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 the 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 @hook 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 @hook TARGET_HANDLE_OPTION
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 @hook TARGET_HANDLE_C_OPTION
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 @hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT
705 @hook TARGET_STRING_OBJECT_REF_TYPE_P
707 @hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG
709 @hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE
710 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
711 but is called when the optimize level is changed via an attribute or
712 pragma or when it is reset at the end of the code affected by the
713 attribute or pragma. It is not called at the beginning of compilation
714 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
715 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
716 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
719 @defmac C_COMMON_OVERRIDE_OPTIONS
720 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
721 but is only used in the C
722 language frontends (C, Objective-C, C++, Objective-C++) and so can be
723 used to alter option flag variables which only exist in those
727 @hook TARGET_OPTION_OPTIMIZATION_TABLE
728 Some machines may desire to change what optimizations are performed for
729 various optimization levels. This variable, if defined, describes
730 options to enable at particular sets of optimization levels. These
731 options are processed once
732 just after the optimization level is determined and before the remainder
733 of the command options have been parsed, so may be overridden by other
734 options passed explicitly.
736 This processing is run once at program startup and when the optimization
737 options are changed via @code{#pragma GCC optimize} or by using the
738 @code{optimize} attribute.
741 @hook TARGET_OPTION_INIT_STRUCT
743 @hook TARGET_OPTION_DEFAULT_PARAMS
746 This hook is called in response to the user invoking
747 @option{--target-help} on the command line. It gives the target a
748 chance to display extra information on the target specific command
749 line options found in its @file{.opt} file.
752 @defmac SWITCHABLE_TARGET
753 Some targets need to switch between substantially different subtargets
754 during compilation. For example, the MIPS target has one subtarget for
755 the traditional MIPS architecture and another for MIPS16. Source code
756 can switch between these two subarchitectures using the @code{mips16}
757 and @code{nomips16} attributes.
759 Such subtargets can differ in things like the set of available
760 registers, the set of available instructions, the costs of various
761 operations, and so on. GCC caches a lot of this type of information
762 in global variables, and recomputing them for each subtarget takes a
763 significant amount of time. The compiler therefore provides a facility
764 for maintaining several versions of the global variables and quickly
765 switching between them; see @file{target-globals.h} for details.
767 Define this macro to 1 if your target needs this facility. The default
771 @node Per-Function Data
772 @section Defining data structures for per-function information.
773 @cindex per-function data
774 @cindex data structures
776 If the target needs to store information on a per-function basis, GCC
777 provides a macro and a couple of variables to allow this. Note, just
778 using statics to store the information is a bad idea, since GCC supports
779 nested functions, so you can be halfway through encoding one function
780 when another one comes along.
782 GCC defines a data structure called @code{struct function} which
783 contains all of the data specific to an individual function. This
784 structure contains a field called @code{machine} whose type is
785 @code{struct machine_function *}, which can be used by targets to point
786 to their own specific data.
788 If a target needs per-function specific data it should define the type
789 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
790 This macro should be used to initialize the function pointer
791 @code{init_machine_status}. This pointer is explained below.
793 One typical use of per-function, target specific data is to create an
794 RTX to hold the register containing the function's return address. This
795 RTX can then be used to implement the @code{__builtin_return_address}
796 function, for level 0.
798 Note---earlier implementations of GCC used a single data area to hold
799 all of the per-function information. Thus when processing of a nested
800 function began the old per-function data had to be pushed onto a
801 stack, and when the processing was finished, it had to be popped off the
802 stack. GCC used to provide function pointers called
803 @code{save_machine_status} and @code{restore_machine_status} to handle
804 the saving and restoring of the target specific information. Since the
805 single data area approach is no longer used, these pointers are no
808 @defmac INIT_EXPANDERS
809 Macro called to initialize any target specific information. This macro
810 is called once per function, before generation of any RTL has begun.
811 The intention of this macro is to allow the initialization of the
812 function pointer @code{init_machine_status}.
815 @deftypevar {void (*)(struct function *)} init_machine_status
816 If this function pointer is non-@code{NULL} it will be called once per
817 function, before function compilation starts, in order to allow the
818 target to perform any target specific initialization of the
819 @code{struct function} structure. It is intended that this would be
820 used to initialize the @code{machine} of that structure.
822 @code{struct machine_function} structures are expected to be freed by GC@.
823 Generally, any memory that they reference must be allocated by using
824 GC allocation, including the structure itself.
828 @section Storage Layout
829 @cindex storage layout
831 Note that the definitions of the macros in this table which are sizes or
832 alignments measured in bits do not need to be constant. They can be C
833 expressions that refer to static variables, such as the @code{target_flags}.
834 @xref{Run-time Target}.
836 @defmac BITS_BIG_ENDIAN
837 Define this macro to have the value 1 if the most significant bit in a
838 byte has the lowest number; otherwise define it to have the value zero.
839 This means that bit-field instructions count from the most significant
840 bit. If the machine has no bit-field instructions, then this must still
841 be defined, but it doesn't matter which value it is defined to. This
842 macro need not be a constant.
844 This macro does not affect the way structure fields are packed into
845 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
848 @defmac BYTES_BIG_ENDIAN
849 Define this macro to have the value 1 if the most significant byte in a
850 word has the lowest number. This macro need not be a constant.
853 @defmac WORDS_BIG_ENDIAN
854 Define this macro to have the value 1 if, in a multiword object, the
855 most significant word has the lowest number. This applies to both
856 memory locations and registers; GCC fundamentally assumes that the
857 order of words in memory is the same as the order in registers. This
858 macro need not be a constant.
861 @defmac FLOAT_WORDS_BIG_ENDIAN
862 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
863 @code{TFmode} floating point numbers are stored in memory with the word
864 containing the sign bit at the lowest address; otherwise define it to
865 have the value 0. This macro need not be a constant.
867 You need not define this macro if the ordering is the same as for
871 @defmac BITS_PER_UNIT
872 Define this macro to be the number of bits in an addressable storage
873 unit (byte). If you do not define this macro the default is 8.
