a
2744 pseudo-register, which could either be in a hard register or in memory.
2745 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2746 in memory and the hard register number if it is in a register.
2748 These macros should not be used in the case where a particular class of
2749 registers can only be copied to memory and not to another class of
2750 registers. In that case, secondary reload registers are not needed and
2751 would not be helpful. Instead, a stack location must be used to perform
2752 the copy and the @code{mov@var{m}} pattern should use memory as an
2753 intermediate storage. This case often occurs between floating-point and
2757 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2758 Certain machines have the property that some registers cannot be copied
2759 to some other registers without using memory. Define this macro on
2760 those machines to be a C expression that is nonzero if objects of mode
2761 @var{m} in registers of @var{class1} can only be copied to registers of
2762 class @var{class2} by storing a register of @var{class1} into memory
2763 and loading that memory location into a register of @var{class2}.
2765 Do not define this macro if its value would always be zero.
2768 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2769 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2770 allocates a stack slot for a memory location needed for register copies.
2771 If this macro is defined, the compiler instead uses the memory location
2772 defined by this macro.
2774 Do not define this macro if you do not define
2775 @code{SECONDARY_MEMORY_NEEDED}.
2778 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2779 When the compiler needs a secondary memory location to copy between two
2780 registers of mode @var{mode}, it normally allocates sufficient memory to
2781 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2782 load operations in a mode that many bits wide and whose class is the
2783 same as that of @var{mode}.
2785 This is right thing to do on most machines because it ensures that all
2786 bits of the register are copied and prevents accesses to the registers
2787 in a narrower mode, which some machines prohibit for floating-point
2790 However, this default behavior is not correct on some machines, such as
2791 the DEC Alpha, that store short integers in floating-point registers
2792 differently than in integer registers. On those machines, the default
2793 widening will not work correctly and you must define this macro to
2794 suppress that widening in some cases. See the file @file{alpha.h} for
2797 Do not define this macro if you do not define
2798 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2799 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2802 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2803 A target hook which returns @code{true} if pseudos that have been assigned
2804 to registers of class @var{rclass} would likely be spilled because
2805 registers of @var{rclass} are needed for spill registers.
2807 The default version of this target hook returns @code{true} if @var{rclass}
2808 has exactly one register and @code{false} otherwise. On most machines, this
2809 default should be used. Only use this target hook to some other expression
2810 if pseudos allocated by @file{local-alloc.c} end up in memory because their
2811 hard registers were needed for spill registers. If this target hook returns
2812 @code{false} for those classes, those pseudos will only be allocated by
2813 @file{global.c}, which knows how to reallocate the pseudo to another
2814 register. If there would not be another register available for reallocation,
2815 you should not change the implementation of this target hook since
2816 the only effect of such implementation would be to slow down register
2820 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2821 A C expression for the maximum number of consecutive registers
2822 of class @var{class} needed to hold a value of mode @var{mode}.
2824 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2825 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2826 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2827 @var{mode})} for all @var{regno} values in the class @var{class}.
2829 This macro helps control the handling of multiple-word values
2833 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2834 If defined, a C expression that returns nonzero for a @var{class} for which
2835 a change from mode @var{from} to mode @var{to} is invalid.
2837 For the example, loading 32-bit integer or floating-point objects into
2838 floating-point registers on the Alpha extends them to 64 bits.
2839 Therefore loading a 64-bit object and then storing it as a 32-bit object
2840 does not store the low-order 32 bits, as would be the case for a normal
2841 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2845 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2846 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2847 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2851 @node Old Constraints
2852 @section Obsolete Macros for Defining Constraints
2853 @cindex defining constraints, obsolete method
2854 @cindex constraints, defining, obsolete method
2856 Machine-specific constraints can be defined with these macros instead
2857 of the machine description constructs described in @ref{Define
2858 Constraints}. This mechanism is obsolete. New ports should not use
2859 it; old ports should convert to the new mechanism.
2861 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2862 For the constraint at the start of @var{str}, which starts with the letter
2863 @var{c}, return the length. This allows you to have register class /
2864 constant / extra constraints that are longer than a single letter;
2865 you don't need to define this macro if you can do with single-letter
2866 constraints only. The definition of this macro should use
2867 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2868 to handle specially.
2869 There are some sanity checks in genoutput.c that check the constraint lengths
2870 for the md file, so you can also use this macro to help you while you are
2871 transitioning from a byzantine single-letter-constraint scheme: when you
2872 return a negative length for a constraint you want to re-use, genoutput
2873 will complain about every instance where it is used in the md file.
2876 @defmac REG_CLASS_FROM_LETTER (@var{char})
2877 A C expression which defines the machine-dependent operand constraint
2878 letters for register classes. If @var{char} is such a letter, the
2879 value should be the register class corresponding to it. Otherwise,
2880 the value should be @code{NO_REGS}. The register letter @samp{r},
2881 corresponding to class @code{GENERAL_REGS}, will not be passed
2882 to this macro; you do not need to handle it.
2885 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2886 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2887 passed in @var{str}, so that you can use suffixes to distinguish between
2891 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2892 A C expression that defines the machine-dependent operand constraint
2893 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2894 particular ranges of integer values. If @var{c} is one of those
2895 letters, the expression should check that @var{value}, an integer, is in
2896 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2897 not one of those letters, the value should be 0 regardless of
2901 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2902 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2903 string passed in @var{str}, so that you can use suffixes to distinguish
2904 between different variants.
2907 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2908 A C expression that defines the machine-dependent operand constraint
2909 letters that specify particular ranges of @code{const_double} values
2910 (@samp{G} or @samp{H}).
2912 If @var{c} is one of those letters, the expression should check that
2913 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2914 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2915 letters, the value should be 0 regardless of @var{value}.
2917 @code{const_double} is used for all floating-point constants and for
2918 @code{DImode} fixed-point constants. A given letter can accept either
2919 or both kinds of values. It can use @code{GET_MODE} to distinguish
2920 between these kinds.
2923 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2924 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2925 string passed in @var{str}, so that you can use suffixes to distinguish
2926 between different variants.
2929 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2930 A C expression that defines the optional machine-dependent constraint
2931 letters that can be used to segregate specific types of operands, usually
2932 memory references, for the target machine. Any letter that is not
2933 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2934 @code{REG_CLASS_FROM_CONSTRAINT}
2935 may be used. Normally this macro will not be defined.
2937 If it is required for a particular target machine, it should return 1
2938 if @var{value} corresponds to the operand type represented by the
2939 constraint letter @var{c}. If @var{c} is not defined as an extra
2940 constraint, the value returned should be 0 regardless of @var{value}.
2942 For example, on the ROMP, load instructions cannot have their output
2943 in r0 if the memory reference contains a symbolic address. Constraint
2944 letter @samp{Q} is defined as representing a memory address that does
2945 @emph{not} contain a symbolic address. An alternative is specified with
2946 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2947 alternative specifies @samp{m} on the input and a register class that
2948 does not include r0 on the output.
2951 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2952 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2953 in @var{str}, so that you can use suffixes to distinguish between different
2957 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2958 A C expression that defines the optional machine-dependent constraint
2959 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2960 be treated like memory constraints by the reload pass.
2962 It should return 1 if the operand type represented by the constraint
2963 at the start of @var{str}, the first letter of which is the letter @var{c},
2964 comprises a subset of all memory references including
2965 all those whose address is simply a base register. This allows the reload
2966 pass to reload an operand, if it does not directly correspond to the operand
2967 type of @var{c}, by copying its address into a base register.
2969 For example, on the S/390, some instructions do not accept arbitrary
2970 memory references, but only those that do not make use of an index
2971 register. The constraint letter @samp{Q} is defined via
2972 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2973 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2974 a @samp{Q} constraint can handle any memory operand, because the
2975 reload pass knows it can be reloaded by copying the memory address
2976 into a base register if required. This is analogous to the way
2977 an @samp{o} constraint can handle any memory operand.
2980 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2981 A C expression that defines the optional machine-dependent constraint
2982 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2983 @code{EXTRA_CONSTRAINT_STR}, that should
2984 be treated like address constraints by the reload pass.
2986 It should return 1 if the operand type represented by the constraint
2987 at the start of @var{str}, which starts with the letter @var{c}, comprises
2988 a subset of all memory addresses including
2989 all those that consist of just a base register. This allows the reload
2990 pass to reload an operand, if it does not directly correspond to the operand
2991 type of @var{str}, by copying it into a base register.
2993 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2994 be used with the @code{address_operand} predicate. It is treated
2995 analogously to the @samp{p} constraint.
2998 @node Stack and Calling
2999 @section Stack Layout and Calling Conventions
3000 @cindex calling conventions
3002 @c prevent bad page break with this line
3003 This describes the stack layout and calling conventions.
3007 * Exception Handling::
3012 * Register Arguments::
3014 * Aggregate Return::
3019 * Stack Smashing Protection::
3023 @subsection Basic Stack Layout
3024 @cindex stack frame layout
3025 @cindex frame layout
3027 @c prevent bad page break with this line
3028 Here is the basic stack layout.
3030 @defmac STACK_GROWS_DOWNWARD
3031 Define this macro if pushing a word onto the stack moves the stack
3032 pointer to a smaller address.
3034 When we say, ``define this macro if @dots{}'', it means that the
3035 compiler checks this macro only with @code{#ifdef} so the precise
3036 definition used does not matter.
3039 @defmac STACK_PUSH_CODE
3040 This macro defines the operation used when something is pushed
3041 on the stack. In RTL, a push operation will be
3042 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3044 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3045 and @code{POST_INC}. Which of these is correct depends on
3046 the stack direction and on whether the stack pointer points
3047 to the last item on the stack or whether it points to the
3048 space for the next item on the stack.
3050 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3051 defined, which is almost always right, and @code{PRE_INC} otherwise,
3052 which is often wrong.
3055 @defmac FRAME_GROWS_DOWNWARD
3056 Define this macro to nonzero value if the addresses of local variable slots
3057 are at negative offsets from the frame pointer.
3060 @defmac ARGS_GROW_DOWNWARD
3061 Define this macro if successive arguments to a function occupy decreasing
3062 addresses on the stack.
3065 @defmac STARTING_FRAME_OFFSET
3066 Offset from the frame pointer to the first local variable slot to be allocated.
3068 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3069 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3070 Otherwise, it is found by adding the length of the first slot to the
3071 value @code{STARTING_FRAME_OFFSET}.
3072 @c i'm not sure if the above is still correct.. had to change it to get
3073 @c rid of an overfull. --mew 2feb93
3076 @defmac STACK_ALIGNMENT_NEEDED
3077 Define to zero to disable final alignment of the stack during reload.
3078 The nonzero default for this macro is suitable for most ports.
3080 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3081 is a register save block following the local block that doesn't require
3082 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3083 stack alignment and do it in the backend.
3086 @defmac STACK_POINTER_OFFSET
3087 Offset from the stack pointer register to the first location at which
3088 outgoing arguments are placed. If not specified, the default value of
3089 zero is used. This is the proper value for most machines.
3091 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3092 the first location at which outgoing arguments are placed.
3095 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3096 Offset from the argument pointer register to the first argument's
3097 address. On some machines it may depend on the data type of the
3100 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3101 the first argument's address.
3104 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3105 Offset from the stack pointer register to an item dynamically allocated
3106 on the stack, e.g., by @code{alloca}.
3108 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3109 length of the outgoing arguments. The default is correct for most
3110 machines. See @file{function.c} for details.
3113 @defmac INITIAL_FRAME_ADDRESS_RTX
3114 A C expression whose value is RTL representing the address of the initial
3115 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3116 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3117 default value will be used. Define this macro in order to make frame pointer
3118 elimination work in the presence of @code{__builtin_frame_address (count)} and
3119 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3122 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3123 A C expression whose value is RTL representing the address in a stack
3124 frame where the pointer to the caller's frame is stored. Assume that
3125 @var{frameaddr} is an RTL expression for the address of the stack frame
3128 If you don't define this macro, the default is to return the value
3129 of @var{frameaddr}---that is, the stack frame address is also the
3130 address of the stack word that points to the previous frame.
3133 @defmac SETUP_FRAME_ADDRESSES
3134 If defined, a C expression that produces the machine-specific code to
3135 setup the stack so that arbitrary frames can be accessed. For example,
3136 on the SPARC, we must flush all of the register windows to the stack
3137 before we can access arbitrary stack frames. You will seldom need to
3141 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3142 This target hook should return an rtx that is used to store
3143 the address of the current frame into the built in @code{setjmp} buffer.
3144 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3145 machines. One reason you may need to define this target hook is if
3146 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3149 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3150 A C expression whose value is RTL representing the value of the frame
3151 address for the current frame. @var{frameaddr} is the frame pointer
3152 of the current frame. This is used for __builtin_frame_address.
3153 You need only define this macro if the frame address is not the same
3154 as the frame pointer. Most machines do not need to define it.
3157 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3158 A C expression whose value is RTL representing the value of the return
3159 address for the frame @var{count} steps up from the current frame, after
3160 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3161 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3162 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3164 The value of the expression must always be the correct address when
3165 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3166 determine the return address of other frames.
3169 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3170 Define this if the return address of a particular stack frame is accessed
3171 from the frame pointer of the previous stack frame.
3174 @defmac INCOMING_RETURN_ADDR_RTX
3175 A C expression whose value is RTL representing the location of the
3176 incoming return address at the beginning of any function, before the
3177 prologue. This RTL is either a @code{REG}, indicating that the return
3178 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3181 You only need to define this macro if you want to support call frame
3182 debugging information like that provided by DWARF 2.
3184 If this RTL is a @code{REG}, you should also define
3185 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3188 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3189 A C expression whose value is an integer giving a DWARF 2 column
3190 number that may be used as an alternative return column. The column
3191 must not correspond to any gcc hard register (that is, it must not
3192 be in the range of @code{DWARF_FRAME_REGNUM}).
3194 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3195 general register, but an alternative column needs to be used for signal
3196 frames. Some targets have also used different frame return columns
3200 @defmac DWARF_ZERO_REG
3201 A C expression whose value is an integer giving a DWARF 2 register
3202 number that is considered to always have the value zero. This should
3203 only be defined if the target has an architected zero register, and
3204 someone decided it was a good idea to use that register number to
3205 terminate the stack backtrace. New ports should avoid this.
3208 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3209 This target hook allows the backend to emit frame-related insns that
3210 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3211 info engine will invoke it on insns of the form
3213 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3217 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3219 to let the backend emit the call frame instructions. @var{label} is
3220 the CFI label attached to the insn, @var{pattern} is the pattern of
3221 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3224 @defmac INCOMING_FRAME_SP_OFFSET
3225 A C expression whose value is an integer giving the offset, in bytes,
3226 from the value of the stack pointer register to the top of the stack
3227 frame at the beginning of any function, before the prologue. The top of
3228 the frame is defined to be the value of the stack pointer in the
3229 previous frame, just before the call instruction.
3231 You only need to define this macro if you want to support call frame
3232 debugging information like that provided by DWARF 2.
3235 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3236 A C expression whose value is an integer giving the offset, in bytes,
3237 from the argument pointer to the canonical frame address (cfa). The
3238 final value should coincide with that calculated by
3239 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3240 during virtual register instantiation.
3242 The default value for this macro is
3243 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3244 which is correct for most machines; in general, the arguments are found
3245 immediately before the stack frame. Note that this is not the case on
3246 some targets that save registers into the caller's frame, such as SPARC
3247 and rs6000, and so such targets need to define this macro.
3249 You only need to define this macro if the default is incorrect, and you
3250 want to support call frame debugging information like that provided by
3254 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3255 If defined, a C expression whose value is an integer giving the offset
3256 in bytes from the frame pointer to the canonical frame address (cfa).
3257 The final value should coincide with that calculated by
3258 @code{INCOMING_FRAME_SP_OFFSET}.
3260 Normally the CFA is calculated as an offset from the argument pointer,
3261 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3262 variable due to the ABI, this may not be possible. If this macro is
3263 defined, it implies that the virtual register instantiation should be
3264 based on the frame pointer instead of the argument pointer. Only one
3265 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3269 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3270 If defined, a C expression whose value is an integer giving the offset
3271 in bytes from the canonical frame address (cfa) to the frame base used
3272 in DWARF 2 debug information. The default is zero. A different value
3273 may reduce the size of debug information on some ports.
3276 @node Exception Handling
3277 @subsection Exception Handling Support
3278 @cindex exception handling
3280 @defmac EH_RETURN_DATA_REGNO (@var{N})
3281 A C expression whose value is the @var{N}th register number used for
3282 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3283 @var{N} registers are usable.
3285 The exception handling library routines communicate with the exception
3286 handlers via a set of agreed upon registers. Ideally these registers
3287 should be call-clobbered; it is possible to use call-saved registers,
3288 but may negatively impact code size. The target must support at least
3289 2 data registers, but should define 4 if there are enough free registers.
3291 You must define this macro if you want to support call frame exception
3292 handling like that provided by DWARF 2.
3295 @defmac EH_RETURN_STACKADJ_RTX
3296 A C expression whose value is RTL representing a location in which
3297 to store a stack adjustment to be applied before function return.
3298 This is used to unwind the stack to an exception handler's call frame.
3299 It will be assigned zero on code paths that return normally.
3301 Typically this is a call-clobbered hard register that is otherwise
3302 untouched by the epilogue, but could also be a stack slot.
3304 Do not define this macro if the stack pointer is saved and restored
3305 by the regular prolog and epilog code in the call frame itself; in
3306 this case, the exception handling library routines will update the
3307 stack location to be restored in place. Otherwise, you must define
3308 this macro if you want to support call frame exception handling like
3309 that provided by DWARF 2.
3312 @defmac EH_RETURN_HANDLER_RTX
3313 A C expression whose value is RTL representing a location in which
3314 to store the address of an exception handler to which we should
3315 return. It will not be assigned on code paths that return normally.
3317 Typically this is the location in the call frame at which the normal
3318 return address is stored. For targets that return by popping an
3319 address off the stack, this might be a memory address just below
3320 the @emph{target} call frame rather than inside the current call
3321 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3322 been assigned, so it may be used to calculate the location of the
3325 Some targets have more complex requirements than storing to an
3326 address calculable during initial code generation. In that case
3327 the @code{eh_return} instruction pattern should be used instead.
3329 If you want to support call frame exception handling, you must
3330 define either this macro or the @code{eh_return} instruction pattern.
3333 @defmac RETURN_ADDR_OFFSET
3334 If defined, an integer-valued C expression for which rtl will be generated
3335 to add it to the exception handler address before it is searched in the
3336 exception handling tables, and to subtract it again from the address before
3337 using it to return to the exception handler.
3340 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3341 This macro chooses the encoding of pointers embedded in the exception
3342 handling sections. If at all possible, this should be defined such
3343 that the exception handling section will not require dynamic relocations,
3344 and so may be read-only.
3346 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3347 @var{global} is true if the symbol may be affected by dynamic relocations.
3348 The macro should return a combination of the @code{DW_EH_PE_*} defines
3349 as found in @file{dwarf2.h}.
3351 If this macro is not defined, pointers will not be encoded but
3352 represented directly.
3355 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3356 This macro allows the target to emit whatever special magic is required
3357 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3358 Generic code takes care of pc-relative and indirect encodings; this must
3359 be defined if the target uses text-relative or data-relative encodings.
3361 This is a C statement that branches to @var{done} if the format was
3362 handled. @var{encoding} is the format chosen, @var{size} is the number
3363 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3367 @defmac MD_UNWIND_SUPPORT
3368 A string specifying a file to be #include'd in unwind-dw2.c. The file
3369 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3372 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3373 This macro allows the target to add CPU and operating system specific
3374 code to the call-frame unwinder for use when there is no unwind data
3375 available. The most common reason to implement this macro is to unwind
3376 through signal frames.
3378 This macro is called from @code{uw_frame_state_for} in
3379 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3380 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3381 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3382 for the address of the code being executed and @code{context->cfa} for
3383 the stack pointer value. If the frame can be decoded, the register
3384 save addresses should be updated in @var{fs} and the macro should
3385 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3386 the macro should evaluate to @code{_URC_END_OF_STACK}.
3388 For proper signal handling in Java this macro is accompanied by
3389 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3392 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3393 This macro allows the target to add operating system specific code to the
3394 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3395 usually used for signal or interrupt frames.
3397 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3398 @var{context} is an @code{_Unwind_Context};
3399 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3400 for the abi and context in the @code{.unwabi} directive. If the
3401 @code{.unwabi} directive can be handled, the register save addresses should
3402 be updated in @var{fs}.
3405 @defmac TARGET_USES_WEAK_UNWIND_INFO
3406 A C expression that evaluates to true if the target requires unwind
3407 info to be given comdat linkage. Define it to be @code{1} if comdat
3408 linkage is necessary. The default is @code{0}.
3411 @node Stack Checking
3412 @subsection Specifying How Stack Checking is Done
3414 GCC will check that stack references are within the boundaries of the
3415 stack, if the option @option{-fstack-check} is specified, in one of
3420 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3421 will assume that you have arranged for full stack checking to be done
3422 at appropriate places in the configuration files. GCC will not do
3423 other special processing.
3426 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3427 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3428 that you have arranged for static stack checking (checking of the
3429 static stack frame of functions) to be done at appropriate places
3430 in the configuration files. GCC will only emit code to do dynamic
3431 stack checking (checking on dynamic stack allocations) using the third
3435 If neither of the above are true, GCC will generate code to periodically
3436 ``probe'' the stack pointer using the values of the macros defined below.
3439 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3440 GCC will change its allocation strategy for large objects if the option
3441 @option{-fstack-check} is specified: they will always be allocated
3442 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3444 @defmac STACK_CHECK_BUILTIN
3445 A nonzero value if stack checking is done by the configuration files in a
3446 machine-dependent manner. You should define this macro if stack checking
3447 is required by the ABI of your machine or if you would like to do stack
3448 checking in some more efficient way than the generic approach. The default
3449 value of this macro is zero.
3452 @defmac STACK_CHECK_STATIC_BUILTIN
3453 A nonzero value if static stack checking is done by the configuration files
3454 in a machine-dependent manner. You should define this macro if you would
3455 like to do static stack checking in some more efficient way than the generic
3456 approach. The default value of this macro is zero.
3459 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3460 An integer specifying the interval at which GCC must generate stack probe
3461 instructions, defined as 2 raised to this integer. You will normally
3462 define this macro so that the interval be no larger than the size of
3463 the ``guard pages'' at the end of a stack area. The default value
3464 of 12 (4096-byte interval) is suitable for most systems.
3467 @defmac STACK_CHECK_MOVING_SP
3468 An integer which is nonzero if GCC should move the stack pointer page by page
3469 when doing probes. This can be necessary on systems where the stack pointer
3470 contains the bottom address of the memory area accessible to the executing
3471 thread at any point in time. In this situation an alternate signal stack
3472 is required in order to be able to recover from a stack overflow. The
3473 default value of this macro is zero.
3476 @defmac STACK_CHECK_PROTECT
3477 The number of bytes of stack needed to recover from a stack overflow, for
3478 languages where such a recovery is supported. The default value of 75 words
3479 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3480 8192 bytes with other exception handling mechanisms should be adequate for
3484 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3485 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3486 in the opposite case.
3488 @defmac STACK_CHECK_MAX_FRAME_SIZE
3489 The maximum size of a stack frame, in bytes. GCC will generate probe
3490 instructions in non-leaf functions to ensure at least this many bytes of
3491 stack are available. If a stack frame is larger than this size, stack
3492 checking will not be reliable and GCC will issue a warning. The
3493 default is chosen so that GCC only generates one instruction on most
3494 systems. You should normally not change the default value of this macro.
3497 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3498 GCC uses this value to generate the above warning message. It
3499 represents the amount of fixed frame used by a function, not including
3500 space for any callee-saved registers, temporaries and user variables.
3501 You need only specify an upper bound for this amount and will normally
3502 use the default of four words.
3505 @defmac STACK_CHECK_MAX_VAR_SIZE
3506 The maximum size, in bytes, of an object that GCC will place in the
3507 fixed area of the stack frame when the user specifies
3508 @option{-fstack-check}.
3509 GCC computed the default from the values of the above macros and you will
3510 normally not need to override that default.
3514 @node Frame Registers
3515 @subsection Registers That Address the Stack Frame
3517 @c prevent bad page break with this line
3518 This discusses registers that address the stack frame.
3520 @defmac STACK_POINTER_REGNUM
3521 The register number of the stack pointer register, which must also be a
3522 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3523 the hardware determines which register this is.
3526 @defmac FRAME_POINTER_REGNUM
3527 The register number of the frame pointer register, which is used to
3528 access automatic variables in the stack frame. On some machines, the
3529 hardware determines which register this is. On other machines, you can
3530 choose any register you wish for this purpose.
3533 @defmac HARD_FRAME_POINTER_REGNUM
3534 On some machines the offset between the frame pointer and starting
3535 offset of the automatic variables is not known until after register
3536 allocation has been done (for example, because the saved registers are
3537 between these two locations). On those machines, define
3538 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3539 be used internally until the offset is known, and define
3540 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3541 used for the frame pointer.
3543 You should define this macro only in the very rare circumstances when it
3544 is not possible to calculate the offset between the frame pointer and
3545 the automatic variables until after register allocation has been
3546 completed. When this macro is defined, you must also indicate in your
3547 definition of @code{ELIMINABLE_REGS} how to eliminate
3548 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3549 or @code{STACK_POINTER_REGNUM}.
3551 Do not define this macro if it would be the same as
3552 @code{FRAME_POINTER_REGNUM}.
3555 @defmac ARG_POINTER_REGNUM
3556 The register number of the arg pointer register, which is used to access
3557 the function's argument list. On some machines, this is the same as the
3558 frame pointer register. On some machines, the hardware determines which
3559 register this is. On other machines, you can choose any register you
3560 wish for this purpose. If this is not the same register as the frame
3561 pointer register, then you must mark it as a fixed register according to
3562 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3563 (@pxref{Elimination}).
3566 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3567 Define this to a preprocessor constant that is nonzero if
3568 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3569 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3570 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3571 definition is not suitable for use in preprocessor conditionals.
3574 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3575 Define this to a preprocessor constant that is nonzero if
3576 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3577 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3578 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3579 definition is not suitable for use in preprocessor conditionals.
3582 @defmac RETURN_ADDRESS_POINTER_REGNUM
3583 The register number of the return address pointer register, which is used to
3584 access the current function's return address from the stack. On some
3585 machines, the return address is not at a fixed offset from the frame
3586 pointer or stack pointer or argument pointer. This register can be defined
3587 to point to the return address on the stack, and then be converted by
3588 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3590 Do not define this macro unless there is no other way to get the return
3591 address from the stack.
3594 @defmac STATIC_CHAIN_REGNUM
3595 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3596 Register numbers used for passing a function's static chain pointer. If
3597 register windows are used, the register number as seen by the called
3598 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3599 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3600 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3603 The static chain register need not be a fixed register.
3605 If the static chain is passed in memory, these macros should not be
3606 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3609 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl}, bool @var{incoming_p})
3610 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3611 targets that may use different static chain locations for different
3612 nested functions. This may be required if the target has function
3613 attributes that affect the calling conventions of the function and
3614 those calling conventions use different static chain locations.
3616 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3618 If the static chain is passed in memory, this hook should be used to
3619 provide rtx giving @code{mem} expressions that denote where they are stored.
3620 Often the @code{mem} expression as seen by the caller will be at an offset
3621 from the stack pointer and the @code{mem} expression as seen by the callee
3622 will be at an offset from the frame pointer.
3623 @findex stack_pointer_rtx
3624 @findex frame_pointer_rtx
3625 @findex arg_pointer_rtx
3626 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3627 @code{arg_pointer_rtx} will have been initialized and should be used
3628 to refer to those items.
3631 @defmac DWARF_FRAME_REGISTERS
3632 This macro specifies the maximum number of hard registers that can be
3633 saved in a call frame. This is used to size data structures used in
3634 DWARF2 exception handling.
3636 Prior to GCC 3.0, this macro was needed in order to establish a stable
3637 exception handling ABI in the face of adding new hard registers for ISA
3638 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3639 in the number of hard registers. Nevertheless, this macro can still be
3640 used to reduce the runtime memory requirements of the exception handling
3641 routines, which can be substantial if the ISA contains a lot of
3642 registers that are not call-saved.
3644 If this macro is not defined, it defaults to
3645 @code{FIRST_PSEUDO_REGISTER}.
3648 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3650 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3651 for backward compatibility in pre GCC 3.0 compiled code.
3653 If this macro is not defined, it defaults to
3654 @code{DWARF_FRAME_REGISTERS}.
3657 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3659 Define this macro if the target's representation for dwarf registers
3660 is different than the internal representation for unwind column.
3661 Given a dwarf register, this macro should return the internal unwind
3662 column number to use instead.
3664 See the PowerPC's SPE target for an example.
3667 @defmac DWARF_FRAME_REGNUM (@var{regno})
3669 Define this macro if the target's representation for dwarf registers
3670 used in .eh_frame or .debug_frame is different from that used in other
3671 debug info sections. Given a GCC hard register number, this macro
3672 should return the .eh_frame register number. The default is
3673 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3677 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3679 Define this macro to map register numbers held in the call frame info
3680 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3681 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3682 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3683 return @code{@var{regno}}.
3688 @subsection Eliminating Frame Pointer and Arg Pointer
3690 @c prevent bad page break with this line
3691 This is about eliminating the frame pointer and arg pointer.
3693 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3694 This target hook should return @code{true} if a function must have and use
3695 a frame pointer. This target hook is called in the reload pass. If its return
3696 value is @code{true} the function will have a frame pointer.
3698 This target hook can in principle examine the current function and decide
3699 according to the facts, but on most machines the constant @code{false} or the
3700 constant @code{true} suffices. Use @code{false} when the machine allows code
3701 to be generated with no frame pointer, and doing so saves some time or space.
3702 Use @code{true} when there is no possible advantage to avoiding a frame
3705 In certain cases, the compiler does not know how to produce valid code
3706 without a frame pointer. The compiler recognizes those cases and
3707 automatically gives the function a frame pointer regardless of what
3708 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3711 In a function that does not require a frame pointer, the frame pointer
3712 register can be allocated for ordinary usage, unless you mark it as a
3713 fixed register. See @code{FIXED_REGISTERS} for more information.
3715 Default return value is @code{false}.
3718 @findex get_frame_size
3719 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3720 A C statement to store in the variable @var{depth-var} the difference
3721 between the frame pointer and the stack pointer values immediately after
3722 the function prologue. The value would be computed from information
3723 such as the result of @code{get_frame_size ()} and the tables of
3724 registers @code{regs_ever_live} and @code{call_used_regs}.
3726 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3727 need not be defined. Otherwise, it must be defined even if
3728 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3729 case, you may set @var{depth-var} to anything.
3732 @defmac ELIMINABLE_REGS
3733 If defined, this macro specifies a table of register pairs used to
3734 eliminate unneeded registers that point into the stack frame. If it is not
3735 defined, the only elimination attempted by the compiler is to replace
3736 references to the frame pointer with references to the stack pointer.
3738 The definition of this macro is a list of structure initializations, each
3739 of which specifies an original and replacement register.
3741 On some machines, the position of the argument pointer is not known until
3742 the compilation is completed. In such a case, a separate hard register
3743 must be used for the argument pointer. This register can be eliminated by
3744 replacing it with either the frame pointer or the argument pointer,
3745 depending on whether or not the frame pointer has been eliminated.
3747 In this case, you might specify:
3749 #define ELIMINABLE_REGS \
3750 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3751 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3752 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3755 Note that the elimination of the argument pointer with the stack pointer is
3756 specified first since that is the preferred elimination.
3759 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3760 This target hook should returns @code{true} if the compiler is allowed to
3761 try to replace register number @var{from_reg} with register number
3762 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3763 is defined, and will usually be @code{true}, since most of the cases
3764 preventing register elimination are things that the compiler already
3767 Default return value is @code{true}.
3770 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3771 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3772 specifies the initial difference between the specified pair of
3773 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3777 @node Stack Arguments
3778 @subsection Passing Function Arguments on the Stack
3779 @cindex arguments on stack
3780 @cindex stack arguments
3782 The macros in this section control how arguments are passed
3783 on the stack. See the following section for other macros that
3784 control passing certain arguments in registers.
3786 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3787 This target hook returns @code{true} if an argument declared in a
3788 prototype as an integral type smaller than @code{int} should actually be
3789 passed as an @code{int}. In addition to avoiding errors in certain
3790 cases of mismatch, it also makes for better code on certain machines.
3791 The default is to not promote prototypes.
3795 A C expression. If nonzero, push insns will be used to pass
3797 If the target machine does not have a push instruction, set it to zero.
3798 That directs GCC to use an alternate strategy: to
3799 allocate the entire argument block and then store the arguments into
3800 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3803 @defmac PUSH_ARGS_REVERSED
3804 A C expression. If nonzero, function arguments will be evaluated from
3805 last to first, rather than from first to last. If this macro is not
3806 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3807 and args grow in opposite directions, and 0 otherwise.
3810 @defmac PUSH_ROUNDING (@var{npushed})
3811 A C expression that is the number of bytes actually pushed onto the
3812 stack when an instruction attempts to push @var{npushed} bytes.
3814 On some machines, the definition
3817 #define PUSH_ROUNDING(BYTES) (BYTES)
3821 will suffice. But on other machines, instructions that appear
3822 to push one byte actually push two bytes in an attempt to maintain
3823 alignment. Then the definition should be
3826 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3829 If the value of this macro has a type, it should be an unsigned type.
3832 @findex current_function_outgoing_args_size
3833 @defmac ACCUMULATE_OUTGOING_ARGS
3834 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3835 will be computed and placed into the variable
3836 @code{current_function_outgoing_args_size}. No space will be pushed
3837 onto the stack for each call; instead, the function prologue should
3838 increase the stack frame size by this amount.