876 @defmac BITS_PER_WORD
877 Number of bits in a word. If you do not define this macro, the default
878 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
881 @defmac MAX_BITS_PER_WORD
882 Maximum number of bits in a word. If this is undefined, the default is
883 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
884 largest value that @code{BITS_PER_WORD} can have at run-time.
887 @defmac UNITS_PER_WORD
888 Number of storage units in a word; normally the size of a general-purpose
889 register, a power of two from 1 or 8.
892 @defmac MIN_UNITS_PER_WORD
893 Minimum number of units in a word. If this is undefined, the default is
894 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
895 smallest value that @code{UNITS_PER_WORD} can have at run-time.
899 Width of a pointer, in bits. You must specify a value no wider than the
900 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
901 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
902 a value the default is @code{BITS_PER_WORD}.
905 @defmac POINTERS_EXTEND_UNSIGNED
906 A C expression that determines how pointers should be extended from
907 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
908 greater than zero if pointers should be zero-extended, zero if they
909 should be sign-extended, and negative if some other sort of conversion
910 is needed. In the last case, the extension is done by the target's
911 @code{ptr_extend} instruction.
913 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
914 and @code{word_mode} are all the same width.
917 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
918 A macro to update @var{m} and @var{unsignedp} when an object whose type
919 is @var{type} and which has the specified mode and signedness is to be
920 stored in a register. This macro is only called when @var{type} is a
923 On most RISC machines, which only have operations that operate on a full
924 register, define this macro to set @var{m} to @code{word_mode} if
925 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
926 cases, only integer modes should be widened because wider-precision
927 floating-point operations are usually more expensive than their narrower
930 For most machines, the macro definition does not change @var{unsignedp}.
931 However, some machines, have instructions that preferentially handle
932 either signed or unsigned quantities of certain modes. For example, on
933 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
934 sign-extend the result to 64 bits. On such machines, set
935 @var{unsignedp} according to which kind of extension is more efficient.
937 Do not define this macro if it would never modify @var{m}.
940 @hook TARGET_PROMOTE_FUNCTION_MODE
941 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
942 function return values. The target hook should return the new mode
943 and possibly change @code{*@var{punsignedp}} if the promotion should
944 change signedness. This function is called only for scalar @emph{or
947 @var{for_return} allows to distinguish the promotion of arguments and
948 return values. If it is @code{1}, a return value is being promoted and
949 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
950 If it is @code{2}, the returned mode should be that of the register in
951 which an incoming parameter is copied, or the outgoing result is computed;
952 then the hook should return the same mode as @code{promote_mode}, though
953 the signedness may be different.
955 The default is to not promote arguments and return values. You can
956 also define the hook to @code{default_promote_function_mode_always_promote}
957 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
960 @defmac PARM_BOUNDARY
961 Normal alignment required for function parameters on the stack, in
962 bits. All stack parameters receive at least this much alignment
963 regardless of data type. On most machines, this is the same as the
967 @defmac STACK_BOUNDARY
968 Define this macro to the minimum alignment enforced by hardware for the
969 stack pointer on this machine. The definition is a C expression for the
970 desired alignment (measured in bits). This value is used as a default
971 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
972 this should be the same as @code{PARM_BOUNDARY}.
975 @defmac PREFERRED_STACK_BOUNDARY
976 Define this macro if you wish to preserve a certain alignment for the
977 stack pointer, greater than what the hardware enforces. The definition
978 is a C expression for the desired alignment (measured in bits). This
979 macro must evaluate to a value equal to or larger than
980 @code{STACK_BOUNDARY}.
983 @defmac INCOMING_STACK_BOUNDARY
984 Define this macro if the incoming stack boundary may be different
985 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
986 to a value equal to or larger than @code{STACK_BOUNDARY}.
989 @defmac FUNCTION_BOUNDARY
990 Alignment required for a function entry point, in bits.
993 @defmac BIGGEST_ALIGNMENT
994 Biggest alignment that any data type can require on this machine, in
995 bits. Note that this is not the biggest alignment that is supported,
996 just the biggest alignment that, when violated, may cause a fault.
999 @defmac MALLOC_ABI_ALIGNMENT
1000 Alignment, in bits, a C conformant malloc implementation has to
1001 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1004 @defmac ATTRIBUTE_ALIGNED_VALUE
1005 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1006 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1009 @defmac MINIMUM_ATOMIC_ALIGNMENT
1010 If defined, the smallest alignment, in bits, that can be given to an
1011 object that can be referenced in one operation, without disturbing any
1012 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1013 on machines that don't have byte or half-word store operations.
1016 @defmac BIGGEST_FIELD_ALIGNMENT
1017 Biggest alignment that any structure or union field can require on this
1018 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1019 structure and union fields only, unless the field alignment has been set
1020 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1023 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1024 An expression for the alignment of a structure field @var{field} if the
1025 alignment computed in the usual way (including applying of
1026 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1027 alignment) is @var{computed}. It overrides alignment only if the
1028 field alignment has not been set by the
1029 @code{__attribute__ ((aligned (@var{n})))} construct.
1032 @defmac MAX_STACK_ALIGNMENT
1033 Biggest stack alignment guaranteed by the backend. Use this macro
1034 to specify the maximum alignment of a variable on stack.
1036 If not defined, the default value is @code{STACK_BOUNDARY}.
1038 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1039 @c But the fix for PR 32893 indicates that we can only guarantee
1040 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1041 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1044 @defmac MAX_OFILE_ALIGNMENT
1045 Biggest alignment supported by the object file format of this machine.
1046 Use this macro to limit the alignment which can be specified using the
1047 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1048 the default value is @code{BIGGEST_ALIGNMENT}.
1050 On systems that use ELF, the default (in @file{config/elfos.h}) is
1051 the largest supported 32-bit ELF section alignment representable on
1052 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1053 On 32-bit ELF the largest supported section alignment in bits is
1054 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1057 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1058 If defined, a C expression to compute the alignment for a variable in
1059 the static store. @var{type} is the data type, and @var{basic-align} is
1060 the alignment that the object would ordinarily have. The value of this
1061 macro is used instead of that alignment to align the object.
1063 If this macro is not defined, then @var{basic-align} is used.
1066 One use of this macro is to increase alignment of medium-size data to
1067 make it all fit in fewer cache lines. Another is to cause character
1068 arrays to be word-aligned so that @code{strcpy} calls that copy
1069 constants to character arrays can be done inline.
1072 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1073 If defined, a C expression to compute the alignment given to a constant
1074 that is being placed in memory. @var{constant} is the constant and
1075 @var{basic-align} is the alignment that the object would ordinarily
1076 have. The value of this macro is used instead of that alignment to
1079 If this macro is not defined, then @var{basic-align} is used.
1081 The typical use of this macro is to increase alignment for string
1082 constants to be word aligned so that @code{strcpy} calls that copy
1083 constants can be done inline.
1086 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1087 If defined, a C expression to compute the alignment for a variable in
1088 the local store. @var{type} is the data type, and @var{basic-align} is
1089 the alignment that the object would ordinarily have. The value of this
1090 macro is used instead of that alignment to align the object.
1092 If this macro is not defined, then @var{basic-align} is used.
1094 One use of this macro is to increase alignment of medium-size data to
1095 make it all fit in fewer cache lines.
1097 If the value of this macro has a type, it should be an unsigned type.
1100 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1101 If defined, a C expression to compute the alignment for stack slot.
1102 @var{type} is the data type, @var{mode} is the widest mode available,
1103 and @var{basic-align} is the alignment that the slot would ordinarily
1104 have. The value of this macro is used instead of that alignment to
1107 If this macro is not defined, then @var{basic-align} is used when
1108 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1111 This macro is to set alignment of stack slot to the maximum alignment
1112 of all possible modes which the slot may have.
1114 If the value of this macro has a type, it should be an unsigned type.
1117 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1118 If defined, a C expression to compute the alignment for a local
1119 variable @var{decl}.
1121 If this macro is not defined, then
1122 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1125 One use of this macro is to increase alignment of medium-size data to
1126 make it all fit in fewer cache lines.
1128 If the value of this macro has a type, it should be an unsigned type.
1131 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1132 If defined, a C expression to compute the minimum required alignment
1133 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1134 @var{mode}, assuming normal alignment @var{align}.
1136 If this macro is not defined, then @var{align} will be used.
1139 @defmac EMPTY_FIELD_BOUNDARY
1140 Alignment in bits to be given to a structure bit-field that follows an
1141 empty field such as @code{int : 0;}.
1143 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1146 @defmac STRUCTURE_SIZE_BOUNDARY
1147 Number of bits which any structure or union's size must be a multiple of.
1148 Each structure or union's size is rounded up to a multiple of this.
1150 If you do not define this macro, the default is the same as
1151 @code{BITS_PER_UNIT}.
1154 @defmac STRICT_ALIGNMENT
1155 Define this macro to be the value 1 if instructions will fail to work
1156 if given data not on the nominal alignment. If instructions will merely
1157 go slower in that case, define this macro as 0.
1160 @defmac PCC_BITFIELD_TYPE_MATTERS
1161 Define this if you wish to imitate the way many other C compilers handle
1162 alignment of bit-fields and the structures that contain them.
1164 The behavior is that the type written for a named bit-field (@code{int},
1165 @code{short}, or other integer type) imposes an alignment for the entire
1166 structure, as if the structure really did contain an ordinary field of
1167 that type. In addition, the bit-field is placed within the structure so
1168 that it would fit within such a field, not crossing a boundary for it.
1170 Thus, on most machines, a named bit-field whose type is written as
1171 @code{int} would not cross a four-byte boundary, and would force
1172 four-byte alignment for the whole structure. (The alignment used may
1173 not be four bytes; it is controlled by the other alignment parameters.)
1175 An unnamed bit-field will not affect the alignment of the containing
1178 If the macro is defined, its definition should be a C expression;
1179 a nonzero value for the expression enables this behavior.
1181 Note that if this macro is not defined, or its value is zero, some
1182 bit-fields may cross more than one alignment boundary. The compiler can
1183 support such references if there are @samp{insv}, @samp{extv}, and
1184 @samp{extzv} insns that can directly reference memory.
1186 The other known way of making bit-fields work is to define
1187 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1188 Then every structure can be accessed with fullwords.
1190 Unless the machine has bit-field instructions or you define
1191 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1192 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1194 If your aim is to make GCC use the same conventions for laying out
1195 bit-fields as are used by another compiler, here is how to investigate
1196 what the other compiler does. Compile and run this program:
1215 printf ("Size of foo1 is %d\n",
1216 sizeof (struct foo1));
1217 printf ("Size of foo2 is %d\n",
1218 sizeof (struct foo2));
1223 If this prints 2 and 5, then the compiler's behavior is what you would
1224 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1227 @defmac BITFIELD_NBYTES_LIMITED
1228 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1229 to aligning a bit-field within the structure.
1232 @hook TARGET_ALIGN_ANON_BITFIELD
1233 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1234 whether unnamed bitfields affect the alignment of the containing
1235 structure. The hook should return true if the structure should inherit
1236 the alignment requirements of an unnamed bitfield's type.
1239 @hook TARGET_NARROW_VOLATILE_BITFIELD
1240 This target hook should return @code{true} if accesses to volatile bitfields
1241 should use the narrowest mode possible. It should return @code{false} if
1242 these accesses should use the bitfield container type.
1244 The default is @code{!TARGET_STRICT_ALIGN}.
1247 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1248 Return 1 if a structure or array containing @var{field} should be accessed using
1251 If @var{field} is the only field in the structure, @var{mode} is its
1252 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1253 case where structures of one field would require the structure's mode to
1254 retain the field's mode.
1256 Normally, this is not needed.
1259 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1260 Define this macro as an expression for the alignment of a type (given
1261 by @var{type} as a tree node) if the alignment computed in the usual
1262 way is @var{computed} and the alignment explicitly specified was
1265 The default is to use @var{specified} if it is larger; otherwise, use
1266 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1269 @defmac MAX_FIXED_MODE_SIZE
1270 An integer expression for the size in bits of the largest integer
1271 machine mode that should actually be used. All integer machine modes of
1272 this size or smaller can be used for structures and unions with the
1273 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1274 (DImode)} is assumed.