3840 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3844 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3845 Define this macro if functions should assume that stack space has been
3846 allocated for arguments even when their values are passed in
3849 The value of this macro is the size, in bytes, of the area reserved for
3850 arguments passed in registers for the function represented by @var{fndecl},
3851 which can be zero if GCC is calling a library function.
3852 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3855 This space can be allocated by the caller, or be a part of the
3856 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3859 @c above is overfull. not sure what to do. --mew 5feb93 did
3860 @c something, not sure if it looks good. --mew 10feb93
3862 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3863 Define this to a nonzero value if it is the responsibility of the
3864 caller to allocate the area reserved for arguments passed in registers
3865 when calling a function of @var{fntype}. @var{fntype} may be NULL
3866 if the function called is a library function.
3868 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3869 whether the space for these arguments counts in the value of
3870 @code{current_function_outgoing_args_size}.
3873 @defmac STACK_PARMS_IN_REG_PARM_AREA
3874 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3875 stack parameters don't skip the area specified by it.
3876 @c i changed this, makes more sens and it should have taken care of the
3877 @c overfull.. not as specific, tho. --mew 5feb93
3879 Normally, when a parameter is not passed in registers, it is placed on the
3880 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3881 suppresses this behavior and causes the parameter to be passed on the
3882 stack in its natural location.
3885 @deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size})
3886 This target hook returns the number of bytes of its own arguments that
3887 a function pops on returning, or 0 if the function pops no arguments
3888 and the caller must therefore pop them all after the function returns.
3890 @var{fundecl} is a C variable whose value is a tree node that describes
3891 the function in question. Normally it is a node of type
3892 @code{FUNCTION_DECL} that describes the declaration of the function.
3893 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3895 @var{funtype} is a C variable whose value is a tree node that
3896 describes the function in question. Normally it is a node of type
3897 @code{FUNCTION_TYPE} that describes the data type of the function.
3898 From this it is possible to obtain the data types of the value and
3899 arguments (if known).
3901 When a call to a library function is being considered, @var{fundecl}
3902 will contain an identifier node for the library function. Thus, if
3903 you need to distinguish among various library functions, you can do so
3904 by their names. Note that ``library function'' in this context means
3905 a function used to perform arithmetic, whose name is known specially
3906 in the compiler and was not mentioned in the C code being compiled.
3908 @var{size} is the number of bytes of arguments passed on the
3909 stack. If a variable number of bytes is passed, it is zero, and
3910 argument popping will always be the responsibility of the calling function.
3912 On the VAX, all functions always pop their arguments, so the definition
3913 of this macro is @var{size}. On the 68000, using the standard
3914 calling convention, no functions pop their arguments, so the value of
3915 the macro is always 0 in this case. But an alternative calling
3916 convention is available in which functions that take a fixed number of
3917 arguments pop them but other functions (such as @code{printf}) pop
3918 nothing (the caller pops all). When this convention is in use,
3919 @var{funtype} is examined to determine whether a function takes a fixed
3920 number of arguments.
3923 @defmac CALL_POPS_ARGS (@var{cum})
3924 A C expression that should indicate the number of bytes a call sequence
3925 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3926 when compiling a function call.
3928 @var{cum} is the variable in which all arguments to the called function
3929 have been accumulated.
3931 On certain architectures, such as the SH5, a call trampoline is used
3932 that pops certain registers off the stack, depending on the arguments
3933 that have been passed to the function. Since this is a property of the
3934 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3938 @node Register Arguments
3939 @subsection Passing Arguments in Registers
3940 @cindex arguments in registers
3941 @cindex registers arguments
3943 This section describes the macros which let you control how various
3944 types of arguments are passed in registers or how they are arranged in
3947 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (CUMULATIVE_ARGS *@var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3948 Return an RTX indicating whether a function argument is passed in a
3949 register and if so, which register.
3951 The arguments are @var{ca}, which summarizes all the previous
3952 arguments; @var{mode}, the machine mode of the argument; @var{type},
3953 the data type of the argument as a tree node or 0 if that is not known
3954 (which happens for C support library functions); and @var{named},
3955 which is @code{true} for an ordinary argument and @code{false} for
3956 nameless arguments that correspond to @samp{@dots{}} in the called
3957 function's prototype. @var{type} can be an incomplete type if a
3958 syntax error has previously occurred.
3960 The return value is usually either a @code{reg} RTX for the hard
3961 register in which to pass the argument, or zero to pass the argument
3964 The value of the expression can also be a @code{parallel} RTX@. This is
3965 used when an argument is passed in multiple locations. The mode of the
3966 @code{parallel} should be the mode of the entire argument. The
3967 @code{parallel} holds any number of @code{expr_list} pairs; each one
3968 describes where part of the argument is passed. In each
3969 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3970 register in which to pass this part of the argument, and the mode of the
3971 register RTX indicates how large this part of the argument is. The
3972 second operand of the @code{expr_list} is a @code{const_int} which gives
3973 the offset in bytes into the entire argument of where this part starts.
3974 As a special exception the first @code{expr_list} in the @code{parallel}
3975 RTX may have a first operand of zero. This indicates that the entire
3976 argument is also stored on the stack.
3978 The last time this hook is called, it is called with @code{MODE ==
3979 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3980 pattern as operands 2 and 3 respectively.
3982 @cindex @file{stdarg.h} and register arguments
3983 The usual way to make the ISO library @file{stdarg.h} work on a
3984 machine where some arguments are usually passed in registers, is to
3985 cause nameless arguments to be passed on the stack instead. This is
3986 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
3987 @var{named} is @code{false}.
3989 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
3990 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
3991 You may use the hook @code{targetm.calls.must_pass_in_stack}
3992 in the definition of this macro to determine if this argument is of a
3993 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3994 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
3995 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3996 defined, the argument will be computed in the stack and then loaded into
4000 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, const_tree @var{type})
4001 This target hook should return @code{true} if we should not pass @var{type}
4002 solely in registers. The file @file{expr.h} defines a
4003 definition that is usually appropriate, refer to @file{expr.h} for additional
4007 @deftypefn {Target Hook} rtx TARGET_FUNCTION_INCOMING_ARG (CUMULATIVE_ARGS *@var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4008 Define this hook if the target machine has ``register windows'', so
4009 that the register in which a function sees an arguments is not
4010 necessarily the same as the one in which the caller passed the
4013 For such machines, @code{TARGET_FUNCTION_ARG} computes the register in
4014 which the caller passes the value, and
4015 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4016 fashion to tell the function being called where the arguments will
4019 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4020 @code{TARGET_FUNCTION_ARG} serves both purposes.
4023 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4024 This target hook returns the number of bytes at the beginning of an
4025 argument that must be put in registers. The value must be zero for
4026 arguments that are passed entirely in registers or that are entirely
4027 pushed on the stack.
4029 On some machines, certain arguments must be passed partially in
4030 registers and partially in memory. On these machines, typically the
4031 first few words of arguments are passed in registers, and the rest
4032 on the stack. If a multi-word argument (a @code{double} or a
4033 structure) crosses that boundary, its first few words must be passed
4034 in registers and the rest must be pushed. This macro tells the
4035 compiler when this occurs, and how many bytes should go in registers.
4037 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4038 register to be used by the caller for this argument; likewise
4039 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4042 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4043 This target hook should return @code{true} if an argument at the
4044 position indicated by @var{cum} should be passed by reference. This
4045 predicate is queried after target independent reasons for being
4046 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4048 If the hook returns true, a copy of that argument is made in memory and a
4049 pointer to the argument is passed instead of the argument itself.
4050 The pointer is passed in whatever way is appropriate for passing a pointer
4054 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4055 The function argument described by the parameters to this hook is
4056 known to be passed by reference. The hook should return true if the
4057 function argument should be copied by the callee instead of copied
4060 For any argument for which the hook returns true, if it can be
4061 determined that the argument is not modified, then a copy need
4064 The default version of this hook always returns false.
4067 @defmac CUMULATIVE_ARGS
4068 A C type for declaring a variable that is used as the first argument
4069 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4070 target machines, the type @code{int} suffices and can hold the number
4071 of bytes of argument so far.
4073 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4074 arguments that have been passed on the stack. The compiler has other
4075 variables to keep track of that. For target machines on which all
4076 arguments are passed on the stack, there is no need to store anything in
4077 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4078 should not be empty, so use @code{int}.
4081 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4082 If defined, this macro is called before generating any code for a
4083 function, but after the @var{cfun} descriptor for the function has been
4084 created. The back end may use this macro to update @var{cfun} to
4085 reflect an ABI other than that which would normally be used by default.
4086 If the compiler is generating code for a compiler-generated function,
4087 @var{fndecl} may be @code{NULL}.
4090 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4091 A C statement (sans semicolon) for initializing the variable
4092 @var{cum} for the state at the beginning of the argument list. The
4093 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4094 is the tree node for the data type of the function which will receive
4095 the args, or 0 if the args are to a compiler support library function.
4096 For direct calls that are not libcalls, @var{fndecl} contain the
4097 declaration node of the function. @var{fndecl} is also set when
4098 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4099 being compiled. @var{n_named_args} is set to the number of named
4100 arguments, including a structure return address if it is passed as a
4101 parameter, when making a call. When processing incoming arguments,
4102 @var{n_named_args} is set to @minus{}1.
4104 When processing a call to a compiler support library function,
4105 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4106 contains the name of the function, as a string. @var{libname} is 0 when
4107 an ordinary C function call is being processed. Thus, each time this
4108 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4109 never both of them at once.
4112 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4113 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4114 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4115 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4116 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4117 0)} is used instead.
4120 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4121 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4122 finding the arguments for the function being compiled. If this macro is
4123 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4125 The value passed for @var{libname} is always 0, since library routines
4126 with special calling conventions are never compiled with GCC@. The
4127 argument @var{libname} exists for symmetry with
4128 @code{INIT_CUMULATIVE_ARGS}.
4129 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4130 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4133 @deftypefn {Target Hook} void TARGET_FUNCTION_ARG_ADVANCE (CUMULATIVE_ARGS *@var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4134 This hook updates the summarizer variable pointed to by @var{ca} to
4135 advance past an argument in the argument list. The values @var{mode},
4136 @var{type} and @var{named} describe that argument. Once this is done,
4137 the variable @var{cum} is suitable for analyzing the @emph{following}
4138 argument with @code{TARGET_FUNCTION_ARG}, etc.
4140 This hook need not do anything if the argument in question was passed
4141 on the stack. The compiler knows how to track the amount of stack space
4142 used for arguments without any special help.
4145 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4146 If defined, a C expression that is the number of bytes to add to the
4147 offset of the argument passed in memory. This is needed for the SPU,
4148 which passes @code{char} and @code{short} arguments in the preferred
4149 slot that is in the middle of the quad word instead of starting at the
4153 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4154 If defined, a C expression which determines whether, and in which direction,
4155 to pad out an argument with extra space. The value should be of type
4156 @code{enum direction}: either @code{upward} to pad above the argument,
4157 @code{downward} to pad below, or @code{none} to inhibit padding.
4159 The @emph{amount} of padding is always just enough to reach the next
4160 multiple of @code{TARGET_FUNCTION_ARG_BOUNDARY}; this macro does not
4163 This macro has a default definition which is right for most systems.
4164 For little-endian machines, the default is to pad upward. For
4165 big-endian machines, the default is to pad downward for an argument of
4166 constant size shorter than an @code{int}, and upward otherwise.
4169 @defmac PAD_VARARGS_DOWN
4170 If defined, a C expression which determines whether the default
4171 implementation of va_arg will attempt to pad down before reading the
4172 next argument, if that argument is smaller than its aligned space as
4173 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4174 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4177 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4178 Specify padding for the last element of a block move between registers and
4179 memory. @var{first} is nonzero if this is the only element. Defining this
4180 macro allows better control of register function parameters on big-endian
4181 machines, without using @code{PARALLEL} rtl. In particular,
4182 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4183 registers, as there is no longer a "wrong" part of a register; For example,
4184 a three byte aggregate may be passed in the high part of a register if so
4188 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4189 This hook returns the alignment boundary, in bits, of an argument
4190 with the specified mode and type. The default hook returns
4191 @code{PARM_BOUNDARY} for all arguments.
4194 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4195 A C expression that is nonzero if @var{regno} is the number of a hard
4196 register in which function arguments are sometimes passed. This does
4197 @emph{not} include implicit arguments such as the static chain and
4198 the structure-value address. On many machines, no registers can be
4199 used for this purpose since all function arguments are pushed on the
4203 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4204 This hook should return true if parameter of type @var{type} are passed
4205 as two scalar parameters. By default, GCC will attempt to pack complex
4206 arguments into the target's word size. Some ABIs require complex arguments
4207 to be split and treated as their individual components. For example, on
4208 AIX64, complex floats should be passed in a pair of floating point
4209 registers, even though a complex float would fit in one 64-bit floating
4212 The default value of this hook is @code{NULL}, which is treated as always
4216 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4217 This hook returns a type node for @code{va_list} for the target.
4218 The default version of the hook returns @code{void*}.
4221 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4222 This target hook is used in function @code{c_common_nodes_and_builtins}
4223 to iterate through the target specific builtin types for va_list. The
4224 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4225 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4227 The arguments @var{pname} and @var{ptree} are used to store the result of
4228 this macro and are set to the name of the va_list builtin type and its
4230 If the return value of this macro is zero, then there is no more element.
4231 Otherwise the @var{IDX} should be increased for the next call of this
4232 macro to iterate through all types.
4235 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4236 This hook returns the va_list type of the calling convention specified by
4238 The default version of this hook returns @code{va_list_type_node}.
4241 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4242 This hook returns the va_list type of the calling convention specified by the
4243 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4247 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, gimple_seq *@var{pre_p}, gimple_seq *@var{post_p})
4248 This hook performs target-specific gimplification of
4249 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4250 arguments to @code{va_arg}; the latter two are as in
4251 @code{gimplify.c:gimplify_expr}.
4254 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4255 Define this to return nonzero if the port can handle pointers
4256 with machine mode @var{mode}. The default version of this
4257 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4260 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (struct ao_ref_s *@var{ref})
4261 Define this to return nonzero if the memory reference @var{ref} may alias with the system C library errno location. The default version of this hook assumes the system C library errno location is either a declaration of type int or accessed by dereferencing a pointer to int.
4264 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4265 Define this to return nonzero if the port is prepared to handle
4266 insns involving scalar mode @var{mode}. For a scalar mode to be
4267 considered supported, all the basic arithmetic and comparisons
4270 The default version of this hook returns true for any mode
4271 required to handle the basic C types (as defined by the port).
4272 Included here are the double-word arithmetic supported by the
4273 code in @file{optabs.c}.
4276 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4277 Define this to return nonzero if the port is prepared to handle
4278 insns involving vector mode @var{mode}. At the very least, it
4279 must have move patterns for this mode.
4282 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (enum machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4283 Return true if GCC should try to use a scalar mode to store an array
4284 of @var{nelems} elements, given that each element has mode @var{mode}.
4285 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4286 and allows GCC to use any defined integer mode.
4288 One use of this hook is to support vector load and store operations
4289 that operate on several homogeneous vectors. For example, ARM NEON
4290 has operations like:
4293 int8x8x3_t vld3_s8 (const int8_t *)
4296 where the return type is defined as:
4299 typedef struct int8x8x3_t
4305 If this hook allows @code{val} to have a scalar mode, then
4306 @code{int8x8x3_t} can have the same mode. GCC can then store
4307 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4310 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (enum machine_mode @var{mode})
4311 Define this to return nonzero for machine modes for which the port has
4312 small register classes. If this target hook returns nonzero for a given
4313 @var{mode}, the compiler will try to minimize the lifetime of registers
4314 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4315 In this case, the hook is expected to return nonzero if it returns nonzero
4318 On some machines, it is risky to let hard registers live across arbitrary
4319 insns. Typically, these machines have instructions that require values
4320 to be in specific registers (like an accumulator), and reload will fail
4321 if the required hard register is used for another purpose across such an
4324 Passes before reload do not know which hard registers will be used
4325 in an instruction, but the machine modes of the registers set or used in
4326 the instruction are already known. And for some machines, register
4327 classes are small for, say, integer registers but not for floating point
4328 registers. For example, the AMD x86-64 architecture requires specific
4329 registers for the legacy x86 integer instructions, but there are many
4330 SSE registers for floating point operations. On such targets, a good
4331 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4332 machine modes but zero for the SSE register classes.
4334 The default version of this hook returns false for any mode. It is always
4335 safe to redefine this hook to return with a nonzero value. But if you
4336 unnecessarily define it, you will reduce the amount of optimizations
4337 that can be performed in some cases. If you do not define this hook
4338 to return a nonzero value when it is required, the compiler will run out
4339 of spill registers and print a fatal error message.
4342 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
4343 If the target has a dedicated flags register, and it needs to use the post-reload comparison elimination pass, then this value should be set appropriately.
4347 @subsection How Scalar Function Values Are Returned
4348 @cindex return values in registers
4349 @cindex values, returned by functions
4350 @cindex scalars, returned as values
4352 This section discusses the macros that control returning scalars as
4353 values---values that can fit in registers.
4355 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4357 Define this to return an RTX representing the place where a function
4358 returns or receives a value of data type @var{ret_type}, a tree node
4359 representing a data type. @var{fn_decl_or_type} is a tree node
4360 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4361 function being called. If @var{outgoing} is false, the hook should
4362 compute the register in which the caller will see the return value.
4363 Otherwise, the hook should return an RTX representing the place where
4364 a function returns a value.
4366 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4367 (Actually, on most machines, scalar values are returned in the same
4368 place regardless of mode.) The value of the expression is usually a
4369 @code{reg} RTX for the hard register where the return value is stored.
4370 The value can also be a @code{parallel} RTX, if the return value is in
4371 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4372 @code{parallel} form. Note that the callee will populate every
4373 location specified in the @code{parallel}, but if the first element of
4374 the @code{parallel} contains the whole return value, callers will use
4375 that element as the canonical location and ignore the others. The m68k
4376 port uses this type of @code{parallel} to return pointers in both
4377 @samp{%a0} (the canonical location) and @samp{%d0}.
4379 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4380 the same promotion rules specified in @code{PROMOTE_MODE} if
4381 @var{valtype} is a scalar type.
4383 If the precise function being called is known, @var{func} is a tree
4384 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4385 pointer. This makes it possible to use a different value-returning
4386 convention for specific functions when all their calls are
4389 Some target machines have ``register windows'' so that the register in
4390 which a function returns its value is not the same as the one in which
4391 the caller sees the value. For such machines, you should return
4392 different RTX depending on @var{outgoing}.
4394 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4395 aggregate data types, because these are returned in another way. See
4396 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4399 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4400 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4401 a new target instead.
4404 @defmac LIBCALL_VALUE (@var{mode})
4405 A C expression to create an RTX representing the place where a library
4406 function returns a value of mode @var{mode}.
4408 Note that ``library function'' in this context means a compiler
4409 support routine, used to perform arithmetic, whose name is known
4410 specially by the compiler and was not mentioned in the C code being
4414 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (enum machine_mode @var{mode}, const_rtx @var{fun})
4415 Define this hook if the back-end needs to know the name of the libcall
4416 function in order to determine where the result should be returned.
4418 The mode of the result is given by @var{mode} and the name of the called
4419 library function is given by @var{fun}. The hook should return an RTX
4420 representing the place where the library function result will be returned.
4422 If this hook is not defined, then LIBCALL_VALUE will be used.
4425 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4426 A C expression that is nonzero if @var{regno} is the number of a hard
4427 register in which the values of called function may come back.
4429 A register whose use for returning values is limited to serving as the
4430 second of a pair (for a value of type @code{double}, say) need not be
4431 recognized by this macro. So for most machines, this definition
4435 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4438 If the machine has register windows, so that the caller and the called
4439 function use different registers for the return value, this macro
4440 should recognize only the caller's register numbers.
4442 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4443 for a new target instead.
4446 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4447 A target hook that return @code{true} if @var{regno} is the number of a hard
4448 register in which the values of called function may come back.
4450 A register whose use for returning values is limited to serving as the
4451 second of a pair (for a value of type @code{double}, say) need not be
4452 recognized by this target hook.
4454 If the machine has register windows, so that the caller and the called
4455 function use different registers for the return value, this target hook
4456 should recognize only the caller's register numbers.
4458 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4461 @defmac APPLY_RESULT_SIZE
4462 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4463 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4464 saving and restoring an arbitrary return value.
4467 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4468 This hook should return true if values of type @var{type} are returned
4469 at the most significant end of a register (in other words, if they are
4470 padded at the least significant end). You can assume that @var{type}
4471 is returned in a register; the caller is required to check this.
4473 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4474 be able to hold the complete return value. For example, if a 1-, 2-
4475 or 3-byte structure is returned at the most significant end of a
4476 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4480 @node Aggregate Return
4481 @subsection How Large Values Are Returned
4482 @cindex aggregates as return values
4483 @cindex large return values
4484 @cindex returning aggregate values
4485 @cindex structure value address
4487 When a function value's mode is @code{BLKmode} (and in some other
4488 cases), the value is not returned according to
4489 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4490 caller passes the address of a block of memory in which the value
4491 should be stored. This address is called the @dfn{structure value
4494 This section describes how to control returning structure values in
4497 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4498 This target hook should return a nonzero value to say to return the
4499 function value in memory, just as large structures are always returned.
4500 Here @var{type} will be the data type of the value, and @var{fntype}
4501 will be the type of the function doing the returning, or @code{NULL} for
4504 Note that values of mode @code{BLKmode} must be explicitly handled
4505 by this function. Also, the option @option{-fpcc-struct-return}
4506 takes effect regardless of this macro. On most systems, it is
4507 possible to leave the hook undefined; this causes a default
4508 definition to be used, whose value is the constant 1 for @code{BLKmode}
4509 values, and 0 otherwise.
4511 Do not use this hook to indicate that structures and unions should always
4512 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4516 @defmac DEFAULT_PCC_STRUCT_RETURN
4517 Define this macro to be 1 if all structure and union return values must be
4518 in memory. Since this results in slower code, this should be defined
4519 only if needed for compatibility with other compilers or with an ABI@.
4520 If you define this macro to be 0, then the conventions used for structure
4521 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4524 If not defined, this defaults to the value 1.
4527 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4528 This target hook should return the location of the structure value
4529 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4530 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4531 be @code{NULL}, for libcalls. You do not need to define this target
4532 hook if the address is always passed as an ``invisible'' first
4535 On some architectures the place where the structure value address
4536 is found by the called function is not the same place that the
4537 caller put it. This can be due to register windows, or it could
4538 be because the function prologue moves it to a different place.
4539 @var{incoming} is @code{1} or @code{2} when the location is needed in
4540 the context of the called function, and @code{0} in the context of
4543 If @var{incoming} is nonzero and the address is to be found on the
4544 stack, return a @code{mem} which refers to the frame pointer. If
4545 @var{incoming} is @code{2}, the result is being used to fetch the
4546 structure value address at the beginning of a function. If you need
4547 to emit adjusting code, you should do it at this point.
4550 @defmac PCC_STATIC_STRUCT_RETURN
4551 Define this macro if the usual system convention on the target machine
4552 for returning structures and unions is for the called function to return
4553 the address of a static variable containing the value.
4555 Do not define this if the usual system convention is for the caller to
4556 pass an address to the subroutine.
4558 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4559 nothing when you use @option{-freg-struct-return} mode.
4562 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4563 This target hook returns the mode to be used when accessing raw return registers in @code{__builtin_return}. Define this macro if the value in @var{reg_raw_mode} is not correct.
4566 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_ARG_MODE (int @var{regno})
4567 This target hook returns the mode to be used when accessing raw argument registers in @code{__builtin_apply_args}. Define this macro if the value in @var{reg_raw_mode} is not correct.
4571 @subsection Caller-Saves Register Allocation
4573 If you enable it, GCC can save registers around function calls. This
4574 makes it possible to use call-clobbered registers to hold variables that
4575 must live across calls.
4577 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4578 A C expression to determine whether it is worthwhile to consider placing
4579 a pseudo-register in a call-clobbered hard register and saving and
4580 restoring it around each function call. The expression should be 1 when
4581 this is worth doing, and 0 otherwise.
4583 If you don't define this macro, a default is used which is good on most
4584 machines: @code{4 * @var{calls} < @var{refs}}.
4587 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4588 A C expression specifying which mode is required for saving @var{nregs}
4589 of a pseudo-register in call-clobbered hard register @var{regno}. If
4590 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4591 returned. For most machines this macro need not be defined since GCC
4592 will select the smallest suitable mode.
4595 @node Function Entry
4596 @subsection Function Entry and Exit
4597 @cindex function entry and exit
4601 This section describes the macros that output function entry
4602 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4604 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4605 If defined, a function that outputs the assembler code for entry to a
4606 function. The prologue is responsible for setting up the stack frame,
4607 initializing the frame pointer register, saving registers that must be
4608 saved, and allocating @var{size} additional bytes of storage for the
4609 local variables. @var{size} is an integer. @var{file} is a stdio
4610 stream to which the assembler code should be output.
4612 The label for the beginning of the function need not be output by this
4613 macro. That has already been done when the macro is run.
4615 @findex regs_ever_live
4616 To determine which registers to save, the macro can refer to the array
4617 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4618 @var{r} is used anywhere within the function. This implies the function
4619 prologue should save register @var{r}, provided it is not one of the
4620 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4621 @code{regs_ever_live}.)
4623 On machines that have ``register windows'', the function entry code does
4624 not save on the stack the registers that are in the windows, even if
4625 they are supposed to be preserved by function calls; instead it takes
4626 appropriate steps to ``push'' the register stack, if any non-call-used
4627 registers are used in the function.
4629 @findex frame_pointer_needed
4630 On machines where functions may or may not have frame-pointers, the
4631 function entry code must vary accordingly; it must set up the frame
4632 pointer if one is wanted, and not otherwise. To determine whether a
4633 frame pointer is in wanted, the macro can refer to the variable
4634 @code{frame_pointer_needed}. The variable's value will be 1 at run
4635 time in a function that needs a frame pointer. @xref{Elimination}.
4637 The function entry code is responsible for allocating any stack space
4638 required for the function. This stack space consists of the regions
4639 listed below. In most cases, these regions are allocated in the
4640 order listed, with the last listed region closest to the top of the
4641 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4642 the highest address if it is not defined). You can use a different order
4643 for a machine if doing so is more convenient or required for
4644 compatibility reasons. Except in cases where required by standard
4645 or by a debugger, there is no reason why the stack layout used by GCC
4646 need agree with that used by other compilers for a machine.
4649 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4650 If defined, a function that outputs assembler code at the end of a
4651 prologue. This should be used when the function prologue is being
4652 emitted as RTL, and you have some extra assembler that needs to be
4653 emitted. @xref{prologue instruction pattern}.
4656 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4657 If defined, a function that outputs assembler code at the start of an
4658 epilogue. This should be used when the function epilogue is being
4659 emitted as RTL, and you have some extra assembler that needs to be
4660 emitted. @xref{epilogue instruction pattern}.
4663 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4664 If defined, a function that outputs the assembler code for exit from a
4665 function. The epilogue is responsible for restoring the saved
4666 registers and stack pointer to their values when the function was
4667 called, and returning control to the caller. This macro takes the
4668 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4669 registers to restore are determined from @code{regs_ever_live} and
4670 @code{CALL_USED_REGISTERS} in the same way.
4672 On some machines, there is a single instruction that does all the work
4673 of returning from the function. On these machines, give that
4674 instruction the name @samp{return} and do not define the macro
4675 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4677 Do not define a pattern named @samp{return} if you want the
4678 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4679 switches to control whether return instructions or epilogues are used,
4680 define a @samp{return} pattern with a validity condition that tests the
4681 target switches appropriately. If the @samp{return} pattern's validity
4682 condition is false, epilogues will be used.
4684 On machines where functions may or may not have frame-pointers, the
4685 function exit code must vary accordingly. Sometimes the code for these
4686 two cases is completely different. To determine whether a frame pointer
4687 is wanted, the macro can refer to the variable
4688 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4689 a function that needs a frame pointer.
4691 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4692 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4693 The C variable @code{current_function_is_leaf} is nonzero for such a
4694 function. @xref{Leaf Functions}.
4696 On some machines, some functions pop their arguments on exit while
4697 others leave that for the caller to do. For example, the 68020 when
4698 given @option{-mrtd} pops arguments in functions that take a fixed
4699 number of arguments.
4701 @findex current_function_pops_args
4702 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4703 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4704 needs to know what was decided. The number of bytes of the current
4705 function's arguments that this function should pop is available in
4706 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4711 @findex current_function_pretend_args_size
4712 A region of @code{current_function_pretend_args_size} bytes of
4713 uninitialized space just underneath the first argument arriving on the
4714 stack. (This may not be at the very start of the allocated stack region
4715 if the calling sequence has pushed anything else since pushing the stack
4716 arguments. But usually, on such machines, nothing else has been pushed
4717 yet, because the function prologue itself does all the pushing.) This
4718 region is used on machines where an argument may be passed partly in
4719 registers and partly in memory, and, in some cases to support the
4720 features in @code{<stdarg.h>}.
4723 An area of memory used to save certain registers used by the function.
4724 The size of this area, which may also include space for such things as
4725 the return address and pointers to previous stack frames, is
4726 machine-specific and usually depends on which registers have been used
4727 in the function. Machines with register windows often do not require
4731 A region of at least @var{size} bytes, possibly rounded up to an allocation
4732 boundary, to contain the local variables of the function. On some machines,
4733 this region and the save area may occur in the opposite order, with the
4734 save area closer to the top of the stack.
4737 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4738 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4739 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4740 argument lists of the function. @xref{Stack Arguments}.
4743 @defmac EXIT_IGNORE_STACK
4744 Define this macro as a C expression that is nonzero if the return
4745 instruction or the function epilogue ignores the value of the stack
4746 pointer; in other words, if it is safe to delete an instruction to
4747 adjust the stack pointer before a return from the function. The
4750 Note that this macro's value is relevant only for functions for which
4751 frame pointers are maintained. It is never safe to delete a final
4752 stack adjustment in a function that has no frame pointer, and the
4753 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4756 @defmac EPILOGUE_USES (@var{regno})
4757 Define this macro as a C expression that is nonzero for registers that are
4758 used by the epilogue or the @samp{return} pattern. The stack and frame
4759 pointer registers are already assumed to be used as needed.
4762 @defmac EH_USES (@var{regno})
4763 Define this macro as a C expression that is nonzero for registers that are
4764 used by the exception handling mechanism, and so should be considered live
4765 on entry to an exception edge.
4768 @defmac DELAY_SLOTS_FOR_EPILOGUE
4769 Define this macro if the function epilogue contains delay slots to which
4770 instructions from the rest of the function can be ``moved''. The
4771 definition should be a C expression whose value is an integer
4772 representing the number of delay slots there.
4775 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4776 A C expression that returns 1 if @var{insn} can be placed in delay
4777 slot number @var{n} of the epilogue.
4779 The argument @var{n} is an integer which identifies the delay slot now
4780 being considered (since different slots may have different rules of
4781 eligibility). It is never negative and is always less than the number
4782 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4783 If you reject a particular insn for a given delay slot, in principle, it
4784 may be reconsidered for a subsequent delay slot. Also, other insns may
4785 (at least in principle) be considered for the so far unfilled delay
4788 @findex current_function_epilogue_delay_list
4789 @findex final_scan_insn
4790 The insns accepted to fill the epilogue delay slots are put in an RTL
4791 list made with @code{insn_list} objects, stored in the variable
4792 @code{current_function_epilogue_delay_list}. The insn for the first
4793 delay slot comes first in the list. Your definition of the macro
4794 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4795 outputting the insns in this list, usually by calling
4796 @code{final_scan_insn}.
4798 You need not define this macro if you did not define
4799 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4802 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function})
4803 A function that outputs the assembler code for a thunk
4804 function, used to implement C++ virtual function calls with multiple
4805 inheritance. The thunk acts as a wrapper around a virtual function,
4806 adjusting the implicit object parameter before handing control off to
4809 First, emit code to add the integer @var{delta} to the location that
4810 contains the incoming first argument. Assume that this argument
4811 contains a pointer, and is the one used to pass the @code{this} pointer
4812 in C++. This is the incoming argument @emph{before} the function prologue,
4813 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4814 all other incoming arguments.
4816 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4817 made after adding @code{delta}. In particular, if @var{p} is the
4818 adjusted pointer, the following adjustment should be made:
4821 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4824 After the additions, emit code to jump to @var{function}, which is a
4825 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4826 not touch the return address. Hence returning from @var{FUNCTION} will
4827 return to whoever called the current @samp{thunk}.
4829 The effect must be as if @var{function} had been called directly with
4830 the adjusted first argument. This macro is responsible for emitting all
4831 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4832 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4834 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4835 have already been extracted from it.) It might possibly be useful on
4836 some targets, but probably not.
4838 If you do not define this macro, the target-independent code in the C++
4839 front end will generate a less efficient heavyweight thunk that calls
4840 @var{function} instead of jumping to it. The generic approach does
4841 not support varargs.
4844 @deftypefn {Target Hook} bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (const_tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, const_tree @var{function})
4845 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4846 to output the assembler code for the thunk function specified by the
4847 arguments it is passed, and false otherwise. In the latter case, the
4848 generic approach will be used by the C++ front end, with the limitations
4853 @subsection Generating Code for Profiling
4854 @cindex profiling, code generation
4856 These macros will help you generate code for profiling.
4858 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4859 A C statement or compound statement to output to @var{file} some
4860 assembler code to call the profiling subroutine @code{mcount}.
4863 The details of how @code{mcount} expects to be called are determined by
4864 your operating system environment, not by GCC@. To figure them out,
4865 compile a small program for profiling using the system's installed C
4866 compiler and look at the assembler code that results.