1277 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1278 If defined, an expression of type @code{enum machine_mode} that
1279 specifies the mode of the save area operand of a
1280 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1281 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1282 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1283 having its mode specified.
1285 You need not define this macro if it always returns @code{Pmode}. You
1286 would most commonly define this macro if the
1287 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1291 @defmac STACK_SIZE_MODE
1292 If defined, an expression of type @code{enum machine_mode} that
1293 specifies the mode of the size increment operand of an
1294 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1296 You need not define this macro if it always returns @code{word_mode}.
1297 You would most commonly define this macro if the @code{allocate_stack}
1298 pattern needs to support both a 32- and a 64-bit mode.
1301 @hook TARGET_LIBGCC_CMP_RETURN_MODE
1302 This target hook should return the mode to be used for the return value
1303 of compare instructions expanded to libgcc calls. If not defined
1304 @code{word_mode} is returned which is the right choice for a majority of
1308 @hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1309 This target hook should return the mode to be used for the shift count operand
1310 of shift instructions expanded to libgcc calls. If not defined
1311 @code{word_mode} is returned which is the right choice for a majority of
1315 @hook TARGET_UNWIND_WORD_MODE
1316 Return machine mode to be used for @code{_Unwind_Word} type.
1317 The default is to use @code{word_mode}.
1320 @defmac ROUND_TOWARDS_ZERO
1321 If defined, this macro should be true if the prevailing rounding
1322 mode is towards zero.
1324 Defining this macro only affects the way @file{libgcc.a} emulates
1325 floating-point arithmetic.
1327 Not defining this macro is equivalent to returning zero.
1330 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1331 This macro should return true if floats with @var{size}
1332 bits do not have a NaN or infinity representation, but use the largest
1333 exponent for normal numbers instead.
1335 Defining this macro only affects the way @file{libgcc.a} emulates
1336 floating-point arithmetic.
1338 The default definition of this macro returns false for all sizes.
1341 @hook TARGET_MS_BITFIELD_LAYOUT_P
1342 This target hook returns @code{true} if bit-fields in the given
1343 @var{record_type} are to be laid out following the rules of Microsoft
1344 Visual C/C++, namely: (i) a bit-field won't share the same storage
1345 unit with the previous bit-field if their underlying types have
1346 different sizes, and the bit-field will be aligned to the highest
1347 alignment of the underlying types of itself and of the previous
1348 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1349 the whole enclosing structure, even if it is unnamed; except that
1350 (iii) a zero-sized bit-field will be disregarded unless it follows
1351 another bit-field of nonzero size. If this hook returns @code{true},
1352 other macros that control bit-field layout are ignored.
1354 When a bit-field is inserted into a packed record, the whole size
1355 of the underlying type is used by one or more same-size adjacent
1356 bit-fields (that is, if its long:3, 32 bits is used in the record,
1357 and any additional adjacent long bit-fields are packed into the same
1358 chunk of 32 bits. However, if the size changes, a new field of that
1359 size is allocated). In an unpacked record, this is the same as using
1360 alignment, but not equivalent when packing.
1362 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1363 the latter will take precedence. If @samp{__attribute__((packed))} is
1364 used on a single field when MS bit-fields are in use, it will take
1365 precedence for that field, but the alignment of the rest of the structure
1366 may affect its placement.
1369 @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1370 Returns true if the target supports decimal floating point.
1373 @hook TARGET_FIXED_POINT_SUPPORTED_P
1374 Returns true if the target supports fixed-point arithmetic.
1377 @hook TARGET_EXPAND_TO_RTL_HOOK
1378 This hook is called just before expansion into rtl, allowing the target
1379 to perform additional initializations or analysis before the expansion.
1380 For example, the rs6000 port uses it to allocate a scratch stack slot
1381 for use in copying SDmode values between memory and floating point
1382 registers whenever the function being expanded has any SDmode
1386 @hook TARGET_INSTANTIATE_DECLS
1387 This hook allows the backend to perform additional instantiations on rtl
1388 that are not actually in any insns yet, but will be later.
1391 @hook TARGET_MANGLE_TYPE
1392 If your target defines any fundamental types, or any types your target
1393 uses should be mangled differently from the default, define this hook
1394 to return the appropriate encoding for these types as part of a C++
1395 mangled name. The @var{type} argument is the tree structure representing
1396 the type to be mangled. The hook may be applied to trees which are
1397 not target-specific fundamental types; it should return @code{NULL}
1398 for all such types, as well as arguments it does not recognize. If the
1399 return value is not @code{NULL}, it must point to a statically-allocated
1402 Target-specific fundamental types might be new fundamental types or
1403 qualified versions of ordinary fundamental types. Encode new
1404 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1405 is the name used for the type in source code, and @var{n} is the
1406 length of @var{name} in decimal. Encode qualified versions of
1407 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1408 @var{name} is the name used for the type qualifier in source code,
1409 @var{n} is the length of @var{name} as above, and @var{code} is the
1410 code used to represent the unqualified version of this type. (See
1411 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1412 codes.) In both cases the spaces are for clarity; do not include any
1413 spaces in your string.
1415 This hook is applied to types prior to typedef resolution. If the mangled
1416 name for a particular type depends only on that type's main variant, you
1417 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1420 The default version of this hook always returns @code{NULL}, which is
1421 appropriate for a target that does not define any new fundamental
1426 @section Layout of Source Language Data Types
1428 These macros define the sizes and other characteristics of the standard
1429 basic data types used in programs being compiled. Unlike the macros in
1430 the previous section, these apply to specific features of C and related
1431 languages, rather than to fundamental aspects of storage layout.
1433 @defmac INT_TYPE_SIZE
1434 A C expression for the size in bits of the type @code{int} on the
1435 target machine. If you don't define this, the default is one word.
1438 @defmac SHORT_TYPE_SIZE
1439 A C expression for the size in bits of the type @code{short} on the
1440 target machine. If you don't define this, the default is half a word.
1441 (If this would be less than one storage unit, it is rounded up to one
1445 @defmac LONG_TYPE_SIZE
1446 A C expression for the size in bits of the type @code{long} on the
1447 target machine. If you don't define this, the default is one word.
1450 @defmac ADA_LONG_TYPE_SIZE
1451 On some machines, the size used for the Ada equivalent of the type
1452 @code{long} by a native Ada compiler differs from that used by C@. In
1453 that situation, define this macro to be a C expression to be used for
1454 the size of that type. If you don't define this, the default is the
1455 value of @code{LONG_TYPE_SIZE}.
1458 @defmac LONG_LONG_TYPE_SIZE
1459 A C expression for the size in bits of the type @code{long long} on the
1460 target machine. If you don't define this, the default is two
1461 words. If you want to support GNU Ada on your machine, the value of this
1462 macro must be at least 64.
1465 @defmac CHAR_TYPE_SIZE
1466 A C expression for the size in bits of the type @code{char} on the
1467 target machine. If you don't define this, the default is
1468 @code{BITS_PER_UNIT}.
1471 @defmac BOOL_TYPE_SIZE
1472 A C expression for the size in bits of the C++ type @code{bool} and
1473 C99 type @code{_Bool} on the target machine. If you don't define
1474 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1477 @defmac FLOAT_TYPE_SIZE
1478 A C expression for the size in bits of the type @code{float} on the
1479 target machine. If you don't define this, the default is one word.
1482 @defmac DOUBLE_TYPE_SIZE
1483 A C expression for the size in bits of the type @code{double} on the
1484 target machine. If you don't define this, the default is two
1488 @defmac LONG_DOUBLE_TYPE_SIZE
1489 A C expression for the size in bits of the type @code{long double} on
1490 the target machine. If you don't define this, the default is two
1494 @defmac SHORT_FRACT_TYPE_SIZE
1495 A C expression for the size in bits of the type @code{short _Fract} on
1496 the target machine. If you don't define this, the default is
1497 @code{BITS_PER_UNIT}.
1500 @defmac FRACT_TYPE_SIZE
1501 A C expression for the size in bits of the type @code{_Fract} on
1502 the target machine. If you don't define this, the default is
1503 @code{BITS_PER_UNIT * 2}.
1506 @defmac LONG_FRACT_TYPE_SIZE
1507 A C expression for the size in bits of the type @code{long _Fract} on
1508 the target machine. If you don't define this, the default is
1509 @code{BITS_PER_UNIT * 4}.
1512 @defmac LONG_LONG_FRACT_TYPE_SIZE
1513 A C expression for the size in bits of the type @code{long long _Fract} on
1514 the target machine. If you don't define this, the default is
1515 @code{BITS_PER_UNIT * 8}.
1518 @defmac SHORT_ACCUM_TYPE_SIZE
1519 A C expression for the size in bits of the type @code{short _Accum} on
1520 the target machine. If you don't define this, the default is
1521 @code{BITS_PER_UNIT * 2}.
1524 @defmac ACCUM_TYPE_SIZE
1525 A C expression for the size in bits of the type @code{_Accum} on
1526 the target machine. If you don't define this, the default is
1527 @code{BITS_PER_UNIT * 4}.
1530 @defmac LONG_ACCUM_TYPE_SIZE
1531 A C expression for the size in bits of the type @code{long _Accum} on
1532 the target machine. If you don't define this, the default is
1533 @code{BITS_PER_UNIT * 8}.
1536 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1537 A C expression for the size in bits of the type @code{long long _Accum} on
1538 the target machine. If you don't define this, the default is
1539 @code{BITS_PER_UNIT * 16}.
1542 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1543 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1544 if you want routines in @file{libgcc2.a} for a size other than
1545 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1546 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1549 @defmac LIBGCC2_HAS_DF_MODE
1550 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1551 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1552 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1553 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1554 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1558 @defmac LIBGCC2_HAS_XF_MODE
1559 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1560 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1561 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1562 is 80 then the default is 1, otherwise it is 0.
1565 @defmac LIBGCC2_HAS_TF_MODE
1566 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1567 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1568 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1569 is 128 then the default is 1, otherwise it is 0.
1576 Define these macros to be the size in bits of the mantissa of
1577 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1578 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1579 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1580 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1581 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1582 @code{DOUBLE_TYPE_SIZE} or
1583 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1586 @defmac TARGET_FLT_EVAL_METHOD
1587 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1588 assuming, if applicable, that the floating-point control word is in its
1589 default state. If you do not define this macro the value of
1590 @code{FLT_EVAL_METHOD} will be zero.
1593 @defmac WIDEST_HARDWARE_FP_SIZE
1594 A C expression for the size in bits of the widest floating-point format
1595 supported by the hardware. If you define this macro, you must specify a
1596 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1597 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1601 @defmac DEFAULT_SIGNED_CHAR
1602 An expression whose value is 1 or 0, according to whether the type
1603 @code{char} should be signed or unsigned by default. The user can
1604 always override this default with the options @option{-fsigned-char}
1605 and @option{-funsigned-char}.
1608 @hook TARGET_DEFAULT_SHORT_ENUMS
1609 This target hook should return true if the compiler should give an
1610 @code{enum} type only as many bytes as it takes to represent the range
1611 of possible values of that type. It should return false if all
1612 @code{enum} types should be allocated like @code{int}.
1614 The default is to return false.
1618 A C expression for a string describing the name of the data type to use
1619 for size values. The typedef name @code{size_t} is defined using the
1620 contents of the string.
1622 The string can contain more than one keyword. If so, separate them with
1623 spaces, and write first any length keyword, then @code{unsigned} if
1624 appropriate, and finally @code{int}. The string must exactly match one
1625 of the data type names defined in the function
1626 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1627 omit @code{int} or change the order---that would cause the compiler to
1630 If you don't define this macro, the default is @code{"long unsigned
1634 @defmac PTRDIFF_TYPE
1635 A C expression for a string describing the name of the data type to use
1636 for the result of subtracting two pointers. The typedef name
1637 @code{ptrdiff_t} is defined using the contents of the string. See
1638 @code{SIZE_TYPE} above for more information.