4868 Older implementations of @code{mcount} expect the address of a counter
4869 variable to be loaded into some register. The name of this variable is
4870 @samp{LP} followed by the number @var{labelno}, so you would generate
4871 the name using @samp{LP%d} in a @code{fprintf}.
4874 @defmac PROFILE_HOOK
4875 A C statement or compound statement to output to @var{file} some assembly
4876 code to call the profiling subroutine @code{mcount} even the target does
4877 not support profiling.
4880 @defmac NO_PROFILE_COUNTERS
4881 Define this macro to be an expression with a nonzero value if the
4882 @code{mcount} subroutine on your system does not need a counter variable
4883 allocated for each function. This is true for almost all modern
4884 implementations. If you define this macro, you must not use the
4885 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4888 @defmac PROFILE_BEFORE_PROLOGUE
4889 Define this macro if the code for function profiling should come before
4890 the function prologue. Normally, the profiling code comes after.
4894 @subsection Permitting tail calls
4897 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4898 True if it is ok to do sibling call optimization for the specified
4899 call expression @var{exp}. @var{decl} will be the called function,
4900 or @code{NULL} if this is an indirect call.
4902 It is not uncommon for limitations of calling conventions to prevent
4903 tail calls to functions outside the current unit of translation, or
4904 during PIC compilation. The hook is used to enforce these restrictions,
4905 as the @code{sibcall} md pattern can not fail, or fall over to a
4906 ``normal'' call. The criteria for successful sibling call optimization
4907 may vary greatly between different architectures.
4910 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4911 Add any hard registers to @var{regs} that are live on entry to the
4912 function. This hook only needs to be defined to provide registers that
4913 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4914 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4915 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4916 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4919 @node Stack Smashing Protection
4920 @subsection Stack smashing protection
4921 @cindex stack smashing protection
4923 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4924 This hook returns a @code{DECL} node for the external variable to use
4925 for the stack protection guard. This variable is initialized by the
4926 runtime to some random value and is used to initialize the guard value
4927 that is placed at the top of the local stack frame. The type of this
4928 variable must be @code{ptr_type_node}.
4930 The default version of this hook creates a variable called
4931 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4934 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4935 This hook returns a tree expression that alerts the runtime that the
4936 stack protect guard variable has been modified. This expression should
4937 involve a call to a @code{noreturn} function.
4939 The default version of this hook invokes a function called
4940 @samp{__stack_chk_fail}, taking no arguments. This function is
4941 normally defined in @file{libgcc2.c}.
4944 @deftypefn {Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
4945 Whether this target supports splitting the stack when the options described in @var{opts} have been passed. This is called after options have been parsed, so the target may reject splitting the stack in some configurations. The default version of this hook returns false. If @var{report} is true, this function may issue a warning or error; if @var{report} is false, it must simply return a value
4949 @section Implementing the Varargs Macros
4950 @cindex varargs implementation
4952 GCC comes with an implementation of @code{<varargs.h>} and
4953 @code{<stdarg.h>} that work without change on machines that pass arguments
4954 on the stack. Other machines require their own implementations of
4955 varargs, and the two machine independent header files must have
4956 conditionals to include it.
4958 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4959 the calling convention for @code{va_start}. The traditional
4960 implementation takes just one argument, which is the variable in which
4961 to store the argument pointer. The ISO implementation of
4962 @code{va_start} takes an additional second argument. The user is
4963 supposed to write the last named argument of the function here.
4965 However, @code{va_start} should not use this argument. The way to find
4966 the end of the named arguments is with the built-in functions described
4969 @defmac __builtin_saveregs ()
4970 Use this built-in function to save the argument registers in memory so
4971 that the varargs mechanism can access them. Both ISO and traditional
4972 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4973 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4975 On some machines, @code{__builtin_saveregs} is open-coded under the
4976 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4977 other machines, it calls a routine written in assembler language,
4978 found in @file{libgcc2.c}.
4980 Code generated for the call to @code{__builtin_saveregs} appears at the
4981 beginning of the function, as opposed to where the call to
4982 @code{__builtin_saveregs} is written, regardless of what the code is.
4983 This is because the registers must be saved before the function starts
4984 to use them for its own purposes.
4985 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4989 @defmac __builtin_next_arg (@var{lastarg})
4990 This builtin returns the address of the first anonymous stack
4991 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4992 returns the address of the location above the first anonymous stack
4993 argument. Use it in @code{va_start} to initialize the pointer for
4994 fetching arguments from the stack. Also use it in @code{va_start} to
4995 verify that the second parameter @var{lastarg} is the last named argument
4996 of the current function.
4999 @defmac __builtin_classify_type (@var{object})
5000 Since each machine has its own conventions for which data types are
5001 passed in which kind of register, your implementation of @code{va_arg}
5002 has to embody these conventions. The easiest way to categorize the
5003 specified data type is to use @code{__builtin_classify_type} together
5004 with @code{sizeof} and @code{__alignof__}.
5006 @code{__builtin_classify_type} ignores the value of @var{object},
5007 considering only its data type. It returns an integer describing what
5008 kind of type that is---integer, floating, pointer, structure, and so on.
5010 The file @file{typeclass.h} defines an enumeration that you can use to
5011 interpret the values of @code{__builtin_classify_type}.
5014 These machine description macros help implement varargs:
5016 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5017 If defined, this hook produces the machine-specific code for a call to
5018 @code{__builtin_saveregs}. This code will be moved to the very
5019 beginning of the function, before any parameter access are made. The
5020 return value of this function should be an RTX that contains the value
5021 to use as the return of @code{__builtin_saveregs}.
5024 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (CUMULATIVE_ARGS *@var{args_so_far}, enum machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
5025 This target hook offers an alternative to using
5026 @code{__builtin_saveregs} and defining the hook
5027 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5028 register arguments into the stack so that all the arguments appear to
5029 have been passed consecutively on the stack. Once this is done, you can
5030 use the standard implementation of varargs that works for machines that
5031 pass all their arguments on the stack.
5033 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5034 structure, containing the values that are obtained after processing the
5035 named arguments. The arguments @var{mode} and @var{type} describe the
5036 last named argument---its machine mode and its data type as a tree node.
5038 The target hook should do two things: first, push onto the stack all the
5039 argument registers @emph{not} used for the named arguments, and second,
5040 store the size of the data thus pushed into the @code{int}-valued
5041 variable pointed to by @var{pretend_args_size}. The value that you
5042 store here will serve as additional offset for setting up the stack
5045 Because you must generate code to push the anonymous arguments at
5046 compile time without knowing their data types,
5047 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5048 have just a single category of argument register and use it uniformly
5051 If the argument @var{second_time} is nonzero, it means that the
5052 arguments of the function are being analyzed for the second time. This
5053 happens for an inline function, which is not actually compiled until the
5054 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5055 not generate any instructions in this case.
5058 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
5059 Define this hook to return @code{true} if the location where a function
5060 argument is passed depends on whether or not it is a named argument.
5062 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5063 is set for varargs and stdarg functions. If this hook returns
5064 @code{true}, the @var{named} argument is always true for named
5065 arguments, and false for unnamed arguments. If it returns @code{false},
5066 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5067 then all arguments are treated as named. Otherwise, all named arguments
5068 except the last are treated as named.
5070 You need not define this hook if it always returns @code{false}.
5073 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (CUMULATIVE_ARGS *@var{ca})
5074 If you need to conditionally change ABIs so that one works with
5075 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5076 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5077 defined, then define this hook to return @code{true} if
5078 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5079 Otherwise, you should not define this hook.
5083 @section Trampolines for Nested Functions
5084 @cindex trampolines for nested functions
5085 @cindex nested functions, trampolines for
5087 A @dfn{trampoline} is a small piece of code that is created at run time
5088 when the address of a nested function is taken. It normally resides on
5089 the stack, in the stack frame of the containing function. These macros
5090 tell GCC how to generate code to allocate and initialize a
5093 The instructions in the trampoline must do two things: load a constant
5094 address into the static chain register, and jump to the real address of
5095 the nested function. On CISC machines such as the m68k, this requires
5096 two instructions, a move immediate and a jump. Then the two addresses
5097 exist in the trampoline as word-long immediate operands. On RISC
5098 machines, it is often necessary to load each address into a register in
5099 two parts. Then pieces of each address form separate immediate
5102 The code generated to initialize the trampoline must store the variable
5103 parts---the static chain value and the function address---into the
5104 immediate operands of the instructions. On a CISC machine, this is
5105 simply a matter of copying each address to a memory reference at the
5106 proper offset from the start of the trampoline. On a RISC machine, it
5107 may be necessary to take out pieces of the address and store them
5110 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5111 This hook is called by @code{assemble_trampoline_template} to output,
5112 on the stream @var{f}, assembler code for a block of data that contains
5113 the constant parts of a trampoline. This code should not include a
5114 label---the label is taken care of automatically.
5116 If you do not define this hook, it means no template is needed
5117 for the target. Do not define this hook on systems where the block move
5118 code to copy the trampoline into place would be larger than the code
5119 to generate it on the spot.
5122 @defmac TRAMPOLINE_SECTION
5123 Return the section into which the trampoline template is to be placed
5124 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5127 @defmac TRAMPOLINE_SIZE
5128 A C expression for the size in bytes of the trampoline, as an integer.
5131 @defmac TRAMPOLINE_ALIGNMENT
5132 Alignment required for trampolines, in bits.
5134 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5135 is used for aligning trampolines.
5138 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5139 This hook is called to initialize a trampoline.
5140 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5141 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5142 RTX for the static chain value that should be passed to the function
5145 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5146 first thing this hook should do is emit a block move into @var{m_tramp}
5147 from the memory block returned by @code{assemble_trampoline_template}.
5148 Note that the block move need only cover the constant parts of the
5149 trampoline. If the target isolates the variable parts of the trampoline
5150 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5152 If the target requires any other actions, such as flushing caches or
5153 enabling stack execution, these actions should be performed after
5154 initializing the trampoline proper.
5157 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5158 This hook should perform any machine-specific adjustment in
5159 the address of the trampoline. Its argument contains the address of the
5160 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5161 the address to be used for a function call should be different from the
5162 address at which the template was stored, the different address should
5163 be returned; otherwise @var{addr} should be returned unchanged.
5164 If this hook is not defined, @var{addr} will be used for function calls.
5167 Implementing trampolines is difficult on many machines because they have
5168 separate instruction and data caches. Writing into a stack location
5169 fails to clear the memory in the instruction cache, so when the program
5170 jumps to that location, it executes the old contents.
5172 Here are two possible solutions. One is to clear the relevant parts of
5173 the instruction cache whenever a trampoline is set up. The other is to
5174 make all trampolines identical, by having them jump to a standard
5175 subroutine. The former technique makes trampoline execution faster; the
5176 latter makes initialization faster.
5178 To clear the instruction cache when a trampoline is initialized, define
5179 the following macro.
5181 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5182 If defined, expands to a C expression clearing the @emph{instruction
5183 cache} in the specified interval. The definition of this macro would
5184 typically be a series of @code{asm} statements. Both @var{beg} and
5185 @var{end} are both pointer expressions.
5188 The operating system may also require the stack to be made executable
5189 before calling the trampoline. To implement this requirement, define
5190 the following macro.
5192 @defmac ENABLE_EXECUTE_STACK
5193 Define this macro if certain operations must be performed before executing
5194 code located on the stack. The macro should expand to a series of C
5195 file-scope constructs (e.g.@: functions) and provide a unique entry point
5196 named @code{__enable_execute_stack}. The target is responsible for
5197 emitting calls to the entry point in the code, for example from the
5198 @code{TARGET_TRAMPOLINE_INIT} hook.
5201 To use a standard subroutine, define the following macro. In addition,
5202 you must make sure that the instructions in a trampoline fill an entire
5203 cache line with identical instructions, or else ensure that the
5204 beginning of the trampoline code is always aligned at the same point in
5205 its cache line. Look in @file{m68k.h} as a guide.
5207 @defmac TRANSFER_FROM_TRAMPOLINE
5208 Define this macro if trampolines need a special subroutine to do their
5209 work. The macro should expand to a series of @code{asm} statements
5210 which will be compiled with GCC@. They go in a library function named
5211 @code{__transfer_from_trampoline}.
5213 If you need to avoid executing the ordinary prologue code of a compiled
5214 C function when you jump to the subroutine, you can do so by placing a
5215 special label of your own in the assembler code. Use one @code{asm}
5216 statement to generate an assembler label, and another to make the label
5217 global. Then trampolines can use that label to jump directly to your
5218 special assembler code.
5222 @section Implicit Calls to Library Routines
5223 @cindex library subroutine names
5224 @cindex @file{libgcc.a}
5226 @c prevent bad page break with this line
5227 Here is an explanation of implicit calls to library routines.
5229 @defmac DECLARE_LIBRARY_RENAMES
5230 This macro, if defined, should expand to a piece of C code that will get
5231 expanded when compiling functions for libgcc.a. It can be used to
5232 provide alternate names for GCC's internal library functions if there
5233 are ABI-mandated names that the compiler should provide.
5236 @findex set_optab_libfunc
5237 @findex init_one_libfunc
5238 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5239 This hook should declare additional library routines or rename
5240 existing ones, using the functions @code{set_optab_libfunc} and
5241 @code{init_one_libfunc} defined in @file{optabs.c}.
5242 @code{init_optabs} calls this macro after initializing all the normal
5245 The default is to do nothing. Most ports don't need to define this hook.
5248 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5249 This macro should return @code{true} if the library routine that
5250 implements the floating point comparison operator @var{comparison} in
5251 mode @var{mode} will return a boolean, and @var{false} if it will
5254 GCC's own floating point libraries return tristates from the
5255 comparison operators, so the default returns false always. Most ports
5256 don't need to define this macro.
5259 @defmac TARGET_LIB_INT_CMP_BIASED
5260 This macro should evaluate to @code{true} if the integer comparison
5261 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5262 operand is smaller than the second, 1 to indicate that they are equal,
5263 and 2 to indicate that the first operand is greater than the second.
5264 If this macro evaluates to @code{false} the comparison functions return
5265 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5266 in @file{libgcc.a}, you do not need to define this macro.
5269 @cindex @code{EDOM}, implicit usage
5272 The value of @code{EDOM} on the target machine, as a C integer constant
5273 expression. If you don't define this macro, GCC does not attempt to
5274 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5275 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5278 If you do not define @code{TARGET_EDOM}, then compiled code reports
5279 domain errors by calling the library function and letting it report the
5280 error. If mathematical functions on your system use @code{matherr} when
5281 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5282 that @code{matherr} is used normally.
5285 @cindex @code{errno}, implicit usage
5286 @defmac GEN_ERRNO_RTX
5287 Define this macro as a C expression to create an rtl expression that
5288 refers to the global ``variable'' @code{errno}. (On certain systems,
5289 @code{errno} may not actually be a variable.) If you don't define this
5290 macro, a reasonable default is used.
5293 @cindex C99 math functions, implicit usage
5294 @defmac TARGET_C99_FUNCTIONS
5295 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5296 @code{sinf} and similarly for other functions defined by C99 standard. The
5297 default is zero because a number of existing systems lack support for these
5298 functions in their runtime so this macro needs to be redefined to one on
5299 systems that do support the C99 runtime.
5302 @cindex sincos math function, implicit usage
5303 @defmac TARGET_HAS_SINCOS
5304 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5305 and @code{cos} with the same argument to a call to @code{sincos}. The
5306 default is zero. The target has to provide the following functions:
5308 void sincos(double x, double *sin, double *cos);
5309 void sincosf(float x, float *sin, float *cos);
5310 void sincosl(long double x, long double *sin, long double *cos);
5314 @defmac NEXT_OBJC_RUNTIME
5315 Define this macro to generate code for Objective-C message sending using
5316 the calling convention of the NeXT system. This calling convention
5317 involves passing the object, the selector and the method arguments all
5318 at once to the method-lookup library function.
5320 The default calling convention passes just the object and the selector
5321 to the lookup function, which returns a pointer to the method.
5324 @node Addressing Modes
5325 @section Addressing Modes
5326 @cindex addressing modes
5328 @c prevent bad page break with this line
5329 This is about addressing modes.
5331 @defmac HAVE_PRE_INCREMENT
5332 @defmacx HAVE_PRE_DECREMENT
5333 @defmacx HAVE_POST_INCREMENT
5334 @defmacx HAVE_POST_DECREMENT
5335 A C expression that is nonzero if the machine supports pre-increment,
5336 pre-decrement, post-increment, or post-decrement addressing respectively.
5339 @defmac HAVE_PRE_MODIFY_DISP
5340 @defmacx HAVE_POST_MODIFY_DISP
5341 A C expression that is nonzero if the machine supports pre- or
5342 post-address side-effect generation involving constants other than
5343 the size of the memory operand.
5346 @defmac HAVE_PRE_MODIFY_REG
5347 @defmacx HAVE_POST_MODIFY_REG
5348 A C expression that is nonzero if the machine supports pre- or
5349 post-address side-effect generation involving a register displacement.
5352 @defmac CONSTANT_ADDRESS_P (@var{x})
5353 A C expression that is 1 if the RTX @var{x} is a constant which
5354 is a valid address. On most machines the default definition of
5355 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5356 is acceptable, but a few machines are more restrictive as to which
5357 constant addresses are supported.
5360 @defmac CONSTANT_P (@var{x})
5361 @code{CONSTANT_P}, which is defined by target-independent code,
5362 accepts integer-values expressions whose values are not explicitly
5363 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5364 expressions and @code{const} arithmetic expressions, in addition to
5365 @code{const_int} and @code{const_double} expressions.
5368 @defmac MAX_REGS_PER_ADDRESS
5369 A number, the maximum number of registers that can appear in a valid
5370 memory address. Note that it is up to you to specify a value equal to
5371 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5375 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5376 A function that returns whether @var{x} (an RTX) is a legitimate memory
5377 address on the target machine for a memory operand of mode @var{mode}.
5379 Legitimate addresses are defined in two variants: a strict variant and a
5380 non-strict one. The @var{strict} parameter chooses which variant is
5381 desired by the caller.
5383 The strict variant is used in the reload pass. It must be defined so
5384 that any pseudo-register that has not been allocated a hard register is
5385 considered a memory reference. This is because in contexts where some
5386 kind of register is required, a pseudo-register with no hard register
5387 must be rejected. For non-hard registers, the strict variant should look
5388 up the @code{reg_renumber} array; it should then proceed using the hard
5389 register number in the array, or treat the pseudo as a memory reference
5390 if the array holds @code{-1}.
5392 The non-strict variant is used in other passes. It must be defined to
5393 accept all pseudo-registers in every context where some kind of
5394 register is required.
5396 Normally, constant addresses which are the sum of a @code{symbol_ref}
5397 and an integer are stored inside a @code{const} RTX to mark them as
5398 constant. Therefore, there is no need to recognize such sums
5399 specifically as legitimate addresses. Normally you would simply
5400 recognize any @code{const} as legitimate.
5402 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5403 sums that are not marked with @code{const}. It assumes that a naked
5404 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5405 naked constant sums as illegitimate addresses, so that none of them will
5406 be given to @code{PRINT_OPERAND_ADDRESS}.
5408 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5409 On some machines, whether a symbolic address is legitimate depends on
5410 the section that the address refers to. On these machines, define the
5411 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5412 into the @code{symbol_ref}, and then check for it here. When you see a
5413 @code{const}, you will have to look inside it to find the
5414 @code{symbol_ref} in order to determine the section. @xref{Assembler
5417 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5418 Some ports are still using a deprecated legacy substitute for
5419 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5423 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5427 and should @code{goto @var{label}} if the address @var{x} is a valid
5428 address on the target machine for a memory operand of mode @var{mode}.
5430 @findex REG_OK_STRICT
5431 Compiler source files that want to use the strict variant of this
5432 macro define the macro @code{REG_OK_STRICT}. You should use an
5433 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5434 that case and the non-strict variant otherwise.
5436 Using the hook is usually simpler because it limits the number of
5437 files that are recompiled when changes are made.
5440 @defmac TARGET_MEM_CONSTRAINT
5441 A single character to be used instead of the default @code{'m'}
5442 character for general memory addresses. This defines the constraint
5443 letter which matches the memory addresses accepted by
5444 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5445 support new address formats in your back end without changing the
5446 semantics of the @code{'m'} constraint. This is necessary in order to
5447 preserve functionality of inline assembly constructs using the
5448 @code{'m'} constraint.
5451 @defmac FIND_BASE_TERM (@var{x})
5452 A C expression to determine the base term of address @var{x},
5453 or to provide a simplified version of @var{x} from which @file{alias.c}
5454 can easily find the base term. This macro is used in only two places:
5455 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5457 It is always safe for this macro to not be defined. It exists so
5458 that alias analysis can understand machine-dependent addresses.
5460 The typical use of this macro is to handle addresses containing
5461 a label_ref or symbol_ref within an UNSPEC@.
5464 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode})
5465 This hook is given an invalid memory address @var{x} for an
5466 operand of mode @var{mode} and should try to return a valid memory
5469 @findex break_out_memory_refs
5470 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5471 and @var{oldx} will be the operand that was given to that function to produce
5474 The code of the hook should not alter the substructure of
5475 @var{x}. If it transforms @var{x} into a more legitimate form, it
5476 should return the new @var{x}.
5478 It is not necessary for this hook to come up with a legitimate address.
5479 The compiler has standard ways of doing so in all cases. In fact, it
5480 is safe to omit this hook or make it return @var{x} if it cannot find
5481 a valid way to legitimize the address. But often a machine-dependent
5482 strategy can generate better code.
5485 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5486 A C compound statement that attempts to replace @var{x}, which is an address
5487 that needs reloading, with a valid memory address for an operand of mode
5488 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5489 It is not necessary to define this macro, but it might be useful for
5490 performance reasons.
5492 For example, on the i386, it is sometimes possible to use a single
5493 reload register instead of two by reloading a sum of two pseudo
5494 registers into a register. On the other hand, for number of RISC
5495 processors offsets are limited so that often an intermediate address
5496 needs to be generated in order to address a stack slot. By defining
5497 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5498 generated for adjacent some stack slots can be made identical, and thus
5501 @emph{Note}: This macro should be used with caution. It is necessary
5502 to know something of how reload works in order to effectively use this,
5503 and it is quite easy to produce macros that build in too much knowledge
5504 of reload internals.
5506 @emph{Note}: This macro must be able to reload an address created by a
5507 previous invocation of this macro. If it fails to handle such addresses
5508 then the compiler may generate incorrect code or abort.
5511 The macro definition should use @code{push_reload} to indicate parts that
5512 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5513 suitable to be passed unaltered to @code{push_reload}.
5515 The code generated by this macro must not alter the substructure of
5516 @var{x}. If it transforms @var{x} into a more legitimate form, it
5517 should assign @var{x} (which will always be a C variable) a new value.
5518 This also applies to parts that you change indirectly by calling
5521 @findex strict_memory_address_p
5522 The macro definition may use @code{strict_memory_address_p} to test if
5523 the address has become legitimate.
5526 If you want to change only a part of @var{x}, one standard way of doing
5527 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5528 single level of rtl. Thus, if the part to be changed is not at the
5529 top level, you'll need to replace first the top level.
5530 It is not necessary for this macro to come up with a legitimate
5531 address; but often a machine-dependent strategy can generate better code.
5534 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr})
5535 This hook returns @code{true} if memory address @var{addr} can have
5536 different meanings depending on the machine mode of the memory
5537 reference it is used for or if the address is valid for some modes
5540 Autoincrement and autodecrement addresses typically have mode-dependent
5541 effects because the amount of the increment or decrement is the size
5542 of the operand being addressed. Some machines have other mode-dependent
5543 addresses. Many RISC machines have no mode-dependent addresses.
5545 You may assume that @var{addr} is a valid address for the machine.
5547 The default version of this hook returns @code{false}.
5550 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5551 A C statement or compound statement with a conditional @code{goto
5552 @var{label};} executed if memory address @var{x} (an RTX) can have
5553 different meanings depending on the machine mode of the memory
5554 reference it is used for or if the address is valid for some modes
5557 Autoincrement and autodecrement addresses typically have mode-dependent
5558 effects because the amount of the increment or decrement is the size
5559 of the operand being addressed. Some machines have other mode-dependent
5560 addresses. Many RISC machines have no mode-dependent addresses.
5562 You may assume that @var{addr} is a valid address for the machine.
5564 These are obsolete macros, replaced by the
5565 @code{TARGET_MODE_DEPENDENT_ADDRESS_P} target hook.
5568 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5569 This hook returns true if @var{x} is a legitimate constant for a
5570 @var{mode}-mode immediate operand on the target machine. You can assume that
5571 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5573 The default definition returns true.
5576 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5577 This hook is used to undo the possibly obfuscating effects of the
5578 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5579 macros. Some backend implementations of these macros wrap symbol
5580 references inside an @code{UNSPEC} rtx to represent PIC or similar
5581 addressing modes. This target hook allows GCC's optimizers to understand
5582 the semantics of these opaque @code{UNSPEC}s by converting them back
5583 into their original form.
5586 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (enum machine_mode @var{mode}, rtx @var{x})
5587 This hook should return true if @var{x} is of a form that cannot (or
5588 should not) be spilled to the constant pool. @var{mode} is the mode
5591 The default version of this hook returns false.
5593 The primary reason to define this hook is to prevent reload from
5594 deciding that a non-legitimate constant would be better reloaded
5595 from the constant pool instead of spilling and reloading a register
5596 holding the constant. This restriction is often true of addresses
5597 of TLS symbols for various targets.
5600 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, const_rtx @var{x})
5601 This hook should return true if pool entries for constant @var{x} can
5602 be placed in an @code{object_block} structure. @var{mode} is the mode
5605 The default version returns false for all constants.
5608 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (unsigned @var{fn}, bool @var{md_fn}, bool @var{sqrt})
5609 This hook should return the DECL of a function that implements reciprocal of
5610 the builtin function with builtin function code @var{fn}, or
5611 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5612 when @var{fn} is a code of a machine-dependent builtin function. When
5613 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5614 of a square root function are performed, and only reciprocals of @code{sqrt}
5618 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5619 This hook should return the DECL of a function @var{f} that given an
5620 address @var{addr} as an argument returns a mask @var{m} that can be
5621 used to extract from two vectors the relevant data that resides in
5622 @var{addr} in case @var{addr} is not properly aligned.
5624 The autovectorizer, when vectorizing a load operation from an address
5625 @var{addr} that may be unaligned, will generate two vector loads from
5626 the two aligned addresses around @var{addr}. It then generates a
5627 @code{REALIGN_LOAD} operation to extract the relevant data from the
5628 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5629 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5630 the third argument, @var{OFF}, defines how the data will be extracted
5631 from these two vectors: if @var{OFF} is 0, then the returned vector is
5632 @var{v2}; otherwise, the returned vector is composed from the last
5633 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5634 @var{OFF} elements of @var{v2}.
5636 If this hook is defined, the autovectorizer will generate a call
5637 to @var{f} (using the DECL tree that this hook returns) and will
5638 use the return value of @var{f} as the argument @var{OFF} to
5639 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5640 should comply with the semantics expected by @code{REALIGN_LOAD}
5642 If this hook is not defined, then @var{addr} will be used as
5643 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5644 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5647 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5648 This hook should return the DECL of a function @var{f} that implements
5649 widening multiplication of the even elements of two input vectors of type @var{x}.
5651 If this hook is defined, the autovectorizer will use it along with the
5652 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5653 widening multiplication in cases that the order of the results does not have to be
5654 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5655 @code{widen_mult_hi/lo} idioms will be used.
5658 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5659 This hook should return the DECL of a function @var{f} that implements
5660 widening multiplication of the odd elements of two input vectors of type @var{x}.
5662 If this hook is defined, the autovectorizer will use it along with the
5663 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5664 widening multiplication in cases that the order of the results does not have to be
5665 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5666 @code{widen_mult_hi/lo} idioms will be used.
5669 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5670 Returns cost of different scalar or vector statements for vectorization cost model.
5671 For vector memory operations the cost may depend on type (@var{vectype}) and
5672 misalignment value (@var{misalign}).
5675 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5676 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5679 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VEC_PERM (tree @var{type}, tree *@var{mask_element_type})
5680 Target builtin that implements vector permute.
5683 @deftypefn {Target Hook} bool TARGET_VECTORIZE_BUILTIN_VEC_PERM_OK (tree @var{vec_type}, tree @var{mask})
5684 Return true if a vector created for @code{builtin_vec_perm} is valid.
5687 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5688 This hook should return the DECL of a function that implements conversion of the
5689 input vector of type @var{src_type} to type @var{dest_type}.
5690 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5691 specifies how the conversion is to be applied
5692 (truncation, rounding, etc.).
5694 If this hook is defined, the autovectorizer will use the
5695 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5696 conversion. Otherwise, it will return @code{NULL_TREE}.
5699 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5700 This hook should return the decl of a function that implements the
5701 vectorized variant of the builtin function with builtin function code
5702 @var{code} or @code{NULL_TREE} if such a function is not available.
5703 The value of @var{fndecl} is the builtin function declaration. The
5704 return type of the vectorized function shall be of vector type
5705 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5708 @deftypefn {Target Hook} bool TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT (enum machine_mode @var{mode}, const_tree @var{type}, int @var{misalignment}, bool @var{is_packed})
5709 This hook should return true if the target supports misaligned vector
5710 store/load of a specific factor denoted in the @var{misalignment}
5711 parameter. The vector store/load should be of machine mode @var{mode} and
5712 the elements in the vectors should be of type @var{type}. @var{is_packed}
5713 parameter is true if the memory access is defined in a packed struct.
5716 @deftypefn {Target Hook} {enum machine_mode} TARGET_VECTORIZE_PREFERRED_SIMD_MODE (enum machine_mode @var{mode})
5717 This hook should return the preferred mode for vectorizing scalar
5718 mode @var{mode}. The default is
5719 equal to @code{word_mode}, because the vectorizer can do some
5720 transformations even in absence of specialized @acronym{SIMD} hardware.
5723 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void)
5724 This hook should return a mask of sizes that should be iterated over
5725 after trying to autovectorize using the vector size derived from the
5726 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5727 The default is zero which means to not iterate over other vector sizes.
5730 @node Anchored Addresses
5731 @section Anchored Addresses
5732 @cindex anchored addresses
5733 @cindex @option{-fsection-anchors}
5735 GCC usually addresses every static object as a separate entity.
5736 For example, if we have:
5740 int foo (void) @{ return a + b + c; @}
5743 the code for @code{foo} will usually calculate three separate symbolic
5744 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5745 it would be better to calculate just one symbolic address and access
5746 the three variables relative to it. The equivalent pseudocode would
5752 register int *xr = &x;
5753 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5757 (which isn't valid C). We refer to shared addresses like @code{x} as
5758 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5760 The hooks below describe the target properties that GCC needs to know
5761 in order to make effective use of section anchors. It won't use
5762 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5763 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5765 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5766 The minimum offset that should be applied to a section anchor.
5767 On most targets, it should be the smallest offset that can be
5768 applied to a base register while still giving a legitimate address
5769 for every mode. The default value is 0.
5772 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5773 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5774 offset that should be applied to section anchors. The default
5778 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5779 Write the assembly code to define section anchor @var{x}, which is a
5780 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5781 The hook is called with the assembly output position set to the beginning
5782 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5784 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5785 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5786 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5787 is @code{NULL}, which disables the use of section anchors altogether.
5790 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
5791 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5792 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5793 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5795 The default version is correct for most targets, but you might need to
5796 intercept this hook to handle things like target-specific attributes
5797 or target-specific sections.
5800 @node Condition Code
5801 @section Condition Code Status
5802 @cindex condition code status
5804 The macros in this section can be split in two families, according to the
5805 two ways of representing condition codes in GCC.
5807 The first representation is the so called @code{(cc0)} representation
5808 (@pxref{Jump Patterns}), where all instructions can have an implicit
5809 clobber of the condition codes. The second is the condition code
5810 register representation, which provides better schedulability for
5811 architectures that do have a condition code register, but on which
5812 most instructions do not affect it. The latter category includes
5815 The implicit clobbering poses a strong restriction on the placement of
5816 the definition and use of the condition code, which need to be in adjacent
5817 insns for machines using @code{(cc0)}. This can prevent important
5818 optimizations on some machines. For example, on the IBM RS/6000, there
5819 is a delay for taken branches unless the condition code register is set
5820 three instructions earlier than the conditional branch. The instruction
5821 scheduler cannot perform this optimization if it is not permitted to
5822 separate the definition and use of the condition code register.
5824 For this reason, it is possible and suggested to use a register to
5825 represent the condition code for new ports. If there is a specific
5826 condition code register in the machine, use a hard register. If the
5827 condition code or comparison result can be placed in any general register,
5828 or if there are multiple condition registers, use a pseudo register.
5829 Registers used to store the condition code value will usually have a mode
5830 that is in class @code{MODE_CC}.
5832 Alternatively, you can use @code{BImode} if the comparison operator is
5833 specified already in the compare instruction. In this case, you are not
5834 interested in most macros in this section.
5837 * CC0 Condition Codes:: Old style representation of condition codes.
5838 * MODE_CC Condition Codes:: Modern representation of condition codes.
5839 * Cond Exec Macros:: Macros to control conditional execution.
5842 @node CC0 Condition Codes
5843 @subsection Representation of condition codes using @code{(cc0)}
5847 The file @file{conditions.h} defines a variable @code{cc_status} to
5848 describe how the condition code was computed (in case the interpretation of
5849 the condition code depends on the instruction that it was set by). This
5850 variable contains the RTL expressions on which the condition code is
5851 currently based, and several standard flags.
5853 Sometimes additional machine-specific flags must be defined in the machine
5854 description header file. It can also add additional machine-specific
5855 information by defining @code{CC_STATUS_MDEP}.
5857 @defmac CC_STATUS_MDEP
5858 C code for a data type which is used for declaring the @code{mdep}
5859 component of @code{cc_status}. It defaults to @code{int}.
5861 This macro is not used on machines that do not use @code{cc0}.
5864 @defmac CC_STATUS_MDEP_INIT
5865 A C expression to initialize the @code{mdep} field to ``empty''.
5866 The default definition does nothing, since most machines don't use
5867 the field anyway. If you want to use the field, you should probably
5868 define this macro to initialize it.
5870 This macro is not used on machines that do not use @code{cc0}.
5873 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5874 A C compound statement to set the components of @code{cc_status}
5875 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5876 this macro's responsibility to recognize insns that set the condition
5877 code as a byproduct of other activity as well as those that explicitly
5880 This macro is not used on machines that do not use @code{cc0}.
5882 If there are insns that do not set the condition code but do alter
5883 other machine registers, this macro must check to see whether they
5884 invalidate the expressions that the condition code is recorded as
5885 reflecting. For example, on the 68000, insns that store in address
5886 registers do not set the condition code, which means that usually
5887 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5888 insns. But suppose that the previous insn set the condition code
5889 based on location @samp{a4@@(102)} and the current insn stores a new
5890 value in @samp{a4}. Although the condition code is not changed by
5891 this, it will no longer be true that it reflects the contents of
5892 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5893 @code{cc_status} in this case to say that nothing is known about the
5894 condition code value.
5896 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5897 with the results of peephole optimization: insns whose patterns are
5898 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5899 constants which are just the operands. The RTL structure of these
5900 insns is not sufficient to indicate what the insns actually do. What
5901 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5902 @code{CC_STATUS_INIT}.
5904 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5905 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5906 @samp{cc}. This avoids having detailed information about patterns in
5907 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5910 @node MODE_CC Condition Codes
5911 @subsection Representation of condition codes using registers
5915 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5916 On many machines, the condition code may be produced by other instructions
5917 than compares, for example the branch can use directly the condition
5918 code set by a subtract instruction. However, on some machines
5919 when the condition code is set this way some bits (such as the overflow
5920 bit) are not set in the same way as a test instruction, so that a different
5921 branch instruction must be used for some conditional branches. When
5922 this happens, use the machine mode of the condition code register to
5923 record different formats of the condition code register. Modes can
5924 also be used to record which compare instruction (e.g. a signed or an
5925 unsigned comparison) produced the condition codes.
5927 If other modes than @code{CCmode} are required, add them to
5928 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
5929 a mode given an operand of a compare. This is needed because the modes
5930 have to be chosen not only during RTL generation but also, for example,
5931 by instruction combination. The result of @code{SELECT_CC_MODE} should
5932 be consistent with the mode used in the patterns; for example to support
5933 the case of the add on the SPARC discussed above, we have the pattern
5937 [(set (reg:CC_NOOV 0)
5939 (plus:SI (match_operand:SI 0 "register_operand" "%r")
5940 (match_operand:SI 1 "arith_operand" "rI"))
5947 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
5948 for comparisons whose argument is a @code{plus}:
5951 #define SELECT_CC_MODE(OP,X,Y) \
5952 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5953 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5954 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5955 || GET_CODE (X) == NEG) \
5956 ? CC_NOOVmode : CCmode))
5959 Another reason to use modes is to retain information on which operands
5960 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
5963 You should define this macro if and only if you define extra CC modes
5964 in @file{@var{machine}-modes.def}.
5967 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5968 On some machines not all possible comparisons are defined, but you can
5969 convert an invalid comparison into a valid one. For example, the Alpha
5970 does not have a @code{GT} comparison, but you can use an @code{LT}
5971 comparison instead and swap the order of the operands.
5973 On such machines, define this macro to be a C statement to do any
5974 required conversions. @var{code} is the initial comparison code
5975 and @var{op0} and @var{op1} are the left and right operands of the
5976 comparison, respectively. You should modify @var{code}, @var{op0}, and
5977 @var{op1} as required.
5979 GCC will not assume that the comparison resulting from this macro is
5980 valid but will see if the resulting insn matches a pattern in the
5983 You need not define this macro if it would never change the comparison
5987 @defmac REVERSIBLE_CC_MODE (@var{mode})
5988 A C expression whose value is one if it is always safe to reverse a
5989 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5990 can ever return @var{mode} for a floating-point inequality comparison,
5991 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5993 You need not define this macro if it would always returns zero or if the
5994 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5995 For example, here is the definition used on the SPARC, where floating-point
5996 inequality comparisons are always given @code{CCFPEmode}:
5999 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
6003 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6004 A C expression whose value is reversed condition code of the @var{code} for
6005 comparison done in CC_MODE @var{mode}. The macro is used only in case
6006 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6007 machine has some non-standard way how to reverse certain conditionals. For
6008 instance in case all floating point conditions are non-trapping, compiler may
6009 freely convert unordered compares to ordered one. Then definition may look
6013 #define REVERSE_CONDITION(CODE, MODE) \
6014 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6015 : reverse_condition_maybe_unordered (CODE))
6019 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6020 On targets which do not use @code{(cc0)}, and which use a hard
6021 register rather than a pseudo-register to hold condition codes, the
6022 regular CSE passes are often not able to identify cases in which the
6023 hard register is set to a common value. Use this hook to enable a
6024 small pass which optimizes such cases. This hook should return true
6025 to enable this pass, and it should set the integers to which its
6026 arguments point to the hard register numbers used for condition codes.
6027 When there is only one such register, as is true on most systems, the
6028 integer pointed to by @var{p2} should be set to
6029 @code{INVALID_REGNUM}.
6031 The default version of this hook returns false.
6034 @deftypefn {Target Hook} {enum machine_mode} TARGET_CC_MODES_COMPATIBLE (enum machine_mode @var{m1}, enum machine_mode @var{m2})
6035 On targets which use multiple condition code modes in class
6036 @code{MODE_CC}, it is sometimes the case that a comparison can be
6037 validly done in more than one mode. On such a system, define this
6038 target hook to take two mode arguments and to return a mode in which
6039 both comparisons may be validly done. If there is no such mode,
6040 return @code{VOIDmode}.
6042 The default version of this hook checks whether the modes are the
6043 same. If they are, it returns that mode. If they are different, it
6044 returns @code{VOIDmode}.
6047 @node Cond Exec Macros
6048 @subsection Macros to control conditional execution
6049 @findex conditional execution
6052 There is one macro that may need to be defined for targets
6053 supporting conditional execution, independent of how they
6054 represent conditional branches.
6056 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6057 A C expression that returns true if the conditional execution predicate
6058 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6059 versa. Define this to return 0 if the target has conditional execution
6060 predicates that cannot be reversed safely. There is no need to validate
6061 that the arguments of op1 and op2 are the same, this is done separately.
6062 If no expansion is specified, this macro is defined as follows:
6065 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6066 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6071 @section Describing Relative Costs of Operations
6072 @cindex costs of instructions
6073 @cindex relative costs
6074 @cindex speed of instructions
6076 These macros let you describe the relative speed of various operations
6077 on the target machine.
6079 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6080 A C expression for the cost of moving data of mode @var{mode} from a
6081 register in class @var{from} to one in class @var{to}. The classes are
6082 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6083 value of 2 is the default; other values are interpreted relative to
6086 It is not required that the cost always equal 2 when @var{from} is the
6087 same as @var{to}; on some machines it is expensive to move between
6088 registers if they are not general registers.
6090 If reload sees an insn consisting of a single @code{set} between two
6091 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6092 classes returns a value of 2, reload does not check to ensure that the
6093 constraints of the insn are met. Setting a cost of other than 2 will
6094 allow reload to verify that the constraints are met. You should do this
6095 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6097 These macros are obsolete, new ports should use the target hook
6098 @code{TARGET_REGISTER_MOVE_COST} instead.
6101 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6102 This target hook should return the cost of moving data of mode @var{mode}
6103 from a register in class @var{from} to one in class @var{to}. The classes
6104 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6105 A value of 2 is the default; other values are interpreted relative to
6108 It is not required that the cost always equal 2 when @var{from} is the
6109 same as @var{to}; on some machines it is expensive to move between
6110 registers if they are not general registers.
6112 If reload sees an insn consisting of a single @code{set} between two
6113 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6114 classes returns a value of 2, reload does not check to ensure that the
6115 constraints of the insn are met. Setting a cost of other than 2 will
6116 allow reload to verify that the constraints are met. You should do this
6117 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6119 The default version of this function returns 2.
6122 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6123 A C expression for the cost of moving data of mode @var{mode} between a
6124 register of class @var{class} and memory; @var{in} is zero if the value
6125 is to be written to memory, nonzero if it is to be read in. This cost
6126 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6127 registers and memory is more expensive than between two registers, you
6128 should define this macro to express the relative cost.
6130 If you do not define this macro, GCC uses a default cost of 4 plus
6131 the cost of copying via a secondary reload register, if one is
6132 needed. If your machine requires a secondary reload register to copy
6133 between memory and a register of @var{class} but the reload mechanism is
6134 more complex than copying via an intermediate, define this macro to
6135 reflect the actual cost of the move.
6137 GCC defines the function @code{memory_move_secondary_cost} if
6138 secondary reloads are needed. It computes the costs due to copying via
6139 a secondary register. If your machine copies from memory using a
6140 secondary register in the conventional way but the default base value of
6141 4 is not correct for your machine, define this macro to add some other
6142 value to the result of that function. The arguments to that function
6143 are the same as to this macro.
6145 These macros are obsolete, new ports should use the target hook
6146 @code{TARGET_MEMORY_MOVE_COST} instead.
6149 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6150 This target hook should return the cost of moving data of mode @var{mode}
6151 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6152 if the value is to be written to memory, @code{true} if it is to be read in.
6153 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6154 If moving between registers and memory is more expensive than between two
6155 registers, you should add this target hook to express the relative cost.
6157 If you do not add this target hook, GCC uses a default cost of 4 plus
6158 the cost of copying via a secondary reload register, if one is
6159 needed. If your machine requires a secondary reload register to copy
6160 between memory and a register of @var{rclass} but the reload mechanism is
6161 more complex than copying via an intermediate, use this target hook to
6162 reflect the actual cost of the move.
6164 GCC defines the function @code{memory_move_secondary_cost} if
6165 secondary reloads are needed. It computes the costs due to copying via
6166 a secondary register. If your machine copies from memory using a
6167 secondary register in the conventional way but the default base value of
6168 4 is not correct for your machine, use this target hook to add some other
6169 value to the result of that function. The arguments to that function
6170 are the same as to this target hook.
6173 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6174 A C expression for the cost of a branch instruction. A value of 1 is
6175 the default; other values are interpreted relative to that. Parameter
6176 @var{speed_p} is true when the branch in question should be optimized
6177 for speed. When it is false, @code{BRANCH_COST} should return a value
6178 optimal for code size rather than performance. @var{predictable_p} is
6179 true for well-predicted branches. On many architectures the
6180 @code{BRANCH_COST} can be reduced then.
6183 Here are additional macros which do not specify precise relative costs,
6184 but only that certain actions are more expensive than GCC would
6187 @defmac SLOW_BYTE_ACCESS
6188 Define this macro as a C expression which is nonzero if accessing less
6189 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6190 faster than accessing a word of memory, i.e., if such access
6191 require more than one instruction or if there is no difference in cost
6192 between byte and (aligned) word loads.
6194 When this macro is not defined, the compiler will access a field by
6195 finding the smallest containing object; when it is defined, a fullword
6196 load will be used if alignment permits. Unless bytes accesses are
6197 faster than word accesses, using word accesses is preferable since it
6198 may eliminate subsequent memory access if subsequent accesses occur to
6199 other fields in the same word of the structure, but to different bytes.
6202 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6203 Define this macro to be the value 1 if memory accesses described by the
6204 @var{mode} and @var{alignment} parameters have a cost many times greater
6205 than aligned accesses, for example if they are emulated in a trap
6208 When this macro is nonzero, the compiler will act as if
6209 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6210 moves. This can cause significantly more instructions to be produced.
6211 Therefore, do not set this macro nonzero if unaligned accesses only add a
6212 cycle or two to the time for a memory access.
6214 If the value of this macro is always zero, it need not be defined. If
6215 this macro is defined, it should produce a nonzero value when
6216 @code{STRICT_ALIGNMENT} is nonzero.
6219 @defmac MOVE_RATIO (@var{speed})
6220 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6221 which a sequence of insns should be generated instead of a
6222 string move insn or a library call. Increasing the value will always
6223 make code faster, but eventually incurs high cost in increased code size.
6225 Note that on machines where the corresponding move insn is a
6226 @code{define_expand} that emits a sequence of insns, this macro counts
6227 the number of such sequences.
6229 The parameter @var{speed} is true if the code is currently being
6230 optimized for speed rather than size.
6232 If you don't define this, a reasonable default is used.
6235 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6236 A C expression used to determine whether @code{move_by_pieces} will be used to
6237 copy a chunk of memory, or whether some other block move mechanism
6238 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6239 than @code{MOVE_RATIO}.
6242 @defmac MOVE_MAX_PIECES
6243 A C expression used by @code{move_by_pieces} to determine the largest unit
6244 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6247 @defmac CLEAR_RATIO (@var{speed})
6248 The threshold of number of scalar move insns, @emph{below} which a sequence
6249 of insns should be generated to clear memory instead of a string clear insn
6250 or a library call. Increasing the value will always make code faster, but
6251 eventually incurs high cost in increased code size.
6253 The parameter @var{speed} is true if the code is currently being
6254 optimized for speed rather than size.
6256 If you don't define this, a reasonable default is used.
6259 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6260 A C expression used to determine whether @code{clear_by_pieces} will be used
6261 to clear a chunk of memory, or whether some other block clear mechanism
6262 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6263 than @code{CLEAR_RATIO}.
6266 @defmac SET_RATIO (@var{speed})
6267 The threshold of number of scalar move insns, @emph{below} which a sequence
6268 of insns should be generated to set memory to a constant value, instead of
6269 a block set insn or a library call.
6270 Increasing the value will always make code faster, but
6271 eventually incurs high cost in increased code size.
6273 The parameter @var{speed} is true if the code is currently being
6274 optimized for speed rather than size.
6276 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6279 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6280 A C expression used to determine whether @code{store_by_pieces} will be
6281 used to set a chunk of memory to a constant value, or whether some
6282 other mechanism will be used. Used by @code{__builtin_memset} when
6283 storing values other than constant zero.
6284 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6285 than @code{SET_RATIO}.
6288 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6289 A C expression used to determine whether @code{store_by_pieces} will be
6290 used to set a chunk of memory to a constant string value, or whether some
6291 other mechanism will be used. Used by @code{__builtin_strcpy} when
6292 called with a constant source string.
6293 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6294 than @code{MOVE_RATIO}.
6297 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6298 A C expression used to determine whether a load postincrement is a good
6299 thing to use for a given mode. Defaults to the value of
6300 @code{HAVE_POST_INCREMENT}.
6303 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6304 A C expression used to determine whether a load postdecrement is a good
6305 thing to use for a given mode. Defaults to the value of
6306 @code{HAVE_POST_DECREMENT}.
6309 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6310 A C expression used to determine whether a load preincrement is a good
6311 thing to use for a given mode. Defaults to the value of
6312 @code{HAVE_PRE_INCREMENT}.
6315 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6316 A C expression used to determine whether a load predecrement is a good
6317 thing to use for a given mode. Defaults to the value of
6318 @code{HAVE_PRE_DECREMENT}.
6321 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6322 A C expression used to determine whether a store postincrement is a good
6323 thing to use for a given mode. Defaults to the value of
6324 @code{HAVE_POST_INCREMENT}.
6327 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6328 A C expression used to determine whether a store postdecrement is a good
6329 thing to use for a given mode. Defaults to the value of
6330 @code{HAVE_POST_DECREMENT}.
6333 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6334 This macro is used to determine whether a store preincrement is a good
6335 thing to use for a given mode. Defaults to the value of
6336 @code{HAVE_PRE_INCREMENT}.
6339 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6340 This macro is used to determine whether a store predecrement is a good
6341 thing to use for a given mode. Defaults to the value of
6342 @code{HAVE_PRE_DECREMENT}.
6345 @defmac NO_FUNCTION_CSE
6346 Define this macro if it is as good or better to call a constant
6347 function address than to call an address kept in a register.
6350 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6351 Define this macro if a non-short-circuit operation produced by
6352 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6353 @code{BRANCH_COST} is greater than or equal to the value 2.
6356 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total}, bool @var{speed})
6357 This target hook describes the relative costs of RTL expressions.
6359 The cost may depend on the precise form of the expression, which is
6360 available for examination in @var{x}, and the rtx code of the expression
6361 in which it is contained, found in @var{outer_code}. @var{code} is the
6362 expression code---redundant, since it can be obtained with
6363 @code{GET_CODE (@var{x})}.
6365 In implementing this hook, you can use the construct
6366 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6369 On entry to the hook, @code{*@var{total}} contains a default estimate
6370 for the cost of the expression. The hook should modify this value as
6371 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6372 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6373 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6375 When optimizing for code size, i.e.@: when @code{speed} is
6376 false, this target hook should be used to estimate the relative
6377 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6379 The hook returns true when all subexpressions of @var{x} have been
6380 processed, and false when @code{rtx_cost} should recurse.
6383 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, bool @var{speed})
6384 This hook computes the cost of an addressing mode that contains
6385 @var{address}. If not defined, the cost is computed from
6386 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6388 For most CISC machines, the default cost is a good approximation of the
6389 true cost of the addressing mode. However, on RISC machines, all
6390 instructions normally have the same length and execution time. Hence
6391 all addresses will have equal costs.
6393 In cases where more than one form of an address is known, the form with
6394 the lowest cost will be used. If multiple forms have the same, lowest,
6395 cost, the one that is the most complex will be used.
6397 For example, suppose an address that is equal to the sum of a register
6398 and a constant is used twice in the same basic block. When this macro
6399 is not defined, the address will be computed in a register and memory
6400 references will be indirect through that register. On machines where
6401 the cost of the addressing mode containing the sum is no higher than
6402 that of a simple indirect reference, this will produce an additional
6403 instruction and possibly require an additional register. Proper
6404 specification of this macro eliminates this overhead for such machines.
6406 This hook is never called with an invalid address.
6408 On machines where an address involving more than one register is as
6409 cheap as an address computation involving only one register, defining
6410 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6411 be live over a region of code where only one would have been if
6412 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6413 should be considered in the definition of this macro. Equivalent costs
6414 should probably only be given to addresses with different numbers of
6415 registers on machines with lots of registers.
6419 @section Adjusting the Instruction Scheduler
6421 The instruction scheduler may need a fair amount of machine-specific
6422 adjustment in order to produce good code. GCC provides several target
6423 hooks for this purpose. It is usually enough to define just a few of
6424 them: try the first ones in this list first.
6426 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6427 This hook returns the maximum number of instructions that can ever
6428 issue at the same time on the target machine. The default is one.
6429 Although the insn scheduler can define itself the possibility of issue
6430 an insn on the same cycle, the value can serve as an additional
6431 constraint to issue insns on the same simulated processor cycle (see
6432 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6433 This value must be constant over the entire compilation. If you need
6434 it to vary depending on what the instructions are, you must use
6435 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6438 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6439 This hook is executed by the scheduler after it has scheduled an insn
6440 from the ready list. It should return the number of insns which can
6441 still be issued in the current cycle. The default is
6442 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6443 @code{USE}, which normally are not counted against the issue rate.
6444 You should define this hook if some insns take more machine resources
6445 than others, so that fewer insns can follow them in the same cycle.
6446 @var{file} is either a null pointer, or a stdio stream to write any
6447 debug output to. @var{verbose} is the verbose level provided by
6448 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6452 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6453 This function corrects the value of @var{cost} based on the
6454 relationship between @var{insn} and @var{dep_insn} through the
6455 dependence @var{link}. It should return the new value. The default
6456 is to make no adjustment to @var{cost}. This can be used for example
6457 to specify to the scheduler using the traditional pipeline description
6458 that an output- or anti-dependence does not incur the same cost as a
6459 data-dependence. If the scheduler using the automaton based pipeline
6460 description, the cost of anti-dependence is zero and the cost of
6461 output-dependence is maximum of one and the difference of latency
6462 times of the first and the second insns. If these values are not
6463 acceptable, you could use the hook to modify them too. See also
6464 @pxref{Processor pipeline description}.
6467 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6468 This hook adjusts the integer scheduling priority @var{priority} of
6469 @var{insn}. It should return the new priority. Increase the priority to
6470 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6471 later. Do not define this hook if you do not need to adjust the
6472 scheduling priorities of insns.
6475 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6476 This hook is executed by the scheduler after it has scheduled the ready
6477 list, to allow the machine description to reorder it (for example to
6478 combine two small instructions together on @samp{VLIW} machines).
6479 @var{file} is either a null pointer, or a stdio stream to write any
6480 debug output to. @var{verbose} is the verbose level provided by
6481 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6482 list of instructions that are ready to be scheduled. @var{n_readyp} is
6483 a pointer to the number of elements in the ready list. The scheduler
6484 reads the ready list in reverse order, starting with
6485 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6486 is the timer tick of the scheduler. You may modify the ready list and
6487 the number of ready insns. The return value is the number of insns that
6488 can issue this cycle; normally this is just @code{issue_rate}. See also
6489 @samp{TARGET_SCHED_REORDER2}.
6492 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6493 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6494 function is called whenever the scheduler starts a new cycle. This one
6495 is called once per iteration over a cycle, immediately after
6496 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6497 return the number of insns to be scheduled in the same cycle. Defining
6498 this hook can be useful if there are frequent situations where
6499 scheduling one insn causes other insns to become ready in the same
6500 cycle. These other insns can then be taken into account properly.
6503 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6504 This hook is called after evaluation forward dependencies of insns in
6505 chain given by two parameter values (@var{head} and @var{tail}
6506 correspondingly) but before insns scheduling of the insn chain. For
6507 example, it can be used for better insn classification if it requires
6508 analysis of dependencies. This hook can use backward and forward
6509 dependencies of the insn scheduler because they are already
6513 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6514 This hook is executed by the scheduler at the beginning of each block of
6515 instructions that are to be scheduled. @var{file} is either a null
6516 pointer, or a stdio stream to write any debug output to. @var{verbose}
6517 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6518 @var{max_ready} is the maximum number of insns in the current scheduling
6519 region that can be live at the same time. This can be used to allocate
6520 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6523 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6524 This hook is executed by the scheduler at the end of each block of
6525 instructions that are to be scheduled. It can be used to perform
6526 cleanup of any actions done by the other scheduling hooks. @var{file}
6527 is either a null pointer, or a stdio stream to write any debug output
6528 to. @var{verbose} is the verbose level provided by
6529 @option{-fsched-verbose-@var{n}}.
6532 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6533 This hook is executed by the scheduler after function level initializations.
6534 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6535 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6536 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6539 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6540 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6541 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6542 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6545 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6546 The hook returns an RTL insn. The automaton state used in the
6547 pipeline hazard recognizer is changed as if the insn were scheduled
6548 when the new simulated processor cycle starts. Usage of the hook may
6549 simplify the automaton pipeline description for some @acronym{VLIW}
6550 processors. If the hook is defined, it is used only for the automaton
6551 based pipeline description. The default is not to change the state
6552 when the new simulated processor cycle starts.
6555 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6556 The hook can be used to initialize data used by the previous hook.
6559 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6560 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6561 to changed the state as if the insn were scheduled when the new
6562 simulated processor cycle finishes.
6565 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6566 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6567 used to initialize data used by the previous hook.
6570 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
6571 The hook to notify target that the current simulated cycle is about to finish.
6572 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6573 to change the state in more complicated situations - e.g., when advancing
6574 state on a single insn is not enough.
6577 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
6578 The hook to notify target that new simulated cycle has just started.
6579 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6580 to change the state in more complicated situations - e.g., when advancing
6581 state on a single insn is not enough.
6584 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6585 This hook controls better choosing an insn from the ready insn queue
6586 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6587 chooses the first insn from the queue. If the hook returns a positive
6588 value, an additional scheduler code tries all permutations of
6589 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6590 subsequent ready insns to choose an insn whose issue will result in
6591 maximal number of issued insns on the same cycle. For the
6592 @acronym{VLIW} processor, the code could actually solve the problem of
6593 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6594 rules of @acronym{VLIW} packing are described in the automaton.
6596 This code also could be used for superscalar @acronym{RISC}
6597 processors. Let us consider a superscalar @acronym{RISC} processor
6598 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6599 @var{B}, some insns can be executed only in pipelines @var{B} or
6600 @var{C}, and one insn can be executed in pipeline @var{B}. The
6601 processor may issue the 1st insn into @var{A} and the 2nd one into
6602 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6603 until the next cycle. If the scheduler issues the 3rd insn the first,
6604 the processor could issue all 3 insns per cycle.
6606 Actually this code demonstrates advantages of the automaton based
6607 pipeline hazard recognizer. We try quickly and easy many insn
6608 schedules to choose the best one.
6610 The default is no multipass scheduling.
6613 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx @var{insn})
6615 This hook controls what insns from the ready insn queue will be
6616 considered for the multipass insn scheduling. If the hook returns
6617 zero for @var{insn}, the insn will be not chosen to
6620 The default is that any ready insns can be chosen to be issued.
6623 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN (void *@var{data}, char *@var{ready_try}, int @var{n_ready}, bool @var{first_cycle_insn_p})
6624 This hook prepares the target backend for a new round of multipass
6628 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE (void *@var{data}, char *@var{ready_try}, int @var{n_ready}, rtx @var{insn}, const void *@var{prev_data})
6629 This hook is called when multipass scheduling evaluates instruction INSN.
6632 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, char *@var{ready_try}, int @var{n_ready})
6633 This is called when multipass scheduling backtracks from evaluation of
6637 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
6638 This hook notifies the target about the result of the concluded current
6639 round of multipass scheduling.
6642 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
6643 This hook initializes target-specific data used in multipass scheduling.
6646 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
6647 This hook finalizes target-specific data used in multipass scheduling.
6650 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *@var{dump}, int @var{verbose}, rtx @var{insn}, int @var{last_clock}, int @var{clock}, int *@var{sort_p})
6651 This hook is called by the insn scheduler before issuing @var{insn}
6652 on cycle @var{clock}. If the hook returns nonzero,
6653 @var{insn} is not issued on this processor cycle. Instead,
6654 the processor cycle is advanced. If *@var{sort_p}
6655 is zero, the insn ready queue is not sorted on the new cycle
6656 start as usually. @var{dump} and @var{verbose} specify the file and
6657 verbosity level to use for debugging output.
6658 @var{last_clock} and @var{clock} are, respectively, the
6659 processor cycle on which the previous insn has been issued,
6660 and the current processor cycle.
6663 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
6664 This hook is used to define which dependences are considered costly by
6665 the target, so costly that it is not advisable to schedule the insns that
6666 are involved in the dependence too close to one another. The parameters
6667 to this hook are as follows: The first parameter @var{_dep} is the dependence
6668 being evaluated. The second parameter @var{cost} is the cost of the
6669 dependence as estimated by the scheduler, and the third
6670 parameter @var{distance} is the distance in cycles between the two insns.
6671 The hook returns @code{true} if considering the distance between the two
6672 insns the dependence between them is considered costly by the target,
6673 and @code{false} otherwise.
6675 Defining this hook can be useful in multiple-issue out-of-order machines,
6676 where (a) it's practically hopeless to predict the actual data/resource
6677 delays, however: (b) there's a better chance to predict the actual grouping
6678 that will be formed, and (c) correctly emulating the grouping can be very
6679 important. In such targets one may want to allow issuing dependent insns
6680 closer to one another---i.e., closer than the dependence distance; however,
6681 not in cases of ``costly dependences'', which this hooks allows to define.
6684 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6685 This hook is called by the insn scheduler after emitting a new instruction to
6686 the instruction stream. The hook notifies a target backend to extend its
6687 per instruction data structures.
6690 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6691 Return a pointer to a store large enough to hold target scheduling context.
6694 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6695 Initialize store pointed to by @var{tc} to hold target scheduling context.
6696 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6697 beginning of the block. Otherwise, copy the current context into @var{tc}.
6700 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6701 Copy target scheduling context pointed to by @var{tc} to the current context.
6704 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6705 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6708 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6709 Deallocate a store for target scheduling context pointed to by @var{tc}.
6712 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6713 This hook is called by the insn scheduler when @var{insn} has only
6714 speculative dependencies and therefore can be scheduled speculatively.
6715 The hook is used to check if the pattern of @var{insn} has a speculative
6716 version and, in case of successful check, to generate that speculative
6717 pattern. The hook should return 1, if the instruction has a speculative form,
6718 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6719 speculation. If the return value equals 1 then @var{new_pat} is assigned
6720 the generated speculative pattern.
6723 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (int @var{dep_status})
6724 This hook is called by the insn scheduler during generation of recovery code
6725 for @var{insn}. It should return @code{true}, if the corresponding check
6726 instruction should branch to recovery code, or @code{false} otherwise.
6729 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6730 This hook is called by the insn scheduler to generate a pattern for recovery
6731 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6732 speculative instruction for which the check should be generated.
6733 @var{label} is either a label of a basic block, where recovery code should
6734 be emitted, or a null pointer, when requested check doesn't branch to
6735 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6736 a pattern for a branchy check corresponding to a simple check denoted by
6737 @var{insn} should be generated. In this case @var{label} can't be null.
6740 @deftypefn {Target Hook} bool TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (const_rtx @var{insn})
6741 This hook is used as a workaround for
6742 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6743 called on the first instruction of the ready list. The hook is used to
6744 discard speculative instructions that stand first in the ready list from
6745 being scheduled on the current cycle. If the hook returns @code{false},
6746 @var{insn} will not be chosen to be issued.
6747 For non-speculative instructions,
6748 the hook should always return @code{true}. For example, in the ia64 backend
6749 the hook is used to cancel data speculative insns when the ALAT table
6753 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
6754 This hook is used by the insn scheduler to find out what features should be
6756 The structure *@var{spec_info} should be filled in by the target.
6757 The structure describes speculation types that can be used in the scheduler.
6760 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6761 This hook is called by the swing modulo scheduler to calculate a
6762 resource-based lower bound which is based on the resources available in
6763 the machine and the resources required by each instruction. The target
6764 backend can use @var{g} to calculate such bound. A very simple lower
6765 bound will be used in case this hook is not implemented: the total number
6766 of instructions divided by the issue rate.
6769 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx @var{insn}, int @var{x})
6770 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6771 is supported in hardware and the condition specified in the parameter is true.
6774 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx @var{insn}, int @var{x})
6775 This hook is called by Haifa Scheduler. It performs the operation specified
6776 in its second parameter.
6780 @section Dividing the Output into Sections (Texts, Data, @dots{})
6781 @c the above section title is WAY too long. maybe cut the part between
6782 @c the (...)? --mew 10feb93
6784 An object file is divided into sections containing different types of
6785 data. In the most common case, there are three sections: the @dfn{text
6786 section}, which holds instructions and read-only data; the @dfn{data
6787 section}, which holds initialized writable data; and the @dfn{bss
6788 section}, which holds uninitialized data. Some systems have other kinds
6791 @file{varasm.c} provides several well-known sections, such as
6792 @code{text_section}, @code{data_section} and @code{bss_section}.
6793 The normal way of controlling a @code{@var{foo}_section} variable
6794 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6795 as described below. The macros are only read once, when @file{varasm.c}
6796 initializes itself, so their values must be run-time constants.
6797 They may however depend on command-line flags.
6799 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6800 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6801 to be string literals.
6803 Some assemblers require a different string to be written every time a
6804 section is selected. If your assembler falls into this category, you
6805 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6806 @code{get_unnamed_section} to set up the sections.
6808 You must always create a @code{text_section}, either by defining
6809 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6810 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6811 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6812 create a distinct @code{readonly_data_section}, the default is to
6813 reuse @code{text_section}.
6815 All the other @file{varasm.c} sections are optional, and are null
6816 if the target does not provide them.
6818 @defmac TEXT_SECTION_ASM_OP
6819 A C expression whose value is a string, including spacing, containing the
6820 assembler operation that should precede instructions and read-only data.
6821 Normally @code{"\t.text"} is right.
6824 @defmac HOT_TEXT_SECTION_NAME
6825 If defined, a C string constant for the name of the section containing most
6826 frequently executed functions of the program. If not defined, GCC will provide
6827 a default definition if the target supports named sections.
6830 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6831 If defined, a C string constant for the name of the section containing unlikely
6832 executed functions in the program.
6835 @defmac DATA_SECTION_ASM_OP
6836 A C expression whose value is a string, including spacing, containing the
6837 assembler operation to identify the following data as writable initialized
6838 data. Normally @code{"\t.data"} is right.
6841 @defmac SDATA_SECTION_ASM_OP
6842 If defined, a C expression whose value is a string, including spacing,
6843 containing the assembler operation to identify the following data as
6844 initialized, writable small data.
6847 @defmac READONLY_DATA_SECTION_ASM_OP
6848 A C expression whose value is a string, including spacing, containing the
6849 assembler operation to identify the following data as read-only initialized
6853 @defmac BSS_SECTION_ASM_OP
6854 If defined, a C expression whose value is a string, including spacing,
6855 containing the assembler operation to identify the following data as
6856 uninitialized global data. If not defined, and
6857 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
6858 uninitialized global data will be output in the data section if
6859 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6863 @defmac SBSS_SECTION_ASM_OP
6864 If defined, a C expression whose value is a string, including spacing,
6865 containing the assembler operation to identify the following data as
6866 uninitialized, writable small data.
6869 @defmac TLS_COMMON_ASM_OP
6870 If defined, a C expression whose value is a string containing the
6871 assembler operation to identify the following data as thread-local
6872 common data. The default is @code{".tls_common"}.
6875 @defmac TLS_SECTION_ASM_FLAG
6876 If defined, a C expression whose value is a character constant
6877 containing the flag used to mark a section as a TLS section. The
6878 default is @code{'T'}.