1640 If you don't define this macro, the default is @code{"long int"}.
1644 A C expression for a string describing the name of the data type to use
1645 for wide characters. The typedef name @code{wchar_t} is defined using
1646 the contents of the string. See @code{SIZE_TYPE} above for more
1649 If you don't define this macro, the default is @code{"int"}.
1652 @defmac WCHAR_TYPE_SIZE
1653 A C expression for the size in bits of the data type for wide
1654 characters. This is used in @code{cpp}, which cannot make use of
1659 A C expression for a string describing the name of the data type to
1660 use for wide characters passed to @code{printf} and returned from
1661 @code{getwc}. The typedef name @code{wint_t} is defined using the
1662 contents of the string. See @code{SIZE_TYPE} above for more
1665 If you don't define this macro, the default is @code{"unsigned int"}.
1669 A C expression for a string describing the name of the data type that
1670 can represent any value of any standard or extended signed integer type.
1671 The typedef name @code{intmax_t} is defined using the contents of the
1672 string. See @code{SIZE_TYPE} above for more information.
1674 If you don't define this macro, the default is the first of
1675 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1676 much precision as @code{long long int}.
1679 @defmac UINTMAX_TYPE
1680 A C expression for a string describing the name of the data type that
1681 can represent any value of any standard or extended unsigned integer
1682 type. The typedef name @code{uintmax_t} is defined using the contents
1683 of the string. See @code{SIZE_TYPE} above for more information.
1685 If you don't define this macro, the default is the first of
1686 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1687 unsigned int"} that has as much precision as @code{long long unsigned
1691 @defmac SIG_ATOMIC_TYPE
1697 @defmacx UINT16_TYPE
1698 @defmacx UINT32_TYPE
1699 @defmacx UINT64_TYPE
1700 @defmacx INT_LEAST8_TYPE
1701 @defmacx INT_LEAST16_TYPE
1702 @defmacx INT_LEAST32_TYPE
1703 @defmacx INT_LEAST64_TYPE
1704 @defmacx UINT_LEAST8_TYPE
1705 @defmacx UINT_LEAST16_TYPE
1706 @defmacx UINT_LEAST32_TYPE
1707 @defmacx UINT_LEAST64_TYPE
1708 @defmacx INT_FAST8_TYPE
1709 @defmacx INT_FAST16_TYPE
1710 @defmacx INT_FAST32_TYPE
1711 @defmacx INT_FAST64_TYPE
1712 @defmacx UINT_FAST8_TYPE
1713 @defmacx UINT_FAST16_TYPE
1714 @defmacx UINT_FAST32_TYPE
1715 @defmacx UINT_FAST64_TYPE
1716 @defmacx INTPTR_TYPE
1717 @defmacx UINTPTR_TYPE
1718 C expressions for the standard types @code{sig_atomic_t},
1719 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1720 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1721 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1722 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1723 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1724 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1725 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1726 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1727 @code{SIZE_TYPE} above for more information.
1729 If any of these macros evaluates to a null pointer, the corresponding
1730 type is not supported; if GCC is configured to provide
1731 @code{<stdint.h>} in such a case, the header provided may not conform
1732 to C99, depending on the type in question. The defaults for all of
1733 these macros are null pointers.
1736 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1737 The C++ compiler represents a pointer-to-member-function with a struct
1744 ptrdiff_t vtable_index;
1751 The C++ compiler must use one bit to indicate whether the function that
1752 will be called through a pointer-to-member-function is virtual.
1753 Normally, we assume that the low-order bit of a function pointer must
1754 always be zero. Then, by ensuring that the vtable_index is odd, we can
1755 distinguish which variant of the union is in use. But, on some
1756 platforms function pointers can be odd, and so this doesn't work. In
1757 that case, we use the low-order bit of the @code{delta} field, and shift
1758 the remainder of the @code{delta} field to the left.
1760 GCC will automatically make the right selection about where to store
1761 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1762 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1763 set such that functions always start at even addresses, but the lowest
1764 bit of pointers to functions indicate whether the function at that
1765 address is in ARM or Thumb mode. If this is the case of your
1766 architecture, you should define this macro to
1767 @code{ptrmemfunc_vbit_in_delta}.
1769 In general, you should not have to define this macro. On architectures
1770 in which function addresses are always even, according to
1771 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1772 @code{ptrmemfunc_vbit_in_pfn}.
1775 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1776 Normally, the C++ compiler uses function pointers in vtables. This
1777 macro allows the target to change to use ``function descriptors''
1778 instead. Function descriptors are found on targets for whom a
1779 function pointer is actually a small data structure. Normally the
1780 data structure consists of the actual code address plus a data
1781 pointer to which the function's data is relative.
1783 If vtables are used, the value of this macro should be the number
1784 of words that the function descriptor occupies.
1787 @defmac TARGET_VTABLE_ENTRY_ALIGN
1788 By default, the vtable entries are void pointers, the so the alignment
1789 is the same as pointer alignment. The value of this macro specifies
1790 the alignment of the vtable entry in bits. It should be defined only
1791 when special alignment is necessary. */
1794 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1795 There are a few non-descriptor entries in the vtable at offsets below
1796 zero. If these entries must be padded (say, to preserve the alignment
1797 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1798 of words in each data entry.
1802 @section Register Usage
1803 @cindex register usage
1805 This section explains how to describe what registers the target machine
1806 has, and how (in general) they can be used.
1808 The description of which registers a specific instruction can use is
1809 done with register classes; see @ref{Register Classes}. For information
1810 on using registers to access a stack frame, see @ref{Frame Registers}.
1811 For passing values in registers, see @ref{Register Arguments}.
1812 For returning values in registers, see @ref{Scalar Return}.
1815 * Register Basics:: Number and kinds of registers.
1816 * Allocation Order:: Order in which registers are allocated.
1817 * Values in Registers:: What kinds of values each reg can hold.