6881 @defmac INIT_SECTION_ASM_OP
6882 If defined, a C expression whose value is a string, including spacing,
6883 containing the assembler operation to identify the following data as
6884 initialization code. If not defined, GCC will assume such a section does
6885 not exist. This section has no corresponding @code{init_section}
6886 variable; it is used entirely in runtime code.
6889 @defmac FINI_SECTION_ASM_OP
6890 If defined, a C expression whose value is a string, including spacing,
6891 containing the assembler operation to identify the following data as
6892 finalization code. If not defined, GCC will assume such a section does
6893 not exist. This section has no corresponding @code{fini_section}
6894 variable; it is used entirely in runtime code.
6897 @defmac INIT_ARRAY_SECTION_ASM_OP
6898 If defined, a C expression whose value is a string, including spacing,
6899 containing the assembler operation to identify the following data as
6900 part of the @code{.init_array} (or equivalent) section. If not
6901 defined, GCC will assume such a section does not exist. Do not define
6902 both this macro and @code{INIT_SECTION_ASM_OP}.
6905 @defmac FINI_ARRAY_SECTION_ASM_OP
6906 If defined, a C expression whose value is a string, including spacing,
6907 containing the assembler operation to identify the following data as
6908 part of the @code{.fini_array} (or equivalent) section. If not
6909 defined, GCC will assume such a section does not exist. Do not define
6910 both this macro and @code{FINI_SECTION_ASM_OP}.
6913 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6914 If defined, an ASM statement that switches to a different section
6915 via @var{section_op}, calls @var{function}, and switches back to
6916 the text section. This is used in @file{crtstuff.c} if
6917 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6918 to initialization and finalization functions from the init and fini
6919 sections. By default, this macro uses a simple function call. Some
6920 ports need hand-crafted assembly code to avoid dependencies on
6921 registers initialized in the function prologue or to ensure that
6922 constant pools don't end up too far way in the text section.
6925 @defmac TARGET_LIBGCC_SDATA_SECTION
6926 If defined, a string which names the section into which small
6927 variables defined in crtstuff and libgcc should go. This is useful
6928 when the target has options for optimizing access to small data, and
6929 you want the crtstuff and libgcc routines to be conservative in what
6930 they expect of your application yet liberal in what your application
6931 expects. For example, for targets with a @code{.sdata} section (like
6932 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6933 require small data support from your application, but use this macro
6934 to put small data into @code{.sdata} so that your application can
6935 access these variables whether it uses small data or not.
6938 @defmac FORCE_CODE_SECTION_ALIGN
6939 If defined, an ASM statement that aligns a code section to some
6940 arbitrary boundary. This is used to force all fragments of the
6941 @code{.init} and @code{.fini} sections to have to same alignment
6942 and thus prevent the linker from having to add any padding.
6945 @defmac JUMP_TABLES_IN_TEXT_SECTION
6946 Define this macro to be an expression with a nonzero value if jump
6947 tables (for @code{tablejump} insns) should be output in the text
6948 section, along with the assembler instructions. Otherwise, the
6949 readonly data section is used.
6951 This macro is irrelevant if there is no separate readonly data section.
6954 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6955 Define this hook if you need to do something special to set up the
6956 @file{varasm.c} sections, or if your target has some special sections
6957 of its own that you need to create.
6959 GCC calls this hook after processing the command line, but before writing
6960 any assembly code, and before calling any of the section-returning hooks
6964 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
6965 Return a mask describing how relocations should be treated when
6966 selecting sections. Bit 1 should be set if global relocations
6967 should be placed in a read-write section; bit 0 should be set if
6968 local relocations should be placed in a read-write section.
6970 The default version of this function returns 3 when @option{-fpic}
6971 is in effect, and 0 otherwise. The hook is typically redefined
6972 when the target cannot support (some kinds of) dynamic relocations
6973 in read-only sections even in executables.
6976 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6977 Return the section into which @var{exp} should be placed. You can
6978 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6979 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6980 requires link-time relocations. Bit 0 is set when variable contains
6981 local relocations only, while bit 1 is set for global relocations.
6982 @var{align} is the constant alignment in bits.
6984 The default version of this function takes care of putting read-only
6985 variables in @code{readonly_data_section}.
6987 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6990 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6991 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6992 for @code{FUNCTION_DECL}s as well as for variables and constants.
6994 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6995 function has been determined to be likely to be called, and nonzero if
6996 it is unlikely to be called.
6999 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7000 Build up a unique section name, expressed as a @code{STRING_CST} node,
7001 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7002 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7003 the initial value of @var{exp} requires link-time relocations.
7005 The default version of this function appends the symbol name to the
7006 ELF section name that would normally be used for the symbol. For
7007 example, the function @code{foo} would be placed in @code{.text.foo}.
7008 Whatever the actual target object format, this is often good enough.
7011 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7012 Return the readonly data section associated with
7013 @samp{DECL_SECTION_NAME (@var{decl})}.
7014 The default version of this function selects @code{.gnu.linkonce.r.name} if
7015 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7016 if function is in @code{.text.name}, and the normal readonly-data section
7020 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7021 Return the section into which a constant @var{x}, of mode @var{mode},
7022 should be placed. You can assume that @var{x} is some kind of
7023 constant in RTL@. The argument @var{mode} is redundant except in the
7024 case of a @code{const_int} rtx. @var{align} is the constant alignment
7027 The default version of this function takes care of putting symbolic
7028 constants in @code{flag_pic} mode in @code{data_section} and everything
7029 else in @code{readonly_data_section}.
7032 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7033 Define this hook if you need to postprocess the assembler name generated
7034 by target-independent code. The @var{id} provided to this hook will be
7035 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7036 or the mangled name of the @var{decl} in C++). The return value of the
7037 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7038 your target system. The default implementation of this hook just
7039 returns the @var{id} provided.
7042 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7043 Define this hook if references to a symbol or a constant must be
7044 treated differently depending on something about the variable or
7045 function named by the symbol (such as what section it is in).
7047 The hook is executed immediately after rtl has been created for
7048 @var{decl}, which may be a variable or function declaration or
7049 an entry in the constant pool. In either case, @var{rtl} is the
7050 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7051 in this hook; that field may not have been initialized yet.
7053 In the case of a constant, it is safe to assume that the rtl is
7054 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7055 will also have this form, but that is not guaranteed. Global
7056 register variables, for instance, will have a @code{reg} for their
7057 rtl. (Normally the right thing to do with such unusual rtl is
7060 The @var{new_decl_p} argument will be true if this is the first time
7061 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7062 be false for subsequent invocations, which will happen for duplicate
7063 declarations. Whether or not anything must be done for the duplicate
7064 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7065 @var{new_decl_p} is always true when the hook is called for a constant.
7067 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7068 The usual thing for this hook to do is to record flags in the
7069 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7070 Historically, the name string was modified if it was necessary to
7071 encode more than one bit of information, but this practice is now
7072 discouraged; use @code{SYMBOL_REF_FLAGS}.
7074 The default definition of this hook, @code{default_encode_section_info}
7075 in @file{varasm.c}, sets a number of commonly-useful bits in
7076 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7077 before overriding it.
7080 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7081 Decode @var{name} and return the real name part, sans
7082 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7086 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7087 Returns true if @var{exp} should be placed into a ``small data'' section.
7088 The default version of this hook always returns false.
7091 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7092 Contains the value true if the target places read-only
7093 ``small data'' into a separate section. The default value is false.
7096 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7097 It returns true if target wants profile code emitted before prologue.
7099 The default version of this hook use the target macro
7100 @code{PROFILE_BEFORE_PROLOGUE}.
7103 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7104 Returns true if @var{exp} names an object for which name resolution
7105 rules must resolve to the current ``module'' (dynamic shared library
7106 or executable image).
7108 The default version of this hook implements the name resolution rules
7109 for ELF, which has a looser model of global name binding than other
7110 currently supported object file formats.
7113 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7114 Contains the value true if the target supports thread-local storage.
7115 The default value is false.
7120 @section Position Independent Code
7121 @cindex position independent code
7124 This section describes macros that help implement generation of position
7125 independent code. Simply defining these macros is not enough to
7126 generate valid PIC; you must also add support to the hook
7127 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7128 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7129 must modify the definition of @samp{movsi} to do something appropriate
7130 when the source operand contains a symbolic address. You may also
7131 need to alter the handling of switch statements so that they use
7133 @c i rearranged the order of the macros above to try to force one of
7134 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7136 @defmac PIC_OFFSET_TABLE_REGNUM
7137 The register number of the register used to address a table of static
7138 data addresses in memory. In some cases this register is defined by a
7139 processor's ``application binary interface'' (ABI)@. When this macro
7140 is defined, RTL is generated for this register once, as with the stack
7141 pointer and frame pointer registers. If this macro is not defined, it
7142 is up to the machine-dependent files to allocate such a register (if
7143 necessary). Note that this register must be fixed when in use (e.g.@:
7144 when @code{flag_pic} is true).
7147 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7148 A C expression that is nonzero if the register defined by
7149 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7150 the default is zero. Do not define
7151 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7154 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7155 A C expression that is nonzero if @var{x} is a legitimate immediate
7156 operand on the target machine when generating position independent code.
7157 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7158 check this. You can also assume @var{flag_pic} is true, so you need not
7159 check it either. You need not define this macro if all constants
7160 (including @code{SYMBOL_REF}) can be immediate operands when generating
7161 position independent code.
7164 @node Assembler Format
7165 @section Defining the Output Assembler Language
7167 This section describes macros whose principal purpose is to describe how
7168 to write instructions in assembler language---rather than what the
7172 * File Framework:: Structural information for the assembler file.
7173 * Data Output:: Output of constants (numbers, strings, addresses).
7174 * Uninitialized Data:: Output of uninitialized variables.
7175 * Label Output:: Output and generation of labels.
7176 * Initialization:: General principles of initialization
7177 and termination routines.
7178 * Macros for Initialization::
7179 Specific macros that control the handling of
7180 initialization and termination routines.
7181 * Instruction Output:: Output of actual instructions.
7182 * Dispatch Tables:: Output of jump tables.
7183 * Exception Region Output:: Output of exception region code.
7184 * Alignment Output:: Pseudo ops for alignment and skipping data.
7187 @node File Framework
7188 @subsection The Overall Framework of an Assembler File
7189 @cindex assembler format
7190 @cindex output of assembler code
7192 @c prevent bad page break with this line
7193 This describes the overall framework of an assembly file.
7195 @findex default_file_start
7196 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7197 Output to @code{asm_out_file} any text which the assembler expects to
7198 find at the beginning of a file. The default behavior is controlled
7199 by two flags, documented below. Unless your target's assembler is
7200 quite unusual, if you override the default, you should call
7201 @code{default_file_start} at some point in your target hook. This
7202 lets other target files rely on these variables.
7205 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7206 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7207 printed as the very first line in the assembly file, unless
7208 @option{-fverbose-asm} is in effect. (If that macro has been defined
7209 to the empty string, this variable has no effect.) With the normal
7210 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7211 assembler that it need not bother stripping comments or extra
7212 whitespace from its input. This allows it to work a bit faster.
7214 The default is false. You should not set it to true unless you have
7215 verified that your port does not generate any extra whitespace or
7216 comments that will cause GAS to issue errors in NO_APP mode.
7219 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7220 If this flag is true, @code{output_file_directive} will be called
7221 for the primary source file, immediately after printing
7222 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7223 this to be done. The default is false.
7226 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7227 Output to @code{asm_out_file} any text which the assembler expects
7228 to find at the end of a file. The default is to output nothing.
7231 @deftypefun void file_end_indicate_exec_stack ()
7232 Some systems use a common convention, the @samp{.note.GNU-stack}
7233 special section, to indicate whether or not an object file relies on
7234 the stack being executable. If your system uses this convention, you
7235 should define @code{TARGET_ASM_FILE_END} to this function. If you
7236 need to do other things in that hook, have your hook function call
7240 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7241 Output to @code{asm_out_file} any text which the assembler expects
7242 to find at the start of an LTO section. The default is to output
7246 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7247 Output to @code{asm_out_file} any text which the assembler expects
7248 to find at the end of an LTO section. The default is to output
7252 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7253 Output to @code{asm_out_file} any text which is needed before emitting
7254 unwind info and debug info at the end of a file. Some targets emit
7255 here PIC setup thunks that cannot be emitted at the end of file,
7256 because they couldn't have unwind info then. The default is to output
7260 @defmac ASM_COMMENT_START
7261 A C string constant describing how to begin a comment in the target
7262 assembler language. The compiler assumes that the comment will end at
7263 the end of the line.
7267 A C string constant for text to be output before each @code{asm}
7268 statement or group of consecutive ones. Normally this is
7269 @code{"#APP"}, which is a comment that has no effect on most
7270 assemblers but tells the GNU assembler that it must check the lines
7271 that follow for all valid assembler constructs.
7275 A C string constant for text to be output after each @code{asm}
7276 statement or group of consecutive ones. Normally this is
7277 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7278 time-saving assumptions that are valid for ordinary compiler output.
7281 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7282 A C statement to output COFF information or DWARF debugging information
7283 which indicates that filename @var{name} is the current source file to
7284 the stdio stream @var{stream}.
7286 This macro need not be defined if the standard form of output
7287 for the file format in use is appropriate.
7290 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7291 Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7293 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7296 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7297 A C statement to output the string @var{string} to the stdio stream
7298 @var{stream}. If you do not call the function @code{output_quoted_string}
7299 in your config files, GCC will only call it to output filenames to
7300 the assembler source. So you can use it to canonicalize the format
7301 of the filename using this macro.
7304 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7305 A C statement to output something to the assembler file to handle a
7306 @samp{#ident} directive containing the text @var{string}. If this
7307 macro is not defined, nothing is output for a @samp{#ident} directive.
7310 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7311 Output assembly directives to switch to section @var{name}. The section
7312 should have attributes as specified by @var{flags}, which is a bit mask
7313 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7314 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7315 this section is associated.
7318 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
7319 Return preferred text (sub)section for function @var{decl}.
7320 Main purpose of this function is to separate cold, normal and hot
7321 functions. @var{startup} is true when function is known to be used only
7322 at startup (from static constructors or it is @code{main()}).
7323 @var{exit} is true when function is known to be used only at exit
7324 (from static destructors).
7325 Return NULL if function should go to default text section.
7328 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
7329 Used by the target to emit any assembler directives or additional labels needed when a function is partitioned between different sections. Output should be written to @var{file}. The function decl is available as @var{decl} and the new section is `cold' if @var{new_is_cold} is @code{true}.
7332 @deftypevr {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7333 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7334 It must not be modified by command-line option processing.
7337 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7338 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7339 This flag is true if we can create zeroed data by switching to a BSS
7340 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7341 This is true on most ELF targets.
7344 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7345 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7346 based on a variable or function decl, a section name, and whether or not the
7347 declaration's initializer may contain runtime relocations. @var{decl} may be
7348 null, in which case read-write data should be assumed.
7350 The default version of this function handles choosing code vs data,
7351 read-only vs read-write data, and @code{flag_pic}. You should only
7352 need to override this if your target has special flags that might be
7353 set via @code{__attribute__}.
7356 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
7357 Provides the target with the ability to record the gcc command line
7358 switches that have been passed to the compiler, and options that are
7359 enabled. The @var{type} argument specifies what is being recorded.
7360 It can take the following values:
7363 @item SWITCH_TYPE_PASSED
7364 @var{text} is a command line switch that has been set by the user.
7366 @item SWITCH_TYPE_ENABLED
7367 @var{text} is an option which has been enabled. This might be as a
7368 direct result of a command line switch, or because it is enabled by
7369 default or because it has been enabled as a side effect of a different
7370 command line switch. For example, the @option{-O2} switch enables
7371 various different individual optimization passes.
7373 @item SWITCH_TYPE_DESCRIPTIVE
7374 @var{text} is either NULL or some descriptive text which should be
7375 ignored. If @var{text} is NULL then it is being used to warn the
7376 target hook that either recording is starting or ending. The first
7377 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7378 warning is for start up and the second time the warning is for
7379 wind down. This feature is to allow the target hook to make any
7380 necessary preparations before it starts to record switches and to
7381 perform any necessary tidying up after it has finished recording
7384 @item SWITCH_TYPE_LINE_START
7385 This option can be ignored by this target hook.
7387 @item SWITCH_TYPE_LINE_END
7388 This option can be ignored by this target hook.
7391 The hook's return value must be zero. Other return values may be
7392 supported in the future.
7394 By default this hook is set to NULL, but an example implementation is
7395 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7396 it records the switches as ASCII text inside a new, string mergeable
7397 section in the assembler output file. The name of the new section is
7398 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7402 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7403 This is the name of the section that will be created by the example
7404 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7410 @subsection Output of Data
7413 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7414 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7415 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7416 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7417 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7418 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7419 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7420 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7421 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7422 These hooks specify assembly directives for creating certain kinds
7423 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7424 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7425 aligned two-byte object, and so on. Any of the hooks may be
7426 @code{NULL}, indicating that no suitable directive is available.
7428 The compiler will print these strings at the start of a new line,
7429 followed immediately by the object's initial value. In most cases,
7430 the string should contain a tab, a pseudo-op, and then another tab.
7433 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7434 The @code{assemble_integer} function uses this hook to output an
7435 integer object. @var{x} is the object's value, @var{size} is its size
7436 in bytes and @var{aligned_p} indicates whether it is aligned. The
7437 function should return @code{true} if it was able to output the
7438 object. If it returns false, @code{assemble_integer} will try to
7439 split the object into smaller parts.
7441 The default implementation of this hook will use the
7442 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7443 when the relevant string is @code{NULL}.
7446 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
7447 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7448 can't deal with, and output assembly code to @var{file} corresponding to
7449 the pattern @var{x}. This may be used to allow machine-dependent
7450 @code{UNSPEC}s to appear within constants.
7452 If target hook fails to recognize a pattern, it must return @code{false},
7453 so that a standard error message is printed. If it prints an error message
7454 itself, by calling, for example, @code{output_operand_lossage}, it may just
7458 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
7459 A C statement to recognize @var{rtx} patterns that
7460 @code{output_addr_const} can't deal with, and output assembly code to
7461 @var{stream} corresponding to the pattern @var{x}. This may be used to
7462 allow machine-dependent @code{UNSPEC}s to appear within constants.
7464 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
7465 @code{goto fail}, so that a standard error message is printed. If it
7466 prints an error message itself, by calling, for example,
7467 @code{output_operand_lossage}, it may just complete normally.
7470 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7471 A C statement to output to the stdio stream @var{stream} an assembler
7472 instruction to assemble a string constant containing the @var{len}
7473 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7474 @code{char *} and @var{len} a C expression of type @code{int}.
7476 If the assembler has a @code{.ascii} pseudo-op as found in the
7477 Berkeley Unix assembler, do not define the macro
7478 @code{ASM_OUTPUT_ASCII}.
7481 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7482 A C statement to output word @var{n} of a function descriptor for
7483 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7484 is defined, and is otherwise unused.
7487 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7488 You may define this macro as a C expression. You should define the
7489 expression to have a nonzero value if GCC should output the constant
7490 pool for a function before the code for the function, or a zero value if
7491 GCC should output the constant pool after the function. If you do
7492 not define this macro, the usual case, GCC will output the constant
7493 pool before the function.
7496 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7497 A C statement to output assembler commands to define the start of the
7498 constant pool for a function. @var{funname} is a string giving
7499 the name of the function. Should the return type of the function
7500 be required, it can be obtained via @var{fundecl}. @var{size}
7501 is the size, in bytes, of the constant pool that will be written
7502 immediately after this call.
7504 If no constant-pool prefix is required, the usual case, this macro need
7508 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7509 A C statement (with or without semicolon) to output a constant in the
7510 constant pool, if it needs special treatment. (This macro need not do
7511 anything for RTL expressions that can be output normally.)
7513 The argument @var{file} is the standard I/O stream to output the
7514 assembler code on. @var{x} is the RTL expression for the constant to
7515 output, and @var{mode} is the machine mode (in case @var{x} is a
7516 @samp{const_int}). @var{align} is the required alignment for the value
7517 @var{x}; you should output an assembler directive to force this much
7520 The argument @var{labelno} is a number to use in an internal label for
7521 the address of this pool entry. The definition of this macro is
7522 responsible for outputting the label definition at the proper place.
7523 Here is how to do this:
7526 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7529 When you output a pool entry specially, you should end with a
7530 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7531 entry from being output a second time in the usual manner.
7533 You need not define this macro if it would do nothing.
7536 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7537 A C statement to output assembler commands to at the end of the constant
7538 pool for a function. @var{funname} is a string giving the name of the
7539 function. Should the return type of the function be required, you can
7540 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7541 constant pool that GCC wrote immediately before this call.
7543 If no constant-pool epilogue is required, the usual case, you need not
7547 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7548 Define this macro as a C expression which is nonzero if @var{C} is
7549 used as a logical line separator by the assembler. @var{STR} points
7550 to the position in the string where @var{C} was found; this can be used if
7551 a line separator uses multiple characters.
7553 If you do not define this macro, the default is that only
7554 the character @samp{;} is treated as a logical line separator.
7557 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7558 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7559 These target hooks are C string constants, describing the syntax in the
7560 assembler for grouping arithmetic expressions. If not overridden, they
7561 default to normal parentheses, which is correct for most assemblers.
7564 These macros are provided by @file{real.h} for writing the definitions
7565 of @code{ASM_OUTPUT_DOUBLE} and the like:
7567 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7568 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7569 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7570 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7571 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7572 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7573 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7574 target's floating point representation, and store its bit pattern in
7575 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7576 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7577 simple @code{long int}. For the others, it should be an array of
7578 @code{long int}. The number of elements in this array is determined
7579 by the size of the desired target floating point data type: 32 bits of
7580 it go in each @code{long int} array element. Each array element holds
7581 32 bits of the result, even if @code{long int} is wider than 32 bits
7582 on the host machine.
7584 The array element values are designed so that you can print them out
7585 using @code{fprintf} in the order they should appear in the target
7589 @node Uninitialized Data
7590 @subsection Output of Uninitialized Variables
7592 Each of the macros in this section is used to do the whole job of
7593 outputting a single uninitialized variable.
7595 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7596 A C statement (sans semicolon) to output to the stdio stream
7597 @var{stream} the assembler definition of a common-label named
7598 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7599 is the size rounded up to whatever alignment the caller wants. It is
7600 possible that @var{size} may be zero, for instance if a struct with no
7601 other member than a zero-length array is defined. In this case, the
7602 backend must output a symbol definition that allocates at least one
7603 byte, both so that the address of the resulting object does not compare
7604 equal to any other, and because some object formats cannot even express
7605 the concept of a zero-sized common symbol, as that is how they represent
7606 an ordinary undefined external.
7608 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7609 output the name itself; before and after that, output the additional
7610 assembler syntax for defining the name, and a newline.
7612 This macro controls how the assembler definitions of uninitialized
7613 common global variables are output.
7616 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7617 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7618 separate, explicit argument. If you define this macro, it is used in
7619 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7620 handling the required alignment of the variable. The alignment is specified
7621 as the number of bits.
7624 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7625 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7626 variable to be output, if there is one, or @code{NULL_TREE} if there
7627 is no corresponding variable. If you define this macro, GCC will use it
7628 in place of both @code{ASM_OUTPUT_COMMON} and
7629 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7630 the variable's decl in order to chose what to output.
7633 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7634 A C statement (sans semicolon) to output to the stdio stream
7635 @var{stream} the assembler definition of uninitialized global @var{decl} named
7636 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
7637 is the alignment specified as the number of bits.
7639 Try to use function @code{asm_output_aligned_bss} defined in file
7640 @file{varasm.c} when defining this macro. If unable, use the expression
7641 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7642 before and after that, output the additional assembler syntax for defining
7643 the name, and a newline.
7645 There are two ways of handling global BSS@. One is to define this macro.
7646 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7647 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7648 You do not need to do both.
7650 Some languages do not have @code{common} data, and require a
7651 non-common form of global BSS in order to handle uninitialized globals
7652 efficiently. C++ is one example of this. However, if the target does
7653 not support global BSS, the front end may choose to make globals
7654 common in order to save space in the object file.
7657 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7658 A C statement (sans semicolon) to output to the stdio stream
7659 @var{stream} the assembler definition of a local-common-label named
7660 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7661 is the size rounded up to whatever alignment the caller wants.
7663 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7664 output the name itself; before and after that, output the additional
7665 assembler syntax for defining the name, and a newline.
7667 This macro controls how the assembler definitions of uninitialized
7668 static variables are output.
7671 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7672 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7673 separate, explicit argument. If you define this macro, it is used in
7674 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7675 handling the required alignment of the variable. The alignment is specified
7676 as the number of bits.
7679 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7680 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7681 variable to be output, if there is one, or @code{NULL_TREE} if there
7682 is no corresponding variable. If you define this macro, GCC will use it
7683 in place of both @code{ASM_OUTPUT_DECL} and
7684 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7685 the variable's decl in order to chose what to output.
7689 @subsection Output and Generation of Labels
7691 @c prevent bad page break with this line
7692 This is about outputting labels.
7694 @findex assemble_name
7695 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7696 A C statement (sans semicolon) to output to the stdio stream
7697 @var{stream} the assembler definition of a label named @var{name}.
7698 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7699 output the name itself; before and after that, output the additional
7700 assembler syntax for defining the name, and a newline. A default
7701 definition of this macro is provided which is correct for most systems.
7704 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7705 A C statement (sans semicolon) to output to the stdio stream
7706 @var{stream} the assembler definition of a label named @var{name} of
7708 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7709 output the name itself; before and after that, output the additional
7710 assembler syntax for defining the name, and a newline. A default
7711 definition of this macro is provided which is correct for most systems.
7713 If this macro is not defined, then the function name is defined in the
7714 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7717 @findex assemble_name_raw
7718 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7719 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7720 to refer to a compiler-generated label. The default definition uses
7721 @code{assemble_name_raw}, which is like @code{assemble_name} except
7722 that it is more efficient.
7726 A C string containing the appropriate assembler directive to specify the
7727 size of a symbol, without any arguments. On systems that use ELF, the
7728 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7729 systems, the default is not to define this macro.
7731 Define this macro only if it is correct to use the default definitions
7732 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7733 for your system. If you need your own custom definitions of those
7734 macros, or if you do not need explicit symbol sizes at all, do not
7738 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7739 A C statement (sans semicolon) to output to the stdio stream
7740 @var{stream} a directive telling the assembler that the size of the
7741 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7742 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7746 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7747 A C statement (sans semicolon) to output to the stdio stream
7748 @var{stream} a directive telling the assembler to calculate the size of
7749 the symbol @var{name} by subtracting its address from the current
7752 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7753 provided. The default assumes that the assembler recognizes a special
7754 @samp{.} symbol as referring to the current address, and can calculate
7755 the difference between this and another symbol. If your assembler does
7756 not recognize @samp{.} or cannot do calculations with it, you will need
7757 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7761 A C string containing the appropriate assembler directive to specify the
7762 type of a symbol, without any arguments. On systems that use ELF, the
7763 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7764 systems, the default is not to define this macro.
7766 Define this macro only if it is correct to use the default definition of
7767 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7768 custom definition of this macro, or if you do not need explicit symbol
7769 types at all, do not define this macro.
7772 @defmac TYPE_OPERAND_FMT
7773 A C string which specifies (using @code{printf} syntax) the format of
7774 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7775 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7776 the default is not to define this macro.
7778 Define this macro only if it is correct to use the default definition of
7779 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7780 custom definition of this macro, or if you do not need explicit symbol
7781 types at all, do not define this macro.
7784 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7785 A C statement (sans semicolon) to output to the stdio stream
7786 @var{stream} a directive telling the assembler that the type of the
7787 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7788 that string is always either @samp{"function"} or @samp{"object"}, but
7789 you should not count on this.
7791 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7792 definition of this macro is provided.
7795 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7796 A C statement (sans semicolon) to output to the stdio stream
7797 @var{stream} any text necessary for declaring the name @var{name} of a
7798 function which is being defined. This macro is responsible for
7799 outputting the label definition (perhaps using
7800 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7801 @code{FUNCTION_DECL} tree node representing the function.
7803 If this macro is not defined, then the function name is defined in the
7804 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7806 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7810 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7811 A C statement (sans semicolon) to output to the stdio stream
7812 @var{stream} any text necessary for declaring the size of a function
7813 which is being defined. The argument @var{name} is the name of the
7814 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7815 representing the function.
7817 If this macro is not defined, then the function size is not defined.
7819 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7823 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7824 A C statement (sans semicolon) to output to the stdio stream
7825 @var{stream} any text necessary for declaring the name @var{name} of an
7826 initialized variable which is being defined. This macro must output the
7827 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7828 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7830 If this macro is not defined, then the variable name is defined in the
7831 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7833 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7834 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7837 @deftypefn {Target Hook} void TARGET_ASM_DECLARE_CONSTANT_NAME (FILE *@var{file}, const char *@var{name}, const_tree @var{expr}, HOST_WIDE_INT @var{size})
7838 A target hook to output to the stdio stream @var{file} any text necessary
7839 for declaring the name @var{name} of a constant which is being defined. This
7840 target hook is responsible for outputting the label definition (perhaps using
7841 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7842 and @var{size} is the size of the constant in bytes. The @var{name}
7843 will be an internal label.
7845 The default version of this target hook, define the @var{name} in the
7846 usual manner as a label (by means of @code{assemble_label}).
7848 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7851 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7852 A C statement (sans semicolon) to output to the stdio stream
7853 @var{stream} any text necessary for claiming a register @var{regno}
7854 for a global variable @var{decl} with name @var{name}.
7856 If you don't define this macro, that is equivalent to defining it to do
7860 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7861 A C statement (sans semicolon) to finish up declaring a variable name
7862 once the compiler has processed its initializer fully and thus has had a
7863 chance to determine the size of an array when controlled by an
7864 initializer. This is used on systems where it's necessary to declare
7865 something about the size of the object.
7867 If you don't define this macro, that is equivalent to defining it to do
7870 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7871 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7874 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7875 This target hook is a function to output to the stdio stream
7876 @var{stream} some commands that will make the label @var{name} global;
7877 that is, available for reference from other files.
7879 The default implementation relies on a proper definition of
7880 @code{GLOBAL_ASM_OP}.
7883 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7884 This target hook is a function to output to the stdio stream
7885 @var{stream} some commands that will make the name associated with @var{decl}
7886 global; that is, available for reference from other files.
7888 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7891 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7892 A C statement (sans semicolon) to output to the stdio stream
7893 @var{stream} some commands that will make the label @var{name} weak;
7894 that is, available for reference from other files but only used if
7895 no other definition is available. Use the expression
7896 @code{assemble_name (@var{stream}, @var{name})} to output the name
7897 itself; before and after that, output the additional assembler syntax
7898 for making that name weak, and a newline.
7900 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7901 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7905 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7906 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7907 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7908 or variable decl. If @var{value} is not @code{NULL}, this C statement
7909 should output to the stdio stream @var{stream} assembler code which
7910 defines (equates) the weak symbol @var{name} to have the value
7911 @var{value}. If @var{value} is @code{NULL}, it should output commands
7912 to make @var{name} weak.
7915 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7916 Outputs a directive that enables @var{name} to be used to refer to
7917 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7918 declaration of @code{name}.
7921 @defmac SUPPORTS_WEAK
7922 A preprocessor constant expression which evaluates to true if the target
7923 supports weak symbols.
7925 If you don't define this macro, @file{defaults.h} provides a default
7926 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7927 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
7930 @defmac TARGET_SUPPORTS_WEAK
7931 A C expression which evaluates to true if the target supports weak symbols.
7933 If you don't define this macro, @file{defaults.h} provides a default
7934 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
7935 this macro if you want to control weak symbol support with a compiler
7936 flag such as @option{-melf}.
7939 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7940 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7941 public symbol such that extra copies in multiple translation units will
7942 be discarded by the linker. Define this macro if your object file
7943 format provides support for this concept, such as the @samp{COMDAT}
7944 section flags in the Microsoft Windows PE/COFF format, and this support
7945 requires changes to @var{decl}, such as putting it in a separate section.
7948 @defmac SUPPORTS_ONE_ONLY
7949 A C expression which evaluates to true if the target supports one-only
7952 If you don't define this macro, @file{varasm.c} provides a default
7953 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7954 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7955 you want to control one-only symbol support with a compiler flag, or if
7956 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7957 be emitted as one-only.
7960 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
7961 This target hook is a function to output to @var{asm_out_file} some
7962 commands that will make the symbol(s) associated with @var{decl} have
7963 hidden, protected or internal visibility as specified by @var{visibility}.
7966 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7967 A C expression that evaluates to true if the target's linker expects
7968 that weak symbols do not appear in a static archive's table of contents.
7969 The default is @code{0}.
7971 Leaving weak symbols out of an archive's table of contents means that,
7972 if a symbol will only have a definition in one translation unit and
7973 will have undefined references from other translation units, that
7974 symbol should not be weak. Defining this macro to be nonzero will
7975 thus have the effect that certain symbols that would normally be weak
7976 (explicit template instantiations, and vtables for polymorphic classes
7977 with noninline key methods) will instead be nonweak.
7979 The C++ ABI requires this macro to be zero. Define this macro for
7980 targets where full C++ ABI compliance is impossible and where linker
7981 restrictions require weak symbols to be left out of a static archive's
7985 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7986 A C statement (sans semicolon) to output to the stdio stream
7987 @var{stream} any text necessary for declaring the name of an external
7988 symbol named @var{name} which is referenced in this compilation but
7989 not defined. The value of @var{decl} is the tree node for the
7992 This macro need not be defined if it does not need to output anything.
7993 The GNU assembler and most Unix assemblers don't require anything.
7996 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
7997 This target hook is a function to output to @var{asm_out_file} an assembler
7998 pseudo-op to declare a library function name external. The name of the
7999 library function is given by @var{symref}, which is a @code{symbol_ref}.
8002 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8003 This target hook is a function to output to @var{asm_out_file} an assembler
8004 directive to annotate @var{symbol} as used. The Darwin target uses the
8005 .no_dead_code_strip directive.
8008 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8009 A C statement (sans semicolon) to output to the stdio stream
8010 @var{stream} a reference in assembler syntax to a label named
8011 @var{name}. This should add @samp{_} to the front of the name, if that
8012 is customary on your operating system, as it is in most Berkeley Unix
8013 systems. This macro is used in @code{assemble_name}.
8016 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8017 Given a symbol @var{name}, perform same mangling as @code{varasm.c}'s @code{assemble_name}, but in memory rather than to a file stream, returning result as an @code{IDENTIFIER_NODE}. Required for correct LTO symtabs. The default implementation calls the @code{TARGET_STRIP_NAME_ENCODING} hook and then prepends the @code{USER_LABEL_PREFIX}, if any.
8020 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8021 A C statement (sans semicolon) to output a reference to
8022 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8023 will be used to output the name of the symbol. This macro may be used
8024 to modify the way a symbol is referenced depending on information
8025 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8028 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8029 A C statement (sans semicolon) to output a reference to @var{buf}, the
8030 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8031 @code{assemble_name} will be used to output the name of the symbol.
8032 This macro is not used by @code{output_asm_label}, or the @code{%l}
8033 specifier that calls it; the intention is that this macro should be set
8034 when it is necessary to output a label differently when its address is
8038 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8039 A function to output to the stdio stream @var{stream} a label whose
8040 name is made from the string @var{prefix} and the number @var{labelno}.
8042 It is absolutely essential that these labels be distinct from the labels
8043 used for user-level functions and variables. Otherwise, certain programs
8044 will have name conflicts with internal labels.
8046 It is desirable to exclude internal labels from the symbol table of the
8047 object file. Most assemblers have a naming convention for labels that
8048 should be excluded; on many systems, the letter @samp{L} at the
8049 beginning of a label has this effect. You should find out what
8050 convention your system uses, and follow it.
8052 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8055 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8056 A C statement to output to the stdio stream @var{stream} a debug info
8057 label whose name is made from the string @var{prefix} and the number
8058 @var{num}. This is useful for VLIW targets, where debug info labels
8059 may need to be treated differently than branch target labels. On some
8060 systems, branch target labels must be at the beginning of instruction
8061 bundles, but debug info labels can occur in the middle of instruction
8064 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8068 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8069 A C statement to store into the string @var{string} a label whose name
8070 is made from the string @var{prefix} and the number @var{num}.
8072 This string, when output subsequently by @code{assemble_name}, should
8073 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8074 with the same @var{prefix} and @var{num}.
8076 If the string begins with @samp{*}, then @code{assemble_name} will
8077 output the rest of the string unchanged. It is often convenient for
8078 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8079 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8080 to output the string, and may change it. (Of course,
8081 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8082 you should know what it does on your machine.)
8085 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8086 A C expression to assign to @var{outvar} (which is a variable of type
8087 @code{char *}) a newly allocated string made from the string
8088 @var{name} and the number @var{number}, with some suitable punctuation
8089 added. Use @code{alloca} to get space for the string.
8091 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8092 produce an assembler label for an internal static variable whose name is
8093 @var{name}. Therefore, the string must be such as to result in valid
8094 assembler code. The argument @var{number} is different each time this
8095 macro is executed; it prevents conflicts between similarly-named
8096 internal static variables in different scopes.
8098 Ideally this string should not be a valid C identifier, to prevent any
8099 conflict with the user's own symbols. Most assemblers allow periods
8100 or percent signs in assembler symbols; putting at least one of these
8101 between the name and the number will suffice.
8103 If this macro is not defined, a default definition will be provided
8104 which is correct for most systems.
8107 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8108 A C statement to output to the stdio stream @var{stream} assembler code
8109 which defines (equates) the symbol @var{name} to have the value @var{value}.
8112 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8113 correct for most systems.
8116 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8117 A C statement to output to the stdio stream @var{stream} assembler code
8118 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8119 to have the value of the tree node @var{decl_of_value}. This macro will
8120 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8121 the tree nodes are available.
8124 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8125 correct for most systems.
8128 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8129 A C statement that evaluates to true if the assembler code which defines
8130 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8131 of the tree node @var{decl_of_value} should be emitted near the end of the
8132 current compilation unit. The default is to not defer output of defines.
8133 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8134 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8137 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8138 A C statement to output to the stdio stream @var{stream} assembler code
8139 which defines (equates) the weak symbol @var{name} to have the value
8140 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8141 an undefined weak symbol.
8143 Define this macro if the target only supports weak aliases; define
8144 @code{ASM_OUTPUT_DEF} instead if possible.
8147 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8148 Define this macro to override the default assembler names used for
8149 Objective-C methods.
8151 The default name is a unique method number followed by the name of the
8152 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8153 the category is also included in the assembler name (e.g.@:
8156 These names are safe on most systems, but make debugging difficult since
8157 the method's selector is not present in the name. Therefore, particular
8158 systems define other ways of computing names.
8160 @var{buf} is an expression of type @code{char *} which gives you a
8161 buffer in which to store the name; its length is as long as
8162 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8163 50 characters extra.
8165 The argument @var{is_inst} specifies whether the method is an instance
8166 method or a class method; @var{class_name} is the name of the class;
8167 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8168 in a category); and @var{sel_name} is the name of the selector.
8170 On systems where the assembler can handle quoted names, you can use this
8171 macro to provide more human-readable names.
8174 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
8175 A C statement (sans semicolon) to output to the stdio stream
8176 @var{stream} commands to declare that the label @var{name} is an
8177 Objective-C class reference. This is only needed for targets whose
8178 linkers have special support for NeXT-style runtimes.
8181 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
8182 A C statement (sans semicolon) to output to the stdio stream
8183 @var{stream} commands to declare that the label @var{name} is an
8184 unresolved Objective-C class reference. This is only needed for targets
8185 whose linkers have special support for NeXT-style runtimes.
8188 @node Initialization
8189 @subsection How Initialization Functions Are Handled
8190 @cindex initialization routines
8191 @cindex termination routines
8192 @cindex constructors, output of
8193 @cindex destructors, output of
8195 The compiled code for certain languages includes @dfn{constructors}
8196 (also called @dfn{initialization routines})---functions to initialize
8197 data in the program when the program is started. These functions need
8198 to be called before the program is ``started''---that is to say, before
8199 @code{main} is called.
8201 Compiling some languages generates @dfn{destructors} (also called
8202 @dfn{termination routines}) that should be called when the program
8205 To make the initialization and termination functions work, the compiler
8206 must output something in the assembler code to cause those functions to
8207 be called at the appropriate time. When you port the compiler to a new
8208 system, you need to specify how to do this.
8210 There are two major ways that GCC currently supports the execution of
8211 initialization and termination functions. Each way has two variants.
8212 Much of the structure is common to all four variations.
8214 @findex __CTOR_LIST__
8215 @findex __DTOR_LIST__
8216 The linker must build two lists of these functions---a list of
8217 initialization functions, called @code{__CTOR_LIST__}, and a list of
8218 termination functions, called @code{__DTOR_LIST__}.
8220 Each list always begins with an ignored function pointer (which may hold
8221 0, @minus{}1, or a count of the function pointers after it, depending on
8222 the environment). This is followed by a series of zero or more function
8223 pointers to constructors (or destructors), followed by a function
8224 pointer containing zero.
8226 Depending on the operating system and its executable file format, either
8227 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8228 time and exit time. Constructors are called in reverse order of the
8229 list; destructors in forward order.
8231 The best way to handle static constructors works only for object file
8232 formats which provide arbitrarily-named sections. A section is set
8233 aside for a list of constructors, and another for a list of destructors.
8234 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8235 object file that defines an initialization function also puts a word in
8236 the constructor section to point to that function. The linker
8237 accumulates all these words into one contiguous @samp{.ctors} section.
8238 Termination functions are handled similarly.
8240 This method will be chosen as the default by @file{target-def.h} if
8241 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8242 support arbitrary sections, but does support special designated
8243 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8244 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8246 When arbitrary sections are available, there are two variants, depending
8247 upon how the code in @file{crtstuff.c} is called. On systems that
8248 support a @dfn{.init} section which is executed at program startup,
8249 parts of @file{crtstuff.c} are compiled into that section. The
8250 program is linked by the @command{gcc} driver like this:
8253 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8256 The prologue of a function (@code{__init}) appears in the @code{.init}
8257 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8258 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8259 files are provided by the operating system or by the GNU C library, but
8260 are provided by GCC for a few targets.
8262 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8263 compiled from @file{crtstuff.c}. They contain, among other things, code
8264 fragments within the @code{.init} and @code{.fini} sections that branch
8265 to routines in the @code{.text} section. The linker will pull all parts
8266 of a section together, which results in a complete @code{__init} function
8267 that invokes the routines we need at startup.
8269 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8272 If no init section is available, when GCC compiles any function called
8273 @code{main} (or more accurately, any function designated as a program
8274 entry point by the language front end calling @code{expand_main_function}),
8275 it inserts a procedure call to @code{__main} as the first executable code
8276 after the function prologue. The @code{__main} function is defined
8277 in @file{libgcc2.c} and runs the global constructors.
8279 In file formats that don't support arbitrary sections, there are again
8280 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8281 and an `a.out' format must be used. In this case,
8282 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8283 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8284 and with the address of the void function containing the initialization
8285 code as its value. The GNU linker recognizes this as a request to add
8286 the value to a @dfn{set}; the values are accumulated, and are eventually
8287 placed in the executable as a vector in the format described above, with
8288 a leading (ignored) count and a trailing zero element.
8289 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8290 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8291 the compilation of @code{main} to call @code{__main} as above, starting
8292 the initialization process.
8294 The last variant uses neither arbitrary sections nor the GNU linker.
8295 This is preferable when you want to do dynamic linking and when using
8296 file formats which the GNU linker does not support, such as `ECOFF'@. In
8297 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8298 termination functions are recognized simply by their names. This requires
8299 an extra program in the linkage step, called @command{collect2}. This program
8300 pretends to be the linker, for use with GCC; it does its job by running
8301 the ordinary linker, but also arranges to include the vectors of
8302 initialization and termination functions. These functions are called
8303 via @code{__main} as described above. In order to use this method,
8304 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8307 The following section describes the specific macros that control and
8308 customize the handling of initialization and termination functions.
8311 @node Macros for Initialization
8312 @subsection Macros Controlling Initialization Routines
8314 Here are the macros that control how the compiler handles initialization
8315 and termination functions:
8317 @defmac INIT_SECTION_ASM_OP
8318 If defined, a C string constant, including spacing, for the assembler
8319 operation to identify the following data as initialization code. If not
8320 defined, GCC will assume such a section does not exist. When you are
8321 using special sections for initialization and termination functions, this
8322 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8323 run the initialization functions.
8326 @defmac HAS_INIT_SECTION
8327 If defined, @code{main} will not call @code{__main} as described above.
8328 This macro should be defined for systems that control start-up code
8329 on a symbol-by-symbol basis, such as OSF/1, and should not
8330 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8333 @defmac LD_INIT_SWITCH
8334 If defined, a C string constant for a switch that tells the linker that
8335 the following symbol is an initialization routine.
8338 @defmac LD_FINI_SWITCH
8339 If defined, a C string constant for a switch that tells the linker that
8340 the following symbol is a finalization routine.
8343 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8344 If defined, a C statement that will write a function that can be
8345 automatically called when a shared library is loaded. The function
8346 should call @var{func}, which takes no arguments. If not defined, and
8347 the object format requires an explicit initialization function, then a
8348 function called @code{_GLOBAL__DI} will be generated.
8350 This function and the following one are used by collect2 when linking a
8351 shared library that needs constructors or destructors, or has DWARF2
8352 exception tables embedded in the code.
8355 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8356 If defined, a C statement that will write a function that can be
8357 automatically called when a shared library is unloaded. The function
8358 should call @var{func}, which takes no arguments. If not defined, and
8359 the object format requires an explicit finalization function, then a
8360 function called @code{_GLOBAL__DD} will be generated.
8363 @defmac INVOKE__main
8364 If defined, @code{main} will call @code{__main} despite the presence of
8365 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8366 where the init section is not actually run automatically, but is still
8367 useful for collecting the lists of constructors and destructors.
8370 @defmac SUPPORTS_INIT_PRIORITY
8371 If nonzero, the C++ @code{init_priority} attribute is supported and the
8372 compiler should emit instructions to control the order of initialization
8373 of objects. If zero, the compiler will issue an error message upon
8374 encountering an @code{init_priority} attribute.
8377 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8378 This value is true if the target supports some ``native'' method of
8379 collecting constructors and destructors to be run at startup and exit.
8380 It is false if we must use @command{collect2}.
8383 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8384 If defined, a function that outputs assembler code to arrange to call
8385 the function referenced by @var{symbol} at initialization time.
8387 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8388 no arguments and with no return value. If the target supports initialization
8389 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8390 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8392 If this macro is not defined by the target, a suitable default will
8393 be chosen if (1) the target supports arbitrary section names, (2) the
8394 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8398 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8399 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8400 functions rather than initialization functions.
8403 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8404 generated for the generated object file will have static linkage.
8406 If your system uses @command{collect2} as the means of processing
8407 constructors, then that program normally uses @command{nm} to scan
8408 an object file for constructor functions to be called.
8410 On certain kinds of systems, you can define this macro to make
8411 @command{collect2} work faster (and, in some cases, make it work at all):
8413 @defmac OBJECT_FORMAT_COFF
8414 Define this macro if the system uses COFF (Common Object File Format)
8415 object files, so that @command{collect2} can assume this format and scan
8416 object files directly for dynamic constructor/destructor functions.
8418 This macro is effective only in a native compiler; @command{collect2} as
8419 part of a cross compiler always uses @command{nm} for the target machine.
8422 @defmac REAL_NM_FILE_NAME
8423 Define this macro as a C string constant containing the file name to use
8424 to execute @command{nm}. The default is to search the path normally for
8429 @command{collect2} calls @command{nm} to scan object files for static
8430 constructors and destructors and LTO info. By default, @option{-n} is
8431 passed. Define @code{NM_FLAGS} to a C string constant if other options
8432 are needed to get the same output format as GNU @command{nm -n}
8436 If your system supports shared libraries and has a program to list the
8437 dynamic dependencies of a given library or executable, you can define
8438 these macros to enable support for running initialization and
8439 termination functions in shared libraries:
8442 Define this macro to a C string constant containing the name of the program
8443 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8446 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8447 Define this macro to be C code that extracts filenames from the output
8448 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8449 of type @code{char *} that points to the beginning of a line of output
8450 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8451 code must advance @var{ptr} to the beginning of the filename on that
8452 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8455 @defmac SHLIB_SUFFIX
8456 Define this macro to a C string constant containing the default shared
8457 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8458 strips version information after this suffix when generating global
8459 constructor and destructor names. This define is only needed on targets
8460 that use @command{collect2} to process constructors and destructors.
8463 @node Instruction Output
8464 @subsection Output of Assembler Instructions
8466 @c prevent bad page break with this line
8467 This describes assembler instruction output.
8469 @defmac REGISTER_NAMES
8470 A C initializer containing the assembler's names for the machine
8471 registers, each one as a C string constant. This is what translates
8472 register numbers in the compiler into assembler language.
8475 @defmac ADDITIONAL_REGISTER_NAMES
8476 If defined, a C initializer for an array of structures containing a name
8477 and a register number. This macro defines additional names for hard
8478 registers, thus allowing the @code{asm} option in declarations to refer
8479 to registers using alternate names.
8482 @defmac OVERLAPPING_REGISTER_NAMES
8483 If defined, a C initializer for an array of structures containing a
8484 name, a register number and a count of the number of consecutive
8485 machine registers the name overlaps. This macro defines additional
8486 names for hard registers, thus allowing the @code{asm} option in
8487 declarations to refer to registers using alternate names. Unlike
8488 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8489 register name implies multiple underlying registers.
8491 This macro should be used when it is important that a clobber in an
8492 @code{asm} statement clobbers all the underlying values implied by the
8493 register name. For example, on ARM, clobbering the double-precision
8494 VFP register ``d0'' implies clobbering both single-precision registers
8498 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8499 Define this macro if you are using an unusual assembler that
8500 requires different names for the machine instructions.
8502 The definition is a C statement or statements which output an
8503 assembler instruction opcode to the stdio stream @var{stream}. The
8504 macro-operand @var{ptr} is a variable of type @code{char *} which
8505 points to the opcode name in its ``internal'' form---the form that is
8506 written in the machine description. The definition should output the
8507 opcode name to @var{stream}, performing any translation you desire, and
8508 increment the variable @var{ptr} to point at the end of the opcode
8509 so that it will not be output twice.
8511 In fact, your macro definition may process less than the entire opcode
8512 name, or more than the opcode name; but if you want to process text
8513 that includes @samp{%}-sequences to substitute operands, you must take
8514 care of the substitution yourself. Just be sure to increment
8515 @var{ptr} over whatever text should not be output normally.
8517 @findex recog_data.operand
8518 If you need to look at the operand values, they can be found as the
8519 elements of @code{recog_data.operand}.
8521 If the macro definition does nothing, the instruction is output
8525 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8526 If defined, a C statement to be executed just prior to the output of
8527 assembler code for @var{insn}, to modify the extracted operands so
8528 they will be output differently.
8530 Here the argument @var{opvec} is the vector containing the operands
8531 extracted from @var{insn}, and @var{noperands} is the number of
8532 elements of the vector which contain meaningful data for this insn.
8533 The contents of this vector are what will be used to convert the insn
8534 template into assembler code, so you can change the assembler output
8535 by changing the contents of the vector.
8537 This macro is useful when various assembler syntaxes share a single
8538 file of instruction patterns; by defining this macro differently, you
8539 can cause a large class of instructions to be output differently (such
8540 as with rearranged operands). Naturally, variations in assembler
8541 syntax affecting individual insn patterns ought to be handled by
8542 writing conditional output routines in those patterns.
8544 If this macro is not defined, it is equivalent to a null statement.
8547 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx @var{insn}, rtx *@var{opvec}, int @var{noperands})
8548 If defined, this target hook is a function which is executed just after the
8549 output of assembler code for @var{insn}, to change the mode of the assembler
8552 Here the argument @var{opvec} is the vector containing the operands
8553 extracted from @var{insn}, and @var{noperands} is the number of
8554 elements of the vector which contain meaningful data for this insn.
8555 The contents of this vector are what was used to convert the insn
8556 template into assembler code, so you can change the assembler mode
8557 by checking the contents of the vector.
8560 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8561 A C compound statement to output to stdio stream @var{stream} the
8562 assembler syntax for an instruction operand @var{x}. @var{x} is an
8565 @var{code} is a value that can be used to specify one of several ways
8566 of printing the operand. It is used when identical operands must be
8567 printed differently depending on the context. @var{code} comes from
8568 the @samp{%} specification that was used to request printing of the
8569 operand. If the specification was just @samp{%@var{digit}} then
8570 @var{code} is 0; if the specification was @samp{%@var{ltr}
8571 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8574 If @var{x} is a register, this macro should print the register's name.
8575 The names can be found in an array @code{reg_names} whose type is
8576 @code{char *[]}. @code{reg_names} is initialized from
8577 @code{REGISTER_NAMES}.
8579 When the machine description has a specification @samp{%@var{punct}}
8580 (a @samp{%} followed by a punctuation character), this macro is called
8581 with a null pointer for @var{x} and the punctuation character for
8585 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8586 A C expression which evaluates to true if @var{code} is a valid
8587 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8588 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8589 punctuation characters (except for the standard one, @samp{%}) are used
8593 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8594 A C compound statement to output to stdio stream @var{stream} the
8595 assembler syntax for an instruction operand that is a memory reference
8596 whose address is @var{x}. @var{x} is an RTL expression.
8598 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8599 On some machines, the syntax for a symbolic address depends on the
8600 section that the address refers to. On these machines, define the hook
8601 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8602 @code{symbol_ref}, and then check for it here. @xref{Assembler
8606 @findex dbr_sequence_length
8607 @defmac DBR_OUTPUT_SEQEND (@var{file})
8608 A C statement, to be executed after all slot-filler instructions have
8609 been output. If necessary, call @code{dbr_sequence_length} to
8610 determine the number of slots filled in a sequence (zero if not
8611 currently outputting a sequence), to decide how many no-ops to output,
8614 Don't define this macro if it has nothing to do, but it is helpful in
8615 reading assembly output if the extent of the delay sequence is made
8616 explicit (e.g.@: with white space).
8619 @findex final_sequence
8620 Note that output routines for instructions with delay slots must be
8621 prepared to deal with not being output as part of a sequence
8622 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8623 found.) The variable @code{final_sequence} is null when not
8624 processing a sequence, otherwise it contains the @code{sequence} rtx
8628 @defmac REGISTER_PREFIX
8629 @defmacx LOCAL_LABEL_PREFIX
8630 @defmacx USER_LABEL_PREFIX
8631 @defmacx IMMEDIATE_PREFIX
8632 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8633 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8634 @file{final.c}). These are useful when a single @file{md} file must
8635 support multiple assembler formats. In that case, the various @file{tm.h}
8636 files can define these macros differently.
8639 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8640 If defined this macro should expand to a series of @code{case}
8641 statements which will be parsed inside the @code{switch} statement of
8642 the @code{asm_fprintf} function. This allows targets to define extra
8643 printf formats which may useful when generating their assembler
8644 statements. Note that uppercase letters are reserved for future
8645 generic extensions to asm_fprintf, and so are not available to target
8646 specific code. The output file is given by the parameter @var{file}.
8647 The varargs input pointer is @var{argptr} and the rest of the format
8648 string, starting the character after the one that is being switched
8649 upon, is pointed to by @var{format}.
8652 @defmac ASSEMBLER_DIALECT
8653 If your target supports multiple dialects of assembler language (such as
8654 different opcodes), define this macro as a C expression that gives the
8655 numeric index of the assembler language dialect to use, with zero as the
8658 If this macro is defined, you may use constructs of the form
8660 @samp{@{option0|option1|option2@dots{}@}}
8663 in the output templates of patterns (@pxref{Output Template}) or in the
8664 first argument of @code{asm_fprintf}. This construct outputs
8665 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8666 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8667 within these strings retain their usual meaning. If there are fewer
8668 alternatives within the braces than the value of
8669 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8671 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8672 @samp{@}} do not have any special meaning when used in templates or
8673 operands to @code{asm_fprintf}.
8675 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8676 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8677 the variations in assembler language syntax with that mechanism. Define
8678 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8679 if the syntax variant are larger and involve such things as different
8680 opcodes or operand order.
8683 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8684 A C expression to output to @var{stream} some assembler code
8685 which will push hard register number @var{regno} onto the stack.
8686 The code need not be optimal, since this macro is used only when
8690 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8691 A C expression to output to @var{stream} some assembler code
8692 which will pop hard register number @var{regno} off of the stack.
8693 The code need not be optimal, since this macro is used only when
8697 @node Dispatch Tables
8698 @subsection Output of Dispatch Tables
8700 @c prevent bad page break with this line
8701 This concerns dispatch tables.
8703 @cindex dispatch table
8704 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8705 A C statement to output to the stdio stream @var{stream} an assembler
8706 pseudo-instruction to generate a difference between two labels.
8707 @var{value} and @var{rel} are the numbers of two internal labels. The
8708 definitions of these labels are output using
8709 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8710 way here. For example,
8713 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8714 @var{value}, @var{rel})
8717 You must provide this macro on machines where the addresses in a
8718 dispatch table are relative to the table's own address. If defined, GCC
8719 will also use this macro on all machines when producing PIC@.
8720 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8721 mode and flags can be read.
8724 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8725 This macro should be provided on machines where the addresses
8726 in a dispatch table are absolute.
8728 The definition should be a C statement to output to the stdio stream
8729 @var{stream} an assembler pseudo-instruction to generate a reference to
8730 a label. @var{value} is the number of an internal label whose
8731 definition is output using @code{(*targetm.asm_out.internal_label)}.
8735 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8739 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8740 Define this if the label before a jump-table needs to be output
8741 specially. The first three arguments are the same as for
8742 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8743 jump-table which follows (a @code{jump_insn} containing an
8744 @code{addr_vec} or @code{addr_diff_vec}).
8746 This feature is used on system V to output a @code{swbeg} statement
8749 If this macro is not defined, these labels are output with
8750 @code{(*targetm.asm_out.internal_label)}.
8753 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8754 Define this if something special must be output at the end of a
8755 jump-table. The definition should be a C statement to be executed
8756 after the assembler code for the table is written. It should write
8757 the appropriate code to stdio stream @var{stream}. The argument
8758 @var{table} is the jump-table insn, and @var{num} is the label-number
8759 of the preceding label.
8761 If this macro is not defined, nothing special is output at the end of
8765 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
8766 This target hook emits a label at the beginning of each FDE@. It
8767 should be defined on targets where FDEs need special labels, and it
8768 should write the appropriate label, for the FDE associated with the
8769 function declaration @var{decl}, to the stdio stream @var{stream}.
8770 The third argument, @var{for_eh}, is a boolean: true if this is for an
8771 exception table. The fourth argument, @var{empty}, is a boolean:
8772 true if this is a placeholder label for an omitted FDE@.
8774 The default is that FDEs are not given nonlocal labels.
8777 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
8778 This target hook emits a label at the beginning of the exception table.
8779 It should be defined on targets where it is desirable for the table
8780 to be broken up according to function.
8782 The default is that no label is emitted.
8785 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
8786 If the target implements @code{TARGET_ASM_UNWIND_EMIT}, this hook may be used to emit a directive to install a personality hook into the unwind info. This hook should not be used if dwarf2 unwind info is used.
8789 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx @var{insn})
8790 This target hook emits assembly directives required to unwind the
8791 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8792 returns @code{UI_TARGET}.
8795 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8796 True if the @code{TARGET_ASM_UNWIND_EMIT} hook should be called before the assembly for @var{insn} has been emitted, false if the hook should be called afterward.
8799 @node Exception Region Output
8800 @subsection Assembler Commands for Exception Regions
8802 @c prevent bad page break with this line
8804 This describes commands marking the start and the end of an exception
8807 @defmac EH_FRAME_SECTION_NAME
8808 If defined, a C string constant for the name of the section containing
8809 exception handling frame unwind information. If not defined, GCC will
8810 provide a default definition if the target supports named sections.
8811 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8813 You should define this symbol if your target supports DWARF 2 frame
8814 unwind information and the default definition does not work.
8817 @defmac EH_FRAME_IN_DATA_SECTION
8818 If defined, DWARF 2 frame unwind information will be placed in the
8819 data section even though the target supports named sections. This
8820 might be necessary, for instance, if the system linker does garbage
8821 collection and sections cannot be marked as not to be collected.
8823 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8827 @defmac EH_TABLES_CAN_BE_READ_ONLY
8828 Define this macro to 1 if your target is such that no frame unwind
8829 information encoding used with non-PIC code will ever require a
8830 runtime relocation, but the linker may not support merging read-only
8831 and read-write sections into a single read-write section.
8834 @defmac MASK_RETURN_ADDR
8835 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8836 that it does not contain any extraneous set bits in it.
8839 @defmac DWARF2_UNWIND_INFO
8840 Define this macro to 0 if your target supports DWARF 2 frame unwind
8841 information, but it does not yet work with exception handling.
8842 Otherwise, if your target supports this information (if it defines
8843 @code{INCOMING_RETURN_ADDR_RTX} and either @code{UNALIGNED_INT_ASM_OP}
8844 or @code{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8847 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
8848 This hook defines the mechanism that will be used for exception handling
8849 by the target. If the target has ABI specified unwind tables, the hook
8850 should return @code{UI_TARGET}. If the target is to use the
8851 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8852 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8853 information, the hook should return @code{UI_DWARF2}.
8855 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8856 This may end up simplifying other parts of target-specific code. The
8857 default implementation of this hook never returns @code{UI_NONE}.
8859 Note that the value returned by this hook should be constant. It should
8860 not depend on anything except the command-line switches described by
8861 @var{opts}. In particular, the
8862 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8863 macros and builtin functions related to exception handling are set up
8864 depending on this setting.
8866 The default implementation of the hook first honors the
8867 @option{--enable-sjlj-exceptions} configure option, then
8868 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
8869 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
8870 must define this hook so that @var{opts} is used correctly.
8873 @deftypevr {Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8874 This variable should be set to @code{true} if the target ABI requires unwinding
8875 tables even when exceptions are not used. It must not be modified by
8876 command-line option processing.
8879 @defmac DONT_USE_BUILTIN_SETJMP
8880 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8881 should use the @code{setjmp}/@code{longjmp} functions from the C library
8882 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8885 @defmac DWARF_CIE_DATA_ALIGNMENT
8886 This macro need only be defined if the target might save registers in the
8887 function prologue at an offset to the stack pointer that is not aligned to
8888 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8889 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8890 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8891 the target supports DWARF 2 frame unwind information.
8894 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8895 Contains the value true if the target should add a zero word onto the
8896 end of a Dwarf-2 frame info section when used for exception handling.
8897 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8901 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8902 Given a register, this hook should return a parallel of registers to
8903 represent where to find the register pieces. Define this hook if the
8904 register and its mode are represented in Dwarf in non-contiguous
8905 locations, or if the register should be represented in more than one
8906 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8907 If not defined, the default is to return @code{NULL_RTX}.
8910 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8911 If some registers are represented in Dwarf-2 unwind information in
8912 multiple pieces, define this hook to fill in information about the
8913 sizes of those pieces in the table used by the unwinder at runtime.
8914 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8915 filling in a single size corresponding to each hard register;
8916 @var{address} is the address of the table.
8919 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8920 This hook is used to output a reference from a frame unwinding table to
8921 the type_info object identified by @var{sym}. It should return @code{true}
8922 if the reference was output. Returning @code{false} will cause the
8923 reference to be output using the normal Dwarf2 routines.
8926 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8927 This flag should be set to @code{true} on targets that use an ARM EABI
8928 based unwinding library, and @code{false} on other targets. This effects
8929 the format of unwinding tables, and how the unwinder in entered after
8930 running a cleanup. The default is @code{false}.
8933 @node Alignment Output
8934 @subsection Assembler Commands for Alignment
8936 @c prevent bad page break with this line
8937 This describes commands for alignment.
8939 @defmac JUMP_ALIGN (@var{label})
8940 The alignment (log base 2) to put in front of @var{label}, which is
8941 a common destination of jumps and has no fallthru incoming edge.
8943 This macro need not be defined if you don't want any special alignment
8944 to be done at such a time. Most machine descriptions do not currently
8947 Unless it's necessary to inspect the @var{label} parameter, it is better
8948 to set the variable @var{align_jumps} in the target's
8949 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8950 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8953 @deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx @var{label})
8954 The maximum number of bytes to skip before @var{label} when applying
8955 @code{JUMP_ALIGN}. This works only if
8956 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8959 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8960 The alignment (log base 2) to put in front of @var{label}, which follows
8963 This macro need not be defined if you don't want any special alignment
8964 to be done at such a time. Most machine descriptions do not currently
8968 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx @var{label})
8969 The maximum number of bytes to skip before @var{label} when applying
8970 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8971 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8974 @defmac LOOP_ALIGN (@var{label})
8975 The alignment (log base 2) to put in front of @var{label}, which follows
8976 a @code{NOTE_INSN_LOOP_BEG} note.
8978 This macro need not be defined if you don't want any special alignment
8979 to be done at such a time. Most machine descriptions do not currently
8982 Unless it's necessary to inspect the @var{label} parameter, it is better
8983 to set the variable @code{align_loops} in the target's
8984 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8985 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8988 @deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx @var{label})
8989 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
8990 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
8994 @defmac LABEL_ALIGN (@var{label})
8995 The alignment (log base 2) to put in front of @var{label}.
8996 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8997 the maximum of the specified values is used.
8999 Unless it's necessary to inspect the @var{label} parameter, it is better
9000 to set the variable @code{align_labels} in the target's
9001 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9002 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9005 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx @var{label})
9006 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
9007 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
9011 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9012 A C statement to output to the stdio stream @var{stream} an assembler
9013 instruction to advance the location counter by @var{nbytes} bytes.
9014 Those bytes should be zero when loaded. @var{nbytes} will be a C
9015 expression of type @code{unsigned HOST_WIDE_INT}.
9018 @defmac ASM_NO_SKIP_IN_TEXT
9019 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9020 text section because it fails to put zeros in the bytes that are skipped.
9021 This is true on many Unix systems, where the pseudo--op to skip bytes
9022 produces no-op instructions rather than zeros when used in the text
9026 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9027 A C statement to output to the stdio stream @var{stream} an assembler
9028 command to advance the location counter to a multiple of 2 to the
9029 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9032 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9033 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9034 for padding, if necessary.
9037 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9038 A C statement to output to the stdio stream @var{stream} an assembler
9039 command to advance the location counter to a multiple of 2 to the
9040 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9041 satisfy the alignment request. @var{power} and @var{max_skip} will be
9042 a C expression of type @code{int}.
9046 @node Debugging Info
9047 @section Controlling Debugging Information Format
9049 @c prevent bad page break with this line
9050 This describes how to specify debugging information.
9053 * All Debuggers:: Macros that affect all debugging formats uniformly.
9054 * DBX Options:: Macros enabling specific options in DBX format.
9055 * DBX Hooks:: Hook macros for varying DBX format.
9056 * File Names and DBX:: Macros controlling output of file names in DBX format.
9057 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9058 * VMS Debug:: Macros for VMS debug format.