1818 * Leaf Functions:: Renumbering registers for leaf functions.
1819 * Stack Registers:: Handling a register stack such as 80387.
1822 @node Register Basics
1823 @subsection Basic Characteristics of Registers
1825 @c prevent bad page break with this line
1826 Registers have various characteristics.
1828 @defmac FIRST_PSEUDO_REGISTER
1829 Number of hardware registers known to the compiler. They receive
1830 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1831 pseudo register's number really is assigned the number
1832 @code{FIRST_PSEUDO_REGISTER}.
1835 @defmac FIXED_REGISTERS
1836 @cindex fixed register
1837 An initializer that says which registers are used for fixed purposes
1838 all throughout the compiled code and are therefore not available for
1839 general allocation. These would include the stack pointer, the frame
1840 pointer (except on machines where that can be used as a general
1841 register when no frame pointer is needed), the program counter on
1842 machines where that is considered one of the addressable registers,
1843 and any other numbered register with a standard use.
1845 This information is expressed as a sequence of numbers, separated by
1846 commas and surrounded by braces. The @var{n}th number is 1 if
1847 register @var{n} is fixed, 0 otherwise.
1849 The table initialized from this macro, and the table initialized by
1850 the following one, may be overridden at run time either automatically,
1851 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1852 the user with the command options @option{-ffixed-@var{reg}},
1853 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1856 @defmac CALL_USED_REGISTERS
1857 @cindex call-used register
1858 @cindex call-clobbered register
1859 @cindex call-saved register
1860 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1861 clobbered (in general) by function calls as well as for fixed
1862 registers. This macro therefore identifies the registers that are not
1863 available for general allocation of values that must live across
1866 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1867 automatically saves it on function entry and restores it on function
1868 exit, if the register is used within the function.
1871 @defmac CALL_REALLY_USED_REGISTERS
1872 @cindex call-used register
1873 @cindex call-clobbered register
1874 @cindex call-saved register
1875 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1876 that the entire set of @code{FIXED_REGISTERS} be included.
1877 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1878 This macro is optional. If not specified, it defaults to the value
1879 of @code{CALL_USED_REGISTERS}.
1882 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1883 @cindex call-used register
1884 @cindex call-clobbered register
1885 @cindex call-saved register
1886 A C expression that is nonzero if it is not permissible to store a
1887 value of mode @var{mode} in hard register number @var{regno} across a
1888 call without some part of it being clobbered. For most machines this
1889 macro need not be defined. It is only required for machines that do not
1890 preserve the entire contents of a register across a call.
1894 @findex call_used_regs
1897 @findex reg_class_contents
1898 @hook TARGET_CONDITIONAL_REGISTER_USAGE
1899 This hook may conditionally modify five variables
1900 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1901 @code{reg_names}, and @code{reg_class_contents}, to take into account
1902 any dependence of these register sets on target flags. The first three
1903 of these are of type @code{char []} (interpreted as Boolean vectors).
1904 @code{global_regs} is a @code{const char *[]}, and
1905 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1906 called, @code{fixed_regs}, @code{call_used_regs},
1907 @code{reg_class_contents}, and @code{reg_names} have been initialized
1908 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1909 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1910 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1911 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1912 command options have been applied.
1914 @cindex disabling certain registers
1915 @cindex controlling register usage
1916 If the usage of an entire class of registers depends on the target
1917 flags, you may indicate this to GCC by using this macro to modify
1918 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1919 registers in the classes which should not be used by GCC@. Also define
1920 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1921 to return @code{NO_REGS} if it
1922 is called with a letter for a class that shouldn't be used.
1924 (However, if this class is not included in @code{GENERAL_REGS} and all
1925 of the insn patterns whose constraints permit this class are
1926 controlled by target switches, then GCC will automatically avoid using
1927 these registers when the target switches are opposed to them.)
1930 @defmac INCOMING_REGNO (@var{out})
1931 Define this macro if the target machine has register windows. This C
1932 expression returns the register number as seen by the called function
1933 corresponding to the register number @var{out} as seen by the calling
1934 function. Return @var{out} if register number @var{out} is not an
1938 @defmac OUTGOING_REGNO (@var{in})
1939 Define this macro if the target machine has register windows. This C
1940 expression returns the register number as seen by the calling function
1941 corresponding to the register number @var{in} as seen by the called
1942 function. Return @var{in} if register number @var{in} is not an inbound
1946 @defmac LOCAL_REGNO (@var{regno})
1947 Define this macro if the target machine has register windows. This C
1948 expression returns true if the register is call-saved but is in the
1949 register window. Unlike most call-saved registers, such registers
1950 need not be explicitly restored on function exit or during non-local
1955 If the program counter has a register number, define this as that
1956 register number. Otherwise, do not define it.
1959 @node Allocation Order
1960 @subsection Order of Allocation of Registers
1961 @cindex order of register allocation
1962 @cindex register allocation order
1964 @c prevent bad page break with this line
1965 Registers are allocated in order.
1967 @defmac REG_ALLOC_ORDER
1968 If defined, an initializer for a vector of integers, containing the
1969 numbers of hard registers in the order in which GCC should prefer
1970 to use them (from most preferred to least).
1972 If this macro is not defined, registers are used lowest numbered first
1973 (all else being equal).
1975 One use of this macro is on machines where the highest numbered
1976 registers must always be saved and the save-multiple-registers
1977 instruction supports only sequences of consecutive registers. On such
1978 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1979 the highest numbered allocable register first.
1982 @defmac ADJUST_REG_ALLOC_ORDER
1983 A C statement (sans semicolon) to choose the order in which to allocate
1984 hard registers for pseudo-registers local to a basic block.
1986 Store the desired register order in the array @code{reg_alloc_order}.
1987 Element 0 should be the register to allocate first; element 1, the next
1988 register; and so on.
1990 The macro body should not assume anything about the contents of
1991 @code{reg_alloc_order} before execution of the macro.