9062 @subsection Macros Affecting All Debugging Formats
9064 @c prevent bad page break with this line
9065 These macros affect all debugging formats.
9067 @defmac DBX_REGISTER_NUMBER (@var{regno})
9068 A C expression that returns the DBX register number for the compiler
9069 register number @var{regno}. In the default macro provided, the value
9070 of this expression will be @var{regno} itself. But sometimes there are
9071 some registers that the compiler knows about and DBX does not, or vice
9072 versa. In such cases, some register may need to have one number in the
9073 compiler and another for DBX@.
9075 If two registers have consecutive numbers inside GCC, and they can be
9076 used as a pair to hold a multiword value, then they @emph{must} have
9077 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9078 Otherwise, debuggers will be unable to access such a pair, because they
9079 expect register pairs to be consecutive in their own numbering scheme.
9081 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9082 does not preserve register pairs, then what you must do instead is
9083 redefine the actual register numbering scheme.
9086 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9087 A C expression that returns the integer offset value for an automatic
9088 variable having address @var{x} (an RTL expression). The default
9089 computation assumes that @var{x} is based on the frame-pointer and
9090 gives the offset from the frame-pointer. This is required for targets
9091 that produce debugging output for DBX or COFF-style debugging output
9092 for SDB and allow the frame-pointer to be eliminated when the
9093 @option{-g} options is used.
9096 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9097 A C expression that returns the integer offset value for an argument
9098 having address @var{x} (an RTL expression). The nominal offset is
9102 @defmac PREFERRED_DEBUGGING_TYPE
9103 A C expression that returns the type of debugging output GCC should
9104 produce when the user specifies just @option{-g}. Define
9105 this if you have arranged for GCC to support more than one format of
9106 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9107 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9108 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9110 When the user specifies @option{-ggdb}, GCC normally also uses the
9111 value of this macro to select the debugging output format, but with two
9112 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9113 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9114 defined, GCC uses @code{DBX_DEBUG}.
9116 The value of this macro only affects the default debugging output; the
9117 user can always get a specific type of output by using @option{-gstabs},
9118 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9122 @subsection Specific Options for DBX Output
9124 @c prevent bad page break with this line
9125 These are specific options for DBX output.
9127 @defmac DBX_DEBUGGING_INFO
9128 Define this macro if GCC should produce debugging output for DBX
9129 in response to the @option{-g} option.
9132 @defmac XCOFF_DEBUGGING_INFO
9133 Define this macro if GCC should produce XCOFF format debugging output
9134 in response to the @option{-g} option. This is a variant of DBX format.
9137 @defmac DEFAULT_GDB_EXTENSIONS
9138 Define this macro to control whether GCC should by default generate
9139 GDB's extended version of DBX debugging information (assuming DBX-format
9140 debugging information is enabled at all). If you don't define the
9141 macro, the default is 1: always generate the extended information
9142 if there is any occasion to.
9145 @defmac DEBUG_SYMS_TEXT
9146 Define this macro if all @code{.stabs} commands should be output while
9147 in the text section.
9150 @defmac ASM_STABS_OP
9151 A C string constant, including spacing, naming the assembler pseudo op to
9152 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9153 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9154 applies only to DBX debugging information format.
9157 @defmac ASM_STABD_OP
9158 A C string constant, including spacing, naming the assembler pseudo op to
9159 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9160 value is the current location. If you don't define this macro,
9161 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9165 @defmac ASM_STABN_OP
9166 A C string constant, including spacing, naming the assembler pseudo op to
9167 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9168 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9169 macro applies only to DBX debugging information format.
9172 @defmac DBX_NO_XREFS
9173 Define this macro if DBX on your system does not support the construct
9174 @samp{xs@var{tagname}}. On some systems, this construct is used to
9175 describe a forward reference to a structure named @var{tagname}.
9176 On other systems, this construct is not supported at all.
9179 @defmac DBX_CONTIN_LENGTH
9180 A symbol name in DBX-format debugging information is normally
9181 continued (split into two separate @code{.stabs} directives) when it
9182 exceeds a certain length (by default, 80 characters). On some
9183 operating systems, DBX requires this splitting; on others, splitting
9184 must not be done. You can inhibit splitting by defining this macro
9185 with the value zero. You can override the default splitting-length by
9186 defining this macro as an expression for the length you desire.
9189 @defmac DBX_CONTIN_CHAR
9190 Normally continuation is indicated by adding a @samp{\} character to
9191 the end of a @code{.stabs} string when a continuation follows. To use
9192 a different character instead, define this macro as a character
9193 constant for the character you want to use. Do not define this macro
9194 if backslash is correct for your system.
9197 @defmac DBX_STATIC_STAB_DATA_SECTION
9198 Define this macro if it is necessary to go to the data section before
9199 outputting the @samp{.stabs} pseudo-op for a non-global static
9203 @defmac DBX_TYPE_DECL_STABS_CODE
9204 The value to use in the ``code'' field of the @code{.stabs} directive
9205 for a typedef. The default is @code{N_LSYM}.
9208 @defmac DBX_STATIC_CONST_VAR_CODE
9209 The value to use in the ``code'' field of the @code{.stabs} directive
9210 for a static variable located in the text section. DBX format does not
9211 provide any ``right'' way to do this. The default is @code{N_FUN}.
9214 @defmac DBX_REGPARM_STABS_CODE
9215 The value to use in the ``code'' field of the @code{.stabs} directive
9216 for a parameter passed in registers. DBX format does not provide any
9217 ``right'' way to do this. The default is @code{N_RSYM}.
9220 @defmac DBX_REGPARM_STABS_LETTER
9221 The letter to use in DBX symbol data to identify a symbol as a parameter
9222 passed in registers. DBX format does not customarily provide any way to
9223 do this. The default is @code{'P'}.
9226 @defmac DBX_FUNCTION_FIRST
9227 Define this macro if the DBX information for a function and its
9228 arguments should precede the assembler code for the function. Normally,
9229 in DBX format, the debugging information entirely follows the assembler
9233 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9234 Define this macro, with value 1, if the value of a symbol describing
9235 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9236 relative to the start of the enclosing function. Normally, GCC uses
9237 an absolute address.
9240 @defmac DBX_LINES_FUNCTION_RELATIVE
9241 Define this macro, with value 1, if the value of a symbol indicating
9242 the current line number (@code{N_SLINE}) should be relative to the
9243 start of the enclosing function. Normally, GCC uses an absolute address.
9246 @defmac DBX_USE_BINCL
9247 Define this macro if GCC should generate @code{N_BINCL} and
9248 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9249 macro also directs GCC to output a type number as a pair of a file
9250 number and a type number within the file. Normally, GCC does not
9251 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9252 number for a type number.
9256 @subsection Open-Ended Hooks for DBX Format
9258 @c prevent bad page break with this line
9259 These are hooks for DBX format.
9261 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9262 Define this macro to say how to output to @var{stream} the debugging
9263 information for the start of a scope level for variable names. The
9264 argument @var{name} is the name of an assembler symbol (for use with
9265 @code{assemble_name}) whose value is the address where the scope begins.
9268 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9269 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9272 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9273 Define this macro if the target machine requires special handling to
9274 output an @code{N_FUN} entry for the function @var{decl}.
9277 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9278 A C statement to output DBX debugging information before code for line
9279 number @var{line} of the current source file to the stdio stream
9280 @var{stream}. @var{counter} is the number of time the macro was
9281 invoked, including the current invocation; it is intended to generate
9282 unique labels in the assembly output.
9284 This macro should not be defined if the default output is correct, or
9285 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9288 @defmac NO_DBX_FUNCTION_END
9289 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9290 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9291 On those machines, define this macro to turn this feature off without
9292 disturbing the rest of the gdb extensions.
9295 @defmac NO_DBX_BNSYM_ENSYM
9296 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9297 extension construct. On those machines, define this macro to turn this
9298 feature off without disturbing the rest of the gdb extensions.
9301 @node File Names and DBX
9302 @subsection File Names in DBX Format
9304 @c prevent bad page break with this line
9305 This describes file names in DBX format.
9307 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9308 A C statement to output DBX debugging information to the stdio stream
9309 @var{stream}, which indicates that file @var{name} is the main source
9310 file---the file specified as the input file for compilation.
9311 This macro is called only once, at the beginning of compilation.
9313 This macro need not be defined if the standard form of output
9314 for DBX debugging information is appropriate.
9316 It may be necessary to refer to a label equal to the beginning of the
9317 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9318 to do so. If you do this, you must also set the variable
9319 @var{used_ltext_label_name} to @code{true}.
9322 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9323 Define this macro, with value 1, if GCC should not emit an indication
9324 of the current directory for compilation and current source language at
9325 the beginning of the file.
9328 @defmac NO_DBX_GCC_MARKER
9329 Define this macro, with value 1, if GCC should not emit an indication
9330 that this object file was compiled by GCC@. The default is to emit
9331 an @code{N_OPT} stab at the beginning of every source file, with
9332 @samp{gcc2_compiled.} for the string and value 0.
9335 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9336 A C statement to output DBX debugging information at the end of
9337 compilation of the main source file @var{name}. Output should be
9338 written to the stdio stream @var{stream}.
9340 If you don't define this macro, nothing special is output at the end
9341 of compilation, which is correct for most machines.
9344 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9345 Define this macro @emph{instead of} defining
9346 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9347 the end of compilation is an @code{N_SO} stab with an empty string,
9348 whose value is the highest absolute text address in the file.
9353 @subsection Macros for SDB and DWARF Output
9355 @c prevent bad page break with this line
9356 Here are macros for SDB and DWARF output.
9358 @defmac SDB_DEBUGGING_INFO
9359 Define this macro if GCC should produce COFF-style debugging output
9360 for SDB in response to the @option{-g} option.
9363 @defmac DWARF2_DEBUGGING_INFO
9364 Define this macro if GCC should produce dwarf version 2 format
9365 debugging output in response to the @option{-g} option.
9367 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
9368 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9369 be emitted for each function. Instead of an integer return the enum
9370 value for the @code{DW_CC_} tag.
9373 To support optional call frame debugging information, you must also
9374 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9375 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9376 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9377 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9380 @defmac DWARF2_FRAME_INFO
9381 Define this macro to a nonzero value if GCC should always output
9382 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9383 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9384 exceptions are enabled, GCC will output this information not matter
9385 how you define @code{DWARF2_FRAME_INFO}.
9388 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
9389 This hook defines the mechanism that will be used for describing frame
9390 unwind information to the debugger. Normally the hook will return
9391 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9392 return @code{UI_NONE} otherwise.
9394 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9395 is disabled in order to always output DWARF 2 frame information.
9397 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9398 This will suppress generation of the normal debug frame unwind information.
9401 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9402 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9403 line debug info sections. This will result in much more compact line number
9404 tables, and hence is desirable if it works.
9407 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
9408 True if the @code{.debug_pubtypes} and @code{.debug_pubnames} sections should be emitted. These sections are not used on most platforms, and in particular GDB does not use them.
9411 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9412 A C statement to issue assembly directives that create a difference
9413 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9416 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9417 A C statement to issue assembly directives that create a difference
9418 between the two given labels in system defined units, e.g. instruction
9419 slots on IA64 VMS, using an integer of the given size.
9422 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9423 A C statement to issue assembly directives that create a
9424 section-relative reference to the given @var{label}, using an integer of the
9425 given @var{size}. The label is known to be defined in the given @var{section}.
9428 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9429 A C statement to issue assembly directives that create a self-relative
9430 reference to the given @var{label}, using an integer of the given @var{size}.
9433 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9434 A C statement to issue assembly directives that create a reference to
9435 the DWARF table identifier @var{label} from the current section. This
9436 is used on some systems to avoid garbage collecting a DWARF table which
9437 is referenced by a function.
9440 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
9441 If defined, this target hook is a function which outputs a DTP-relative
9442 reference to the given TLS symbol of the specified size.
9445 @defmac PUT_SDB_@dots{}
9446 Define these macros to override the assembler syntax for the special
9447 SDB assembler directives. See @file{sdbout.c} for a list of these
9448 macros and their arguments. If the standard syntax is used, you need
9449 not define them yourself.
9453 Some assemblers do not support a semicolon as a delimiter, even between
9454 SDB assembler directives. In that case, define this macro to be the
9455 delimiter to use (usually @samp{\n}). It is not necessary to define
9456 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9460 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9461 Define this macro to allow references to unknown structure,
9462 union, or enumeration tags to be emitted. Standard COFF does not
9463 allow handling of unknown references, MIPS ECOFF has support for
9467 @defmac SDB_ALLOW_FORWARD_REFERENCES
9468 Define this macro to allow references to structure, union, or
9469 enumeration tags that have not yet been seen to be handled. Some
9470 assemblers choke if forward tags are used, while some require it.
9473 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9474 A C statement to output SDB debugging information before code for line
9475 number @var{line} of the current source file to the stdio stream
9476 @var{stream}. The default is to emit an @code{.ln} directive.
9481 @subsection Macros for VMS Debug Format
9483 @c prevent bad page break with this line
9484 Here are macros for VMS debug format.
9486 @defmac VMS_DEBUGGING_INFO
9487 Define this macro if GCC should produce debugging output for VMS
9488 in response to the @option{-g} option. The default behavior for VMS
9489 is to generate minimal debug info for a traceback in the absence of
9490 @option{-g} unless explicitly overridden with @option{-g0}. This
9491 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9492 @code{TARGET_OPTION_OVERRIDE}.
9495 @node Floating Point
9496 @section Cross Compilation and Floating Point
9497 @cindex cross compilation and floating point
9498 @cindex floating point and cross compilation
9500 While all modern machines use twos-complement representation for integers,
9501 there are a variety of representations for floating point numbers. This
9502 means that in a cross-compiler the representation of floating point numbers
9503 in the compiled program may be different from that used in the machine
9504 doing the compilation.
9506 Because different representation systems may offer different amounts of
9507 range and precision, all floating point constants must be represented in
9508 the target machine's format. Therefore, the cross compiler cannot
9509 safely use the host machine's floating point arithmetic; it must emulate
9510 the target's arithmetic. To ensure consistency, GCC always uses
9511 emulation to work with floating point values, even when the host and
9512 target floating point formats are identical.
9514 The following macros are provided by @file{real.h} for the compiler to
9515 use. All parts of the compiler which generate or optimize
9516 floating-point calculations must use these macros. They may evaluate
9517 their operands more than once, so operands must not have side effects.
9519 @defmac REAL_VALUE_TYPE
9520 The C data type to be used to hold a floating point value in the target
9521 machine's format. Typically this is a @code{struct} containing an
9522 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9526 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9527 Compares for equality the two values, @var{x} and @var{y}. If the target
9528 floating point format supports negative zeroes and/or NaNs,
9529 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9530 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9533 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9534 Tests whether @var{x} is less than @var{y}.
9537 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9538 Truncates @var{x} to a signed integer, rounding toward zero.
9541 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9542 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9543 @var{x} is negative, returns zero.
9546 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9547 Converts @var{string} into a floating point number in the target machine's
9548 representation for mode @var{mode}. This routine can handle both
9549 decimal and hexadecimal floating point constants, using the syntax
9550 defined by the C language for both.
9553 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9554 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9557 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9558 Determines whether @var{x} represents infinity (positive or negative).
9561 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9562 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9565 @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})
9566 Calculates an arithmetic operation on the two floating point values
9567 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9570 The operation to be performed is specified by @var{code}. Only the
9571 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9572 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9574 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9575 target's floating point format cannot represent infinity, it will call
9576 @code{abort}. Callers should check for this situation first, using
9577 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9580 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9581 Returns the negative of the floating point value @var{x}.
9584 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9585 Returns the absolute value of @var{x}.
9588 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9589 Truncates the floating point value @var{x} to fit in @var{mode}. The
9590 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9591 appropriate bit pattern to be output as a floating constant whose
9592 precision accords with mode @var{mode}.
9595 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9596 Converts a floating point value @var{x} into a double-precision integer
9597 which is then stored into @var{low} and @var{high}. If the value is not
9598 integral, it is truncated.
9601 @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})
9602 Converts a double-precision integer found in @var{low} and @var{high},
9603 into a floating point value which is then stored into @var{x}. The
9604 value is truncated to fit in mode @var{mode}.
9607 @node Mode Switching
9608 @section Mode Switching Instructions
9609 @cindex mode switching
9610 The following macros control mode switching optimizations:
9612 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9613 Define this macro if the port needs extra instructions inserted for mode
9614 switching in an optimizing compilation.
9616 For an example, the SH4 can perform both single and double precision
9617 floating point operations, but to perform a single precision operation,
9618 the FPSCR PR bit has to be cleared, while for a double precision
9619 operation, this bit has to be set. Changing the PR bit requires a general
9620 purpose register as a scratch register, hence these FPSCR sets have to
9621 be inserted before reload, i.e.@: you can't put this into instruction emitting
9622 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9624 You can have multiple entities that are mode-switched, and select at run time
9625 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9626 return nonzero for any @var{entity} that needs mode-switching.
9627 If you define this macro, you also have to define
9628 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9629 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9630 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9634 @defmac NUM_MODES_FOR_MODE_SWITCHING
9635 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9636 initializer for an array of integers. Each initializer element
9637 N refers to an entity that needs mode switching, and specifies the number
9638 of different modes that might need to be set for this entity.
9639 The position of the initializer in the initializer---starting counting at
9640 zero---determines the integer that is used to refer to the mode-switched
9642 In macros that take mode arguments / yield a mode result, modes are
9643 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9644 switch is needed / supplied.
9647 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9648 @var{entity} is an integer specifying a mode-switched entity. If
9649 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9650 return an integer value not larger than the corresponding element in
9651 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9652 be switched into prior to the execution of @var{insn}.
9655 @defmac MODE_AFTER (@var{mode}, @var{insn})
9656 If this macro is defined, it is evaluated for every @var{insn} during
9657 mode switching. It determines the mode that an insn results in (if
9658 different from the incoming mode).
9661 @defmac MODE_ENTRY (@var{entity})
9662 If this macro is defined, it is evaluated for every @var{entity} that needs
9663 mode switching. It should evaluate to an integer, which is a mode that
9664 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9665 is defined then @code{MODE_EXIT} must be defined.
9668 @defmac MODE_EXIT (@var{entity})
9669 If this macro is defined, it is evaluated for every @var{entity} that needs
9670 mode switching. It should evaluate to an integer, which is a mode that
9671 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9672 is defined then @code{MODE_ENTRY} must be defined.
9675 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9676 This macro specifies the order in which modes for @var{entity} are processed.
9677 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9678 lowest. The value of the macro should be an integer designating a mode
9679 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9680 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9681 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9684 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9685 Generate one or more insns to set @var{entity} to @var{mode}.
9686 @var{hard_reg_live} is the set of hard registers live at the point where
9687 the insn(s) are to be inserted.
9690 @node Target Attributes
9691 @section Defining target-specific uses of @code{__attribute__}
9692 @cindex target attributes
9693 @cindex machine attributes
9694 @cindex attributes, target-specific
9696 Target-specific attributes may be defined for functions, data and types.
9697 These are described using the following target hooks; they also need to
9698 be documented in @file{extend.texi}.
9700 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9701 If defined, this target hook points to an array of @samp{struct
9702 attribute_spec} (defined in @file{tree.h}) specifying the machine
9703 specific attributes for this target and some of the restrictions on the
9704 entities to which these attributes are applied and the arguments they
9708 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
9709 If defined, this target hook is a function which returns true if the
9710 machine-specific attribute named @var{name} expects an identifier
9711 given as its first argument to be passed on as a plain identifier, not
9712 subjected to name lookup. If this is not defined, the default is
9713 false for all machine-specific attributes.
9716 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
9717 If defined, this target hook is a function which returns zero if the attributes on
9718 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9719 and two if they are nearly compatible (which causes a warning to be
9720 generated). If this is not defined, machine-specific attributes are
9721 supposed always to be compatible.
9724 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9725 If defined, this target hook is a function which assigns default attributes to
9726 the newly defined @var{type}.
9729 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9730 Define this target hook if the merging of type attributes needs special
9731 handling. If defined, the result is a list of the combined
9732 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9733 that @code{comptypes} has already been called and returned 1. This
9734 function may call @code{merge_attributes} to handle machine-independent
9738 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9739 Define this target hook if the merging of decl attributes needs special
9740 handling. If defined, the result is a list of the combined
9741 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9742 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9743 when this is needed are when one attribute overrides another, or when an
9744 attribute is nullified by a subsequent definition. This function may
9745 call @code{merge_attributes} to handle machine-independent merging.
9747 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9748 If the only target-specific handling you require is @samp{dllimport}
9749 for Microsoft Windows targets, you should define the macro
9750 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9751 will then define a function called
9752 @code{merge_dllimport_decl_attributes} which can then be defined as
9753 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9754 add @code{handle_dll_attribute} in the attribute table for your port
9755 to perform initial processing of the @samp{dllimport} and
9756 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9757 @file{i386/i386.c}, for example.
9760 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
9761 @var{decl} is a variable or function with @code{__attribute__((dllimport))} specified. Use this hook if the target needs to add extra validation checks to @code{handle_dll_attribute}.
9764 @defmac TARGET_DECLSPEC
9765 Define this macro to a nonzero value if you want to treat
9766 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9767 default, this behavior is enabled only for targets that define
9768 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9769 of @code{__declspec} is via a built-in macro, but you should not rely
9770 on this implementation detail.
9773 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9774 Define this target hook if you want to be able to add attributes to a decl
9775 when it is being created. This is normally useful for back ends which
9776 wish to implement a pragma by using the attributes which correspond to
9777 the pragma's effect. The @var{node} argument is the decl which is being
9778 created. The @var{attr_ptr} argument is a pointer to the attribute list
9779 for this decl. The list itself should not be modified, since it may be
9780 shared with other decls, but attributes may be chained on the head of
9781 the list and @code{*@var{attr_ptr}} modified to point to the new
9782 attributes, or a copy of the list may be made if further changes are
9786 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
9788 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9789 into the current function, despite its having target-specific
9790 attributes, @code{false} otherwise. By default, if a function has a
9791 target specific attribute attached to it, it will not be inlined.
9794 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9795 This hook is called to parse the @code{attribute(option("..."))}, and
9796 it allows the function to set different target machine compile time
9797 options for the current function that might be different than the
9798 options specified on the command line. The hook should return
9799 @code{true} if the options are valid.
9801 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9802 the function declaration to hold a pointer to a target specific
9803 @var{struct cl_target_option} structure.
9806 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr})
9807 This hook is called to save any additional target specific information
9808 in the @var{struct cl_target_option} structure for function specific
9810 @xref{Option file format}.
9813 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct cl_target_option *@var{ptr})
9814 This hook is called to restore any additional target specific
9815 information in the @var{struct cl_target_option} structure for
9816 function specific options.
9819 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
9820 This hook is called to print any additional target specific
9821 information in the @var{struct cl_target_option} structure for
9822 function specific options.
9825 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
9826 This target hook parses the options for @code{#pragma GCC option} to
9827 set the machine specific options for functions that occur later in the
9828 input stream. The options should be the same as handled by the
9829 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9832 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
9833 Sometimes certain combinations of command options do not make sense on
9834 a particular target machine. You can override the hook
9835 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9836 once just after all the command options have been parsed.
9838 Don't use this hook to turn on various extra optimizations for
9839 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9841 If you need to do something whenever the optimization level is
9842 changed via the optimize attribute or pragma, see
9843 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9846 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
9847 This target hook returns @code{false} if the @var{caller} function
9848 cannot inline @var{callee}, based on target specific information. By
9849 default, inlining is not allowed if the callee function has function
9850 specific target options and the caller does not use the same options.
9854 @section Emulating TLS
9855 @cindex Emulated TLS
9857 For targets whose psABI does not provide Thread Local Storage via
9858 specific relocations and instruction sequences, an emulation layer is
9859 used. A set of target hooks allows this emulation layer to be
9860 configured for the requirements of a particular target. For instance
9861 the psABI may in fact specify TLS support in terms of an emulation
9864 The emulation layer works by creating a control object for every TLS
9865 object. To access the TLS object, a lookup function is provided
9866 which, when given the address of the control object, will return the
9867 address of the current thread's instance of the TLS object.
9869 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9870 Contains the name of the helper function that uses a TLS control
9871 object to locate a TLS instance. The default causes libgcc's
9872 emulated TLS helper function to be used.
9875 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9876 Contains the name of the helper function that should be used at
9877 program startup to register TLS objects that are implicitly
9878 initialized to zero. If this is @code{NULL}, all TLS objects will
9879 have explicit initializers. The default causes libgcc's emulated TLS
9880 registration function to be used.
9883 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9884 Contains the name of the section in which TLS control variables should
9885 be placed. The default of @code{NULL} allows these to be placed in
9889 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9890 Contains the name of the section in which TLS initializers should be
9891 placed. The default of @code{NULL} allows these to be placed in any
9895 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9896 Contains the prefix to be prepended to TLS control variable names.
9897 The default of @code{NULL} uses a target-specific prefix.
9900 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9901 Contains the prefix to be prepended to TLS initializer objects. The
9902 default of @code{NULL} uses a target-specific prefix.
9905 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
9906 Specifies a function that generates the FIELD_DECLs for a TLS control
9907 object type. @var{type} is the RECORD_TYPE the fields are for and
9908 @var{name} should be filled with the structure tag, if the default of
9909 @code{__emutls_object} is unsuitable. The default creates a type suitable
9910 for libgcc's emulated TLS function.
9913 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
9914 Specifies a function that generates the CONSTRUCTOR to initialize a
9915 TLS control object. @var{var} is the TLS control object, @var{decl}
9916 is the TLS object and @var{tmpl_addr} is the address of the
9917 initializer. The default initializes libgcc's emulated TLS control object.
9920 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
9921 Specifies whether the alignment of TLS control variable objects is
9922 fixed and should not be increased as some backends may do to optimize
9923 single objects. The default is false.
9926 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9927 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9928 may be used to describe emulated TLS control objects.
9931 @node MIPS Coprocessors
9932 @section Defining coprocessor specifics for MIPS targets.
9933 @cindex MIPS coprocessor-definition macros
9935 The MIPS specification allows MIPS implementations to have as many as 4
9936 coprocessors, each with as many as 32 private registers. GCC supports
9937 accessing these registers and transferring values between the registers
9938 and memory using asm-ized variables. For example:
9941 register unsigned int cp0count asm ("c0r1");
9947 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9948 names may be added as described below, or the default names may be
9949 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9951 Coprocessor registers are assumed to be epilogue-used; sets to them will
9952 be preserved even if it does not appear that the register is used again
9953 later in the function.
9955 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9956 the FPU@. One accesses COP1 registers through standard mips
9957 floating-point support; they are not included in this mechanism.
9959 There is one macro used in defining the MIPS coprocessor interface which
9960 you may want to override in subtargets; it is described below.
9962 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9963 A comma-separated list (with leading comma) of pairs describing the
9964 alternate names of coprocessor registers. The format of each entry should be
9966 @{ @var{alternatename}, @var{register_number}@}
9972 @section Parameters for Precompiled Header Validity Checking
9973 @cindex parameters, precompiled headers
9975 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
9976 This hook returns a pointer to the data needed by
9977 @code{TARGET_PCH_VALID_P} and sets
9978 @samp{*@var{sz}} to the size of the data in bytes.
9981 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
9982 This hook checks whether the options used to create a PCH file are
9983 compatible with the current settings. It returns @code{NULL}
9984 if so and a suitable error message if not. Error messages will
9985 be presented to the user and must be localized using @samp{_(@var{msg})}.
9987 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9988 when the PCH file was created and @var{sz} is the size of that data in bytes.
9989 It's safe to assume that the data was created by the same version of the
9990 compiler, so no format checking is needed.
9992 The default definition of @code{default_pch_valid_p} should be
9993 suitable for most targets.
9996 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
9997 If this hook is nonnull, the default implementation of
9998 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9999 of @code{target_flags}. @var{pch_flags} specifies the value that
10000 @code{target_flags} had when the PCH file was created. The return
10001 value is the same as for @code{TARGET_PCH_VALID_P}.
10005 @section C++ ABI parameters
10006 @cindex parameters, c++ abi
10008 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10009 Define this hook to override the integer type used for guard variables.
10010 These are used to implement one-time construction of static objects. The
10011 default is long_long_integer_type_node.
10014 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10015 This hook determines how guard variables are used. It should return
10016 @code{false} (the default) if the first byte should be used. A return value of
10017 @code{true} indicates that only the least significant bit should be used.
10020 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10021 This hook returns the size of the cookie to use when allocating an array
10022 whose elements have the indicated @var{type}. Assumes that it is already
10023 known that a cookie is needed. The default is
10024 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10025 IA64/Generic C++ ABI@.
10028 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10029 This hook should return @code{true} if the element size should be stored in
10030 array cookies. The default is to return @code{false}.
10033 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10034 If defined by a backend this hook allows the decision made to export
10035 class @var{type} to be overruled. Upon entry @var{import_export}
10036 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10037 to be imported and 0 otherwise. This function should return the
10038 modified value and perform any other actions necessary to support the
10039 backend's targeted operating system.
10042 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10043 This hook should return @code{true} if constructors and destructors return
10044 the address of the object created/destroyed. The default is to return
10048 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10049 This hook returns true if the key method for a class (i.e., the method
10050 which, if defined in the current translation unit, causes the virtual
10051 table to be emitted) may be an inline function. Under the standard
10052 Itanium C++ ABI the key method may be an inline function so long as
10053 the function is not declared inline in the class definition. Under
10054 some variants of the ABI, an inline function can never be the key
10055 method. The default is to return @code{true}.
10058 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10059 @var{decl} is a virtual table, virtual table table, typeinfo object, or other similar implicit class data object that will be emitted with external linkage in this translation unit. No ELF visibility has been explicitly specified. If the target needs to specify a visibility other than that of the containing class, use this hook to set @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
10062 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10063 This hook returns true (the default) if virtual tables and other
10064 similar implicit class data objects are always COMDAT if they have
10065 external linkage. If this hook returns false, then class data for
10066 classes whose virtual table will be emitted in only one translation
10067 unit will not be COMDAT.
10070 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10071 This hook returns true (the default) if the RTTI information for
10072 the basic types which is defined in the C++ runtime should always
10073 be COMDAT, false if it should not be COMDAT.
10076 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10077 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10078 should be used to register static destructors when @option{-fuse-cxa-atexit}
10079 is in effect. The default is to return false to use @code{__cxa_atexit}.
10082 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10083 This hook returns true if the target @code{atexit} function can be used
10084 in the same manner as @code{__cxa_atexit} to register C++ static
10085 destructors. This requires that @code{atexit}-registered functions in
10086 shared libraries are run in the correct order when the libraries are
10087 unloaded. The default is to return false.
10090 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10091 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been defined. Use this hook to make adjustments to the class (eg, tweak visibility or perform any other required target modifications).
10094 @node Named Address Spaces
10095 @section Adding support for named address spaces
10096 @cindex named address spaces
10098 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10099 standards committee, @cite{Programming Languages - C - Extensions to
10100 support embedded processors}, specifies a syntax for embedded
10101 processors to specify alternate address spaces. You can configure a
10102 GCC port to support section 5.1 of the draft report to add support for
10103 address spaces other than the default address space. These address
10104 spaces are new keywords that are similar to the @code{volatile} and
10105 @code{const} type attributes.
10107 Pointers to named address spaces can have a different size than
10108 pointers to the generic address space.
10110 For example, the SPU port uses the @code{__ea} address space to refer
10111 to memory in the host processor, rather than memory local to the SPU
10112 processor. Access to memory in the @code{__ea} address space involves
10113 issuing DMA operations to move data between the host processor and the
10114 local processor memory address space. Pointers in the @code{__ea}
10115 address space are either 32 bits or 64 bits based on the
10116 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10119 Internally, address spaces are represented as a small integer in the
10120 range 0 to 15 with address space 0 being reserved for the generic
10123 To register a named address space qualifier keyword with the C front end,
10124 the target may call the @code{c_register_addr_space} routine. For example,
10125 the SPU port uses the following to declare @code{__ea} as the keyword for
10126 named address space #1:
10128 #define ADDR_SPACE_EA 1
10129 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10132 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10133 Define this to return the machine mode to use for pointers to
10134 @var{address_space} if the target supports named address spaces.
10135 The default version of this hook returns @code{ptr_mode} for the
10136 generic address space only.
10139 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10140 Define this to return the machine mode to use for addresses in
10141 @var{address_space} if the target supports named address spaces.
10142 The default version of this hook returns @code{Pmode} for the
10143 generic address space only.
10146 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (enum machine_mode @var{mode}, addr_space_t @var{as})
10147 Define this to return nonzero if the port can handle pointers
10148 with machine mode @var{mode} to address space @var{as}. This target
10149 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10150 except that it includes explicit named address space support. The default
10151 version of this hook returns true for the modes returned by either the
10152 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10153 target hooks for the given address space.
10156 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{exp}, bool @var{strict}, addr_space_t @var{as})
10157 Define this to return true if @var{exp} is a valid address for mode
10158 @var{mode} in the named address space @var{as}. The @var{strict}
10159 parameter says whether strict addressing is in effect after reload has
10160 finished. This target hook is the same as the
10161 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10162 explicit named address space support.
10165 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode}, addr_space_t @var{as})
10166 Define this to modify an invalid address @var{x} to be a valid address
10167 with mode @var{mode} in the named address space @var{as}. This target
10168 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10169 except that it includes explicit named address space support.
10172 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{superset}, addr_space_t @var{subset})
10173 Define this to return whether the @var{subset} named address space is
10174 contained within the @var{superset} named address space. Pointers to
10175 a named address space that is a subset of another named address space
10176 will be converted automatically without a cast if used together in
10177 arithmetic operations. Pointers to a superset address space can be
10178 converted to pointers to a subset address space via explicit casts.
10181 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10182 Define this to convert the pointer expression represented by the RTL
10183 @var{op} with type @var{from_type} that points to a named address
10184 space to a new pointer expression with type @var{to_type} that points
10185 to a different named address space. When this hook it called, it is
10186 guaranteed that one of the two address spaces is a subset of the other,
10187 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10191 @section Miscellaneous Parameters
10192 @cindex parameters, miscellaneous
10194 @c prevent bad page break with this line
10195 Here are several miscellaneous parameters.
10197 @defmac HAS_LONG_COND_BRANCH
10198 Define this boolean macro to indicate whether or not your architecture
10199 has conditional branches that can span all of memory. It is used in
10200 conjunction with an optimization that partitions hot and cold basic
10201 blocks into separate sections of the executable. If this macro is
10202 set to false, gcc will convert any conditional branches that attempt
10203 to cross between sections into unconditional branches or indirect jumps.
10206 @defmac HAS_LONG_UNCOND_BRANCH
10207 Define this boolean macro to indicate whether or not your architecture
10208 has unconditional branches that can span all of memory. It is used in
10209 conjunction with an optimization that partitions hot and cold basic
10210 blocks into separate sections of the executable. If this macro is
10211 set to false, gcc will convert any unconditional branches that attempt
10212 to cross between sections into indirect jumps.
10215 @defmac CASE_VECTOR_MODE
10216 An alias for a machine mode name. This is the machine mode that
10217 elements of a jump-table should have.
10220 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10221 Optional: return the preferred mode for an @code{addr_diff_vec}
10222 when the minimum and maximum offset are known. If you define this,
10223 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10224 To make this work, you also have to define @code{INSN_ALIGN} and
10225 make the alignment for @code{addr_diff_vec} explicit.
10226 The @var{body} argument is provided so that the offset_unsigned and scale
10227 flags can be updated.
10230 @defmac CASE_VECTOR_PC_RELATIVE
10231 Define this macro to be a C expression to indicate when jump-tables
10232 should contain relative addresses. You need not define this macro if
10233 jump-tables never contain relative addresses, or jump-tables should
10234 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10238 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10239 This function return the smallest number of different values for which it
10240 is best to use a jump-table instead of a tree of conditional branches.
10241 The default is four for machines with a @code{casesi} instruction and
10242 five otherwise. This is best for most machines.
10245 @defmac CASE_USE_BIT_TESTS
10246 Define this macro to be a C expression to indicate whether C switch
10247 statements may be implemented by a sequence of bit tests. This is
10248 advantageous on processors that can efficiently implement left shift
10249 of 1 by the number of bits held in a register, but inappropriate on
10250 targets that would require a loop. By default, this macro returns
10251 @code{true} if the target defines an @code{ashlsi3} pattern, and
10252 @code{false} otherwise.
10255 @defmac WORD_REGISTER_OPERATIONS
10256 Define this macro if operations between registers with integral mode
10257 smaller than a word are always performed on the entire register.
10258 Most RISC machines have this property and most CISC machines do not.
10261 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10262 Define this macro to be a C expression indicating when insns that read
10263 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10264 bits outside of @var{mem_mode} to be either the sign-extension or the
10265 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10266 of @var{mem_mode} for which the
10267 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10268 @code{UNKNOWN} for other modes.
10270 This macro is not called with @var{mem_mode} non-integral or with a width
10271 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10272 value in this case. Do not define this macro if it would always return
10273 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10274 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10276 You may return a non-@code{UNKNOWN} value even if for some hard registers
10277 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10278 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10279 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10280 integral mode larger than this but not larger than @code{word_mode}.
10282 You must return @code{UNKNOWN} if for some hard registers that allow this
10283 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10284 @code{word_mode}, but that they can change to another integral mode that
10285 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10288 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10289 Define this macro if loading short immediate values into registers sign
10293 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
10294 Define this macro if the same instructions that convert a floating
10295 point number to a signed fixed point number also convert validly to an
10299 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
10300 When @option{-ffast-math} is in effect, GCC tries to optimize
10301 divisions by the same divisor, by turning them into multiplications by
10302 the reciprocal. This target hook specifies the minimum number of divisions
10303 that should be there for GCC to perform the optimization for a variable
10304 of mode @var{mode}. The default implementation returns 3 if the machine
10305 has an instruction for the division, and 2 if it does not.
10309 The maximum number of bytes that a single instruction can move quickly
10310 between memory and registers or between two memory locations.
10313 @defmac MAX_MOVE_MAX
10314 The maximum number of bytes that a single instruction can move quickly
10315 between memory and registers or between two memory locations. If this
10316 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10317 constant value that is the largest value that @code{MOVE_MAX} can have
10321 @defmac SHIFT_COUNT_TRUNCATED
10322 A C expression that is nonzero if on this machine the number of bits
10323 actually used for the count of a shift operation is equal to the number
10324 of bits needed to represent the size of the object being shifted. When
10325 this macro is nonzero, the compiler will assume that it is safe to omit
10326 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10327 truncates the count of a shift operation. On machines that have
10328 instructions that act on bit-fields at variable positions, which may
10329 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10330 also enables deletion of truncations of the values that serve as
10331 arguments to bit-field instructions.
10333 If both types of instructions truncate the count (for shifts) and
10334 position (for bit-field operations), or if no variable-position bit-field
10335 instructions exist, you should define this macro.
10337 However, on some machines, such as the 80386 and the 680x0, truncation
10338 only applies to shift operations and not the (real or pretended)
10339 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10340 such machines. Instead, add patterns to the @file{md} file that include
10341 the implied truncation of the shift instructions.
10343 You need not define this macro if it would always have the value of zero.
10346 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10347 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
10348 This function describes how the standard shift patterns for @var{mode}
10349 deal with shifts by negative amounts or by more than the width of the mode.
10350 @xref{shift patterns}.
10352 On many machines, the shift patterns will apply a mask @var{m} to the
10353 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10354 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10355 this is true for mode @var{mode}, the function should return @var{m},
10356 otherwise it should return 0. A return value of 0 indicates that no
10357 particular behavior is guaranteed.
10359 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10360 @emph{not} apply to general shift rtxes; it applies only to instructions
10361 that are generated by the named shift patterns.
10363 The default implementation of this function returns
10364 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10365 and 0 otherwise. This definition is always safe, but if
10366 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10367 nevertheless truncate the shift count, you may get better code
10371 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10372 A C expression which is nonzero if on this machine it is safe to
10373 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10374 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10375 operating on it as if it had only @var{outprec} bits.
10377 On many machines, this expression can be 1.
10379 @c rearranged this, removed the phrase "it is reported that". this was
10380 @c to fix an overfull hbox. --mew 10feb93
10381 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10382 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10383 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10384 such cases may improve things.
10387 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
10388 The representation of an integral mode can be such that the values
10389 are always extended to a wider integral mode. Return
10390 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10391 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10392 otherwise. (Currently, none of the targets use zero-extended
10393 representation this way so unlike @code{LOAD_EXTEND_OP},
10394 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10395 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10396 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10397 widest integral mode and currently we take advantage of this fact.)
10399 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10400 value even if the extension is not performed on certain hard registers
10401 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10402 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10404 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10405 describe two related properties. If you define
10406 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10407 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10410 In order to enforce the representation of @code{mode},
10411 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10415 @defmac STORE_FLAG_VALUE
10416 A C expression describing the value returned by a comparison operator
10417 with an integral mode and stored by a store-flag instruction
10418 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10419 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10420 comparison operators whose results have a @code{MODE_INT} mode.
10422 A value of 1 or @minus{}1 means that the instruction implementing the
10423 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10424 and 0 when the comparison is false. Otherwise, the value indicates
10425 which bits of the result are guaranteed to be 1 when the comparison is
10426 true. This value is interpreted in the mode of the comparison
10427 operation, which is given by the mode of the first operand in the
10428 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10429 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10432 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10433 generate code that depends only on the specified bits. It can also
10434 replace comparison operators with equivalent operations if they cause
10435 the required bits to be set, even if the remaining bits are undefined.
10436 For example, on a machine whose comparison operators return an
10437 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10438 @samp{0x80000000}, saying that just the sign bit is relevant, the
10442 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10446 can be converted to
10449 (ashift:SI @var{x} (const_int @var{n}))
10453 where @var{n} is the appropriate shift count to move the bit being
10454 tested into the sign bit.
10456 There is no way to describe a machine that always sets the low-order bit
10457 for a true value, but does not guarantee the value of any other bits,
10458 but we do not know of any machine that has such an instruction. If you
10459 are trying to port GCC to such a machine, include an instruction to
10460 perform a logical-and of the result with 1 in the pattern for the
10461 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10463 Often, a machine will have multiple instructions that obtain a value
10464 from a comparison (or the condition codes). Here are rules to guide the
10465 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10470 Use the shortest sequence that yields a valid definition for
10471 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10472 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10473 comparison operators to do so because there may be opportunities to
10474 combine the normalization with other operations.
10477 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10478 slightly preferred on machines with expensive jumps and 1 preferred on
10482 As a second choice, choose a value of @samp{0x80000001} if instructions
10483 exist that set both the sign and low-order bits but do not define the
10487 Otherwise, use a value of @samp{0x80000000}.
10490 Many machines can produce both the value chosen for
10491 @code{STORE_FLAG_VALUE} and its negation in the same number of
10492 instructions. On those machines, you should also define a pattern for
10493 those cases, e.g., one matching
10496 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10499 Some machines can also perform @code{and} or @code{plus} operations on
10500 condition code values with less instructions than the corresponding
10501 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10502 machines, define the appropriate patterns. Use the names @code{incscc}
10503 and @code{decscc}, respectively, for the patterns which perform
10504 @code{plus} or @code{minus} operations on condition code values. See
10505 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10506 find such instruction sequences on other machines.
10508 If this macro is not defined, the default value, 1, is used. You need
10509 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10510 instructions, or if the value generated by these instructions is 1.
10513 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10514 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10515 returned when comparison operators with floating-point results are true.
10516 Define this macro on machines that have comparison operations that return
10517 floating-point values. If there are no such operations, do not define
10521 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10522 A C expression that gives a rtx representing the nonzero true element
10523 for vector comparisons. The returned rtx should be valid for the inner
10524 mode of @var{mode} which is guaranteed to be a vector mode. Define
10525 this macro on machines that have vector comparison operations that
10526 return a vector result. If there are no such operations, do not define
10527 this macro. Typically, this macro is defined as @code{const1_rtx} or
10528 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10529 the compiler optimizing such vector comparison operations for the
10533 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10534 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10535 A C expression that indicates whether the architecture defines a value
10536 for @code{clz} or @code{ctz} with a zero operand.
10537 A result of @code{0} indicates the value is undefined.
10538 If the value is defined for only the RTL expression, the macro should
10539 evaluate to @code{1}; if the value applies also to the corresponding optab
10540 entry (which is normally the case if it expands directly into
10541 the corresponding RTL), then the macro should evaluate to @code{2}.
10542 In the cases where the value is defined, @var{value} should be set to
10545 If this macro is not defined, the value of @code{clz} or
10546 @code{ctz} at zero is assumed to be undefined.
10548 This macro must be defined if the target's expansion for @code{ffs}
10549 relies on a particular value to get correct results. Otherwise it
10550 is not necessary, though it may be used to optimize some corner cases, and
10551 to provide a default expansion for the @code{ffs} optab.
10553 Note that regardless of this macro the ``definedness'' of @code{clz}
10554 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10555 visible to the user. Thus one may be free to adjust the value at will
10556 to match the target expansion of these operations without fear of
10561 An alias for the machine mode for pointers. On most machines, define
10562 this to be the integer mode corresponding to the width of a hardware
10563 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10564 On some machines you must define this to be one of the partial integer
10565 modes, such as @code{PSImode}.
10567 The width of @code{Pmode} must be at least as large as the value of
10568 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10569 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10573 @defmac FUNCTION_MODE
10574 An alias for the machine mode used for memory references to functions
10575 being called, in @code{call} RTL expressions. On most CISC machines,
10576 where an instruction can begin at any byte address, this should be
10577 @code{QImode}. On most RISC machines, where all instructions have fixed
10578 size and alignment, this should be a mode with the same size and alignment
10579 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10582 @defmac STDC_0_IN_SYSTEM_HEADERS
10583 In normal operation, the preprocessor expands @code{__STDC__} to the
10584 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10585 hosts, like Solaris, the system compiler uses a different convention,
10586 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10587 strict conformance to the C Standard.
10589 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10590 convention when processing system header files, but when processing user
10591 files @code{__STDC__} will always expand to 1.
10594 @defmac NO_IMPLICIT_EXTERN_C
10595 Define this macro if the system header files support C++ as well as C@.
10596 This macro inhibits the usual method of using system header files in
10597 C++, which is to pretend that the file's contents are enclosed in
10598 @samp{extern "C" @{@dots{}@}}.
10603 @defmac REGISTER_TARGET_PRAGMAS ()
10604 Define this macro if you want to implement any target-specific pragmas.
10605 If defined, it is a C expression which makes a series of calls to
10606 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10607 for each pragma. The macro may also do any
10608 setup required for the pragmas.
10610 The primary reason to define this macro is to provide compatibility with
10611 other compilers for the same target. In general, we discourage
10612 definition of target-specific pragmas for GCC@.
10614 If the pragma can be implemented by attributes then you should consider
10615 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10617 Preprocessor macros that appear on pragma lines are not expanded. All
10618 @samp{#pragma} directives that do not match any registered pragma are
10619 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10622 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10623 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10625 Each call to @code{c_register_pragma} or
10626 @code{c_register_pragma_with_expansion} establishes one pragma. The
10627 @var{callback} routine will be called when the preprocessor encounters a
10631 #pragma [@var{space}] @var{name} @dots{}
10634 @var{space} is the case-sensitive namespace of the pragma, or
10635 @code{NULL} to put the pragma in the global namespace. The callback
10636 routine receives @var{pfile} as its first argument, which can be passed
10637 on to cpplib's functions if necessary. You can lex tokens after the
10638 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10639 callback will be silently ignored. The end of the line is indicated by
10640 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10641 arguments of pragmas registered with
10642 @code{c_register_pragma_with_expansion} but not on the arguments of
10643 pragmas registered with @code{c_register_pragma}.
10645 Note that the use of @code{pragma_lex} is specific to the C and C++
10646 compilers. It will not work in the Java or Fortran compilers, or any
10647 other language compilers for that matter. Thus if @code{pragma_lex} is going
10648 to be called from target-specific code, it must only be done so when
10649 building the C and C++ compilers. This can be done by defining the
10650 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10651 target entry in the @file{config.gcc} file. These variables should name
10652 the target-specific, language-specific object file which contains the
10653 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10654 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10655 how to build this object file.
10658 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10659 Define this macro if macros should be expanded in the
10660 arguments of @samp{#pragma pack}.
10663 @deftypevr {Target Hook} bool TARGET_HANDLE_PRAGMA_EXTERN_PREFIX
10664 True if @code{#pragma extern_prefix} is to be supported.
10667 @defmac TARGET_DEFAULT_PACK_STRUCT
10668 If your target requires a structure packing default other than 0 (meaning
10669 the machine default), define this macro to the necessary value (in bytes).
10670 This must be a value that would also be valid to use with
10671 @samp{#pragma pack()} (that is, a small power of two).
10674 @defmac DOLLARS_IN_IDENTIFIERS
10675 Define this macro to control use of the character @samp{$} in
10676 identifier names for the C family of languages. 0 means @samp{$} is
10677 not allowed by default; 1 means it is allowed. 1 is the default;
10678 there is no need to define this macro in that case.
10681 @defmac NO_DOLLAR_IN_LABEL
10682 Define this macro if the assembler does not accept the character
10683 @samp{$} in label names. By default constructors and destructors in
10684 G++ have @samp{$} in the identifiers. If this macro is defined,
10685 @samp{.} is used instead.
10688 @defmac NO_DOT_IN_LABEL
10689 Define this macro if the assembler does not accept the character
10690 @samp{.} in label names. By default constructors and destructors in G++
10691 have names that use @samp{.}. If this macro is defined, these names
10692 are rewritten to avoid @samp{.}.
10695 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10696 Define this macro as a C expression that is nonzero if it is safe for the
10697 delay slot scheduler to place instructions in the delay slot of @var{insn},
10698 even if they appear to use a resource set or clobbered in @var{insn}.
10699 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10700 every @code{call_insn} has this behavior. On machines where some @code{insn}
10701 or @code{jump_insn} is really a function call and hence has this behavior,
10702 you should define this macro.
10704 You need not define this macro if it would always return zero.
10707 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10708 Define this macro as a C expression that is nonzero if it is safe for the
10709 delay slot scheduler to place instructions in the delay slot of @var{insn},
10710 even if they appear to set or clobber a resource referenced in @var{insn}.
10711 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10712 some @code{insn} or @code{jump_insn} is really a function call and its operands
10713 are registers whose use is actually in the subroutine it calls, you should
10714 define this macro. Doing so allows the delay slot scheduler to move
10715 instructions which copy arguments into the argument registers into the delay
10716 slot of @var{insn}.
10718 You need not define this macro if it would always return zero.
10721 @defmac MULTIPLE_SYMBOL_SPACES
10722 Define this macro as a C expression that is nonzero if, in some cases,
10723 global symbols from one translation unit may not be bound to undefined
10724 symbols in another translation unit without user intervention. For
10725 instance, under Microsoft Windows symbols must be explicitly imported
10726 from shared libraries (DLLs).
10728 You need not define this macro if it would always evaluate to zero.
10731 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10732 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10733 any hard regs the port wishes to automatically clobber for an asm.
10734 It should return the result of the last @code{tree_cons} used to add a
10735 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10736 corresponding parameters to the asm and may be inspected to avoid
10737 clobbering a register that is an input or output of the asm. You can use
10738 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10739 for overlap with regards to asm-declared registers.
10742 @defmac MATH_LIBRARY
10743 Define this macro as a C string constant for the linker argument to link
10744 in the system math library, minus the initial @samp{"-l"}, or
10745 @samp{""} if the target does not have a
10746 separate math library.
10748 You need only define this macro if the default of @samp{"m"} is wrong.
10751 @defmac LIBRARY_PATH_ENV
10752 Define this macro as a C string constant for the environment variable that
10753 specifies where the linker should look for libraries.
10755 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10759 @defmac TARGET_POSIX_IO
10760 Define this macro if the target supports the following POSIX@ file
10761 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10762 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10763 to use file locking when exiting a program, which avoids race conditions
10764 if the program has forked. It will also create directories at run-time
10765 for cross-profiling.
10768 @defmac MAX_CONDITIONAL_EXECUTE
10770 A C expression for the maximum number of instructions to execute via
10771 conditional execution instructions instead of a branch. A value of
10772 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10773 1 if it does use cc0.
10776 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10777 Used if the target needs to perform machine-dependent modifications on the
10778 conditionals used for turning basic blocks into conditionally executed code.
10779 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10780 contains information about the currently processed blocks. @var{true_expr}
10781 and @var{false_expr} are the tests that are used for converting the
10782 then-block and the else-block, respectively. Set either @var{true_expr} or
10783 @var{false_expr} to a null pointer if the tests cannot be converted.
10786 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10787 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10788 if-statements into conditions combined by @code{and} and @code{or} operations.
10789 @var{bb} contains the basic block that contains the test that is currently
10790 being processed and about to be turned into a condition.
10793 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10794 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10795 be converted to conditional execution format. @var{ce_info} points to
10796 a data structure, @code{struct ce_if_block}, which contains information
10797 about the currently processed blocks.
10800 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10801 A C expression to perform any final machine dependent modifications in
10802 converting code to conditional execution. The involved basic blocks
10803 can be found in the @code{struct ce_if_block} structure that is pointed
10804 to by @var{ce_info}.
10807 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10808 A C expression to cancel any machine dependent modifications in
10809 converting code to conditional execution. The involved basic blocks
10810 can be found in the @code{struct ce_if_block} structure that is pointed
10811 to by @var{ce_info}.
10814 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10815 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10816 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10819 @defmac IFCVT_EXTRA_FIELDS
10820 If defined, it should expand to a set of field declarations that will be
10821 added to the @code{struct ce_if_block} structure. These should be initialized
10822 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10825 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
10826 If non-null, this hook performs a target-specific pass over the
10827 instruction stream. The compiler will run it at all optimization levels,
10828 just before the point at which it normally does delayed-branch scheduling.
10830 The exact purpose of the hook varies from target to target. Some use
10831 it to do transformations that are necessary for correctness, such as
10832 laying out in-function constant pools or avoiding hardware hazards.
10833 Others use it as an opportunity to do some machine-dependent optimizations.
10835 You need not implement the hook if it has nothing to do. The default
10836 definition is null.
10839 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
10840 Define this hook if you have any machine-specific built-in functions
10841 that need to be defined. It should be a function that performs the
10844 Machine specific built-in functions can be useful to expand special machine
10845 instructions that would otherwise not normally be generated because
10846 they have no equivalent in the source language (for example, SIMD vector
10847 instructions or prefetch instructions).
10849 To create a built-in function, call the function
10850 @code{lang_hooks.builtin_function}
10851 which is defined by the language front end. You can use any type nodes set
10852 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
10853 only language front ends that use those two functions will call
10854 @samp{TARGET_INIT_BUILTINS}.
10857 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
10858 Define this hook if you have any machine-specific built-in functions
10859 that need to be defined. It should be a function that returns the
10860 builtin function declaration for the builtin function code @var{code}.
10861 If there is no such builtin and it cannot be initialized at this time
10862 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10863 If @var{code} is out of range the function should return
10864 @code{error_mark_node}.
10867 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10869 Expand a call to a machine specific built-in function that was set up by
10870 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10871 function call; the result should go to @var{target} if that is
10872 convenient, and have mode @var{mode} if that is convenient.
10873 @var{subtarget} may be used as the target for computing one of
10874 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10875 ignored. This function should return the result of the call to the
10879 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
10880 Select a replacement for a machine specific built-in function that
10881 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10882 @emph{before} regular type checking, and so allows the target to
10883 implement a crude form of function overloading. @var{fndecl} is the
10884 declaration of the built-in function. @var{arglist} is the list of
10885 arguments passed to the built-in function. The result is a
10886 complete expression that implements the operation, usually
10887 another @code{CALL_EXPR}.
10888 @var{arglist} really has type @samp{VEC(tree,gc)*}
10891 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
10892 Fold a call to a machine specific built-in function that was set up by
10893 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10894 built-in function. @var{n_args} is the number of arguments passed to
10895 the function; the arguments themselves are pointed to by @var{argp}.
10896 The result is another tree containing a simplified expression for the
10897 call's result. If @var{ignore} is true the value will be ignored.
10900 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const_rtx @var{insn})
10902 Take an instruction in @var{insn} and return NULL if it is valid within a
10903 low-overhead loop, otherwise return a string explaining why doloop
10904 could not be applied.
10906 Many targets use special registers for low-overhead looping. For any
10907 instruction that clobbers these this function should return a string indicating
10908 the reason why the doloop could not be applied.
10909 By default, the RTL loop optimizer does not use a present doloop pattern for
10910 loops containing function calls or branch on table instructions.
10913 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10915 Take a branch insn in @var{branch1} and another in @var{branch2}.
10916 Return true if redirecting @var{branch1} to the destination of
10917 @var{branch2} is possible.
10919 On some targets, branches may have a limited range. Optimizing the
10920 filling of delay slots can result in branches being redirected, and this
10921 may in turn cause a branch offset to overflow.
10924 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
10925 This target hook returns @code{true} if @var{x} is considered to be commutative.
10926 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10927 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10928 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10931 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
10933 When the initial value of a hard register has been copied in a pseudo
10934 register, it is often not necessary to actually allocate another register
10935 to this pseudo register, because the original hard register or a stack slot
10936 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10937 is called at the start of register allocation once for each hard register
10938 that had its initial value copied by using
10939 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10940 Possible values are @code{NULL_RTX}, if you don't want
10941 to do any special allocation, a @code{REG} rtx---that would typically be
10942 the hard register itself, if it is known not to be clobbered---or a
10944 If you are returning a @code{MEM}, this is only a hint for the allocator;
10945 it might decide to use another register anyways.
10946 You may use @code{current_function_leaf_function} in the hook, functions
10947 that use @code{REG_N_SETS}, to determine if the hard
10948 register in question will not be clobbered.
10949 The default value of this hook is @code{NULL}, which disables any special
10953 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
10954 This target hook returns nonzero if @var{x}, an @code{unspec} or
10955 @code{unspec_volatile} operation, might cause a trap. Targets can use
10956 this hook to enhance precision of analysis for @code{unspec} and
10957 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10958 to analyze inner elements of @var{x} in which case @var{flags} should be
10962 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
10963 The compiler invokes this hook whenever it changes its current function
10964 context (@code{cfun}). You can define this function if
10965 the back end needs to perform any initialization or reset actions on a
10966 per-function basis. For example, it may be used to implement function
10967 attributes that affect register usage or code generation patterns.
10968 The argument @var{decl} is the declaration for the new function context,
10969 and may be null to indicate that the compiler has left a function context
10970 and is returning to processing at the top level.
10971 The default hook function does nothing.
10973 GCC sets @code{cfun} to a dummy function context during initialization of
10974 some parts of the back end. The hook function is not invoked in this
10975 situation; you need not worry about the hook being invoked recursively,
10976 or when the back end is in a partially-initialized state.
10977 @code{cfun} might be @code{NULL} to indicate processing at top level,
10978 outside of any function scope.
10981 @defmac TARGET_OBJECT_SUFFIX
10982 Define this macro to be a C string representing the suffix for object
10983 files on your target machine. If you do not define this macro, GCC will
10984 use @samp{.o} as the suffix for object files.
10987 @defmac TARGET_EXECUTABLE_SUFFIX
10988 Define this macro to be a C string representing the suffix to be
10989 automatically added to executable files on your target machine. If you
10990 do not define this macro, GCC will use the null string as the suffix for
10994 @defmac COLLECT_EXPORT_LIST
10995 If defined, @code{collect2} will scan the individual object files
10996 specified on its command line and create an export list for the linker.
10997 Define this macro for systems like AIX, where the linker discards
10998 object files that are not referenced from @code{main} and uses export
11002 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11003 Define this macro to a C expression representing a variant of the
11004 method call @var{mdecl}, if Java Native Interface (JNI) methods
11005 must be invoked differently from other methods on your target.
11006 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11007 the @code{stdcall} calling convention and this macro is then
11008 defined as this expression:
11011 build_type_attribute_variant (@var{mdecl},
11013 (get_identifier ("stdcall"),
11018 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11019 This target hook returns @code{true} past the point in which new jump
11020 instructions could be created. On machines that require a register for
11021 every jump such as the SHmedia ISA of SH5, this point would typically be
11022 reload, so this target hook should be defined to a function such as:
11026 cannot_modify_jumps_past_reload_p ()
11028 return (reload_completed || reload_in_progress);
11033 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11034 This target hook returns a register class for which branch target register
11035 optimizations should be applied. All registers in this class should be
11036 usable interchangeably. After reload, registers in this class will be
11037 re-allocated and loads will be hoisted out of loops and be subjected
11038 to inter-block scheduling.
11041 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11042 Branch target register optimization will by default exclude callee-saved
11044 that are not already live during the current function; if this target hook
11045 returns true, they will be included. The target code must than make sure
11046 that all target registers in the class returned by
11047 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11048 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11049 epilogues have already been generated. Note, even if you only return
11050 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11051 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11052 to reserve space for caller-saved target registers.
11055 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11056 This target hook returns true if the target supports conditional execution.
11057 This target hook is required only when the target has several different
11058 modes and they have different conditional execution capability, such as ARM.
11061 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop})
11062 This target hook returns a new value for the number of times @var{loop}
11063 should be unrolled. The parameter @var{nunroll} is the number of times
11064 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11065 the loop, which is going to be checked for unrolling. This target hook
11066 is required only when the target has special constraints like maximum
11067 number of memory accesses.
11070 @defmac POWI_MAX_MULTS
11071 If defined, this macro is interpreted as a signed integer C expression
11072 that specifies the maximum number of floating point multiplications
11073 that should be emitted when expanding exponentiation by an integer
11074 constant inline. When this value is defined, exponentiation requiring
11075 more than this number of multiplications is implemented by calling the
11076 system library's @code{pow}, @code{powf} or @code{powl} routines.
11077 The default value places no upper bound on the multiplication count.
11080 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11081 This target hook should register any extra include files for the
11082 target. The parameter @var{stdinc} indicates if normal include files
11083 are present. The parameter @var{sysroot} is the system root directory.
11084 The parameter @var{iprefix} is the prefix for the gcc directory.
11087 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11088 This target hook should register any extra include files for the
11089 target before any standard headers. The parameter @var{stdinc}
11090 indicates if normal include files are present. The parameter
11091 @var{sysroot} is the system root directory. The parameter
11092 @var{iprefix} is the prefix for the gcc directory.
11095 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11096 This target hook should register special include paths for the target.
11097 The parameter @var{path} is the include to register. On Darwin
11098 systems, this is used for Framework includes, which have semantics
11099 that are different from @option{-I}.
11102 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11103 This target macro returns @code{true} if it is safe to use a local alias
11104 for a virtual function @var{fndecl} when constructing thunks,
11105 @code{false} otherwise. By default, the macro returns @code{true} for all
11106 functions, if a target supports aliases (i.e.@: defines
11107 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11110 @defmac TARGET_FORMAT_TYPES
11111 If defined, this macro is the name of a global variable containing
11112 target-specific format checking information for the @option{-Wformat}
11113 option. The default is to have no target-specific format checks.
11116 @defmac TARGET_N_FORMAT_TYPES
11117 If defined, this macro is the number of entries in
11118 @code{TARGET_FORMAT_TYPES}.
11121 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11122 If defined, this macro is the name of a global variable containing
11123 target-specific format overrides for the @option{-Wformat} option. The
11124 default is to have no target-specific format overrides. If defined,
11125 @code{TARGET_FORMAT_TYPES} must be defined, too.
11128 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11129 If defined, this macro specifies the number of entries in
11130 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11133 @defmac TARGET_OVERRIDES_FORMAT_INIT
11134 If defined, this macro specifies the optional initialization
11135 routine for target specific customizations of the system printf
11136 and scanf formatter settings.
11139 @deftypevr {Target Hook} bool TARGET_RELAXED_ORDERING
11140 If set to @code{true}, means that the target's memory model does not
11141 guarantee that loads which do not depend on one another will access
11142 main memory in the order of the instruction stream; if ordering is
11143 important, an explicit memory barrier must be used. This is true of
11144 many recent processors which implement a policy of ``relaxed,''
11145 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11146 and ia64. The default is @code{false}.
11149 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11150 If defined, this macro returns the diagnostic message when it is
11151 illegal to pass argument @var{val} to function @var{funcdecl}
11152 with prototype @var{typelist}.
11155 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11156 If defined, this macro returns the diagnostic message when it is
11157 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11158 if validity should be determined by the front end.
11161 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11162 If defined, this macro returns the diagnostic message when it is
11163 invalid to apply operation @var{op} (where unary plus is denoted by
11164 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11165 if validity should be determined by the front end.
11168 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11169 If defined, this macro returns the diagnostic message when it is
11170 invalid to apply operation @var{op} to operands of types @var{type1}
11171 and @var{type2}, or @code{NULL} if validity should be determined by
11175 @deftypefn {Target Hook} {const char *} TARGET_INVALID_PARAMETER_TYPE (const_tree @var{type})
11176 If defined, this macro returns the diagnostic message when it is
11177 invalid for functions to include parameters of type @var{type},
11178 or @code{NULL} if validity should be determined by
11179 the front end. This is currently used only by the C and C++ front ends.
11182 @deftypefn {Target Hook} {const char *} TARGET_INVALID_RETURN_TYPE (const_tree @var{type})
11183 If defined, this macro returns the diagnostic message when it is
11184 invalid for functions to have return type @var{type},
11185 or @code{NULL} if validity should be determined by
11186 the front end. This is currently used only by the C and C++ front ends.
11189 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11190 If defined, this target hook returns the type to which values of
11191 @var{type} should be promoted when they appear in expressions,
11192 analogous to the integer promotions, or @code{NULL_TREE} to use the
11193 front end's normal promotion rules. This hook is useful when there are
11194 target-specific types with special promotion rules.
11195 This is currently used only by the C and C++ front ends.
11198 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11199 If defined, this hook returns the result of converting @var{expr} to
11200 @var{type}. It should return the converted expression,
11201 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11202 This hook is useful when there are target-specific types with special
11204 This is currently used only by the C and C++ front ends.
11207 @defmac TARGET_USE_JCR_SECTION
11208 This macro determines whether to use the JCR section to register Java
11209 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11210 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11214 This macro determines the size of the objective C jump buffer for the
11215 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11218 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11219 Define this macro if any target-specific attributes need to be attached
11220 to the functions in @file{libgcc} that provide low-level support for
11221 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11222 and the associated definitions of those functions.
11225 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11226 Define this macro to update the current function stack boundary if
11230 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11231 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11232 different argument pointer register is needed to access the function's
11233 argument list due to stack realignment. Return @code{NULL} if no DRAP
11237 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11238 When optimization is disabled, this hook indicates whether or not
11239 arguments should be allocated to stack slots. Normally, GCC allocates
11240 stacks slots for arguments when not optimizing in order to make
11241 debugging easier. However, when a function is declared with
11242 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11243 cannot safely move arguments from the registers in which they are passed
11244 to the stack. Therefore, this hook should return true in general, but
11245 false for naked functions. The default implementation always returns true.
11248 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11249 On some architectures it can take multiple instructions to synthesize
11250 a constant. If there is another constant already in a register that
11251 is close enough in value then it is preferable that the new constant
11252 is computed from this register using immediate addition or
11253 subtraction. We accomplish this through CSE. Besides the value of
11254 the constant we also add a lower and an upper constant anchor to the
11255 available expressions. These are then queried when encountering new
11256 constants. The anchors are computed by rounding the constant up and
11257 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11258 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11259 accepted by immediate-add plus one. We currently assume that the
11260 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11261 MIPS, where add-immediate takes a 16-bit signed value,
11262 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11263 is zero, which disables this optimization. @end deftypevr