1993 On most machines, it is not necessary to define this macro.
1996 @defmac HONOR_REG_ALLOC_ORDER
1997 Normally, IRA tries to estimate the costs for saving a register in the
1998 prologue and restoring it in the epilogue. This discourages it from
1999 using call-saved registers. If a machine wants to ensure that IRA
2000 allocates registers in the order given by REG_ALLOC_ORDER even if some
2001 call-saved registers appear earlier than call-used ones, this macro
2005 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2006 In some case register allocation order is not enough for the
2007 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2008 If this macro is defined, it should return a floating point value
2009 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2010 be increased by approximately the pseudo's usage frequency times the
2011 value returned by this macro. Not defining this macro is equivalent
2012 to having it always return @code{0.0}.
2014 On most machines, it is not necessary to define this macro.
2017 @node Values in Registers
2018 @subsection How Values Fit in Registers
2020 This section discusses the macros that describe which kinds of values
2021 (specifically, which machine modes) each register can hold, and how many
2022 consecutive registers are needed for a given mode.
2024 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2025 A C expression for the number of consecutive hard registers, starting
2026 at register number @var{regno}, required to hold a value of mode
2027 @var{mode}. This macro must never return zero, even if a register
2028 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2029 and/or CANNOT_CHANGE_MODE_CLASS instead.
2031 On a machine where all registers are exactly one word, a suitable
2032 definition of this macro is
2035 #define HARD_REGNO_NREGS(REGNO, MODE) \
2036 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2041 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2042 A C expression that is nonzero if a value of mode @var{mode}, stored
2043 in memory, ends with padding that causes it to take up more space than
2044 in registers starting at register number @var{regno} (as determined by
2045 multiplying GCC's notion of the size of the register when containing
2046 this mode by the number of registers returned by
2047 @code{HARD_REGNO_NREGS}). By default this is zero.
2049 For example, if a floating-point value is stored in three 32-bit
2050 registers but takes up 128 bits in memory, then this would be
2053 This macros only needs to be defined if there are cases where
2054 @code{subreg_get_info}
2055 would otherwise wrongly determine that a @code{subreg} can be
2056 represented by an offset to the register number, when in fact such a
2057 @code{subreg} would contain some of the padding not stored in
2058 registers and so not be representable.
2061 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2062 For values of @var{regno} and @var{mode} for which
2063 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2064 returning the greater number of registers required to hold the value
2065 including any padding. In the example above, the value would be four.
2068 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2069 Define this macro if the natural size of registers that hold values
2070 of mode @var{mode} is not the word size. It is a C expression that
2071 should give the natural size in bytes for the specified mode. It is
2072 used by the register allocator to try to optimize its results. This
2073 happens for example on SPARC 64-bit where the natural size of
2074 floating-point registers is still 32-bit.
2077 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2078 A C expression that is nonzero if it is permissible to store a value
2079 of mode @var{mode} in hard register number @var{regno} (or in several
2080 registers starting with that one). For a machine where all registers
2081 are equivalent, a suitable definition is
2084 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2087 You need not include code to check for the numbers of fixed registers,
2088 because the allocation mechanism considers them to be always occupied.
2090 @cindex register pairs
2091 On some machines, double-precision values must be kept in even/odd
2092 register pairs. You can implement that by defining this macro to reject
2093 odd register numbers for such modes.
2095 The minimum requirement for a mode to be OK in a register is that the
2096 @samp{mov@var{mode}} instruction pattern support moves between the
2097 register and other hard register in the same class and that moving a
2098 value into the register and back out not alter it.
2100 Since the same instruction used to move @code{word_mode} will work for
2101 all narrower integer modes, it is not necessary on any machine for
2102 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2103 you define patterns @samp{movhi}, etc., to take advantage of this. This
2104 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2105 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2108 Many machines have special registers for floating point arithmetic.
2109 Often people assume that floating point machine modes are allowed only
2110 in floating point registers. This is not true. Any registers that
2111 can hold integers can safely @emph{hold} a floating point machine
2112 mode, whether or not floating arithmetic can be done on it in those
2113 registers. Integer move instructions can be used to move the values.
2115 On some machines, though, the converse is true: fixed-point machine
2116 modes may not go in floating registers. This is true if the floating
2117 registers normalize any value stored in them, because storing a
2118 non-floating value there would garble it. In this case,
2119 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2120 floating registers. But if the floating registers do not automatically
2121 normalize, if you can store any bit pattern in one and retrieve it
2122 unchanged without a trap, then any machine mode may go in a floating
2123 register, so you can define this macro to say so.
2125 The primary significance of special floating registers is rather that
2126 they are the registers acceptable in floating point arithmetic
2127 instructions. However, this is of no concern to
2128 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2129 constraints for those instructions.
2131 On some machines, the floating registers are especially slow to access,
2132 so that it is better to store a value in a stack frame than in such a
2133 register if floating point arithmetic is not being done. As long as the
2134 floating registers are not in class @code{GENERAL_REGS}, they will not
2135 be used unless some pattern's constraint asks for one.
2138 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2139 A C expression that is nonzero if it is OK to rename a hard register
2140 @var{from} to another hard register @var{to}.
2142 One common use of this macro is to prevent renaming of a register to
2143 another register that is not saved by a prologue in an interrupt
2146 The default is always nonzero.
2149 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2150 A C expression that is nonzero if a value of mode
2151 @var{mode1} is accessible in mode @var{mode2} without copying.
2153 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2154 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2155 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2156 should be nonzero. If they differ for any @var{r}, you should define
2157 this macro to return zero unless some other mechanism ensures the
2158 accessibility of the value in a narrower mode.
2160 You should define this macro to return nonzero in as many cases as
2161 possible since doing so will allow GCC to perform better register
2165 @hook TARGET_HARD_REGNO_SCRATCH_OK
2166 This target hook should return @code{true} if it is OK to use a hard register
2167 @var{regno} as scratch reg in peephole2.
2169 One common use of this macro is to prevent using of a register that
2170 is not saved by a prologue in an interrupt handler.
2172 The default version of this hook always returns @code{true}.
2175 @defmac AVOID_CCMODE_COPIES
2176 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2177 registers. You should only define this macro if support for copying to/from
2178 @code{CCmode} is incomplete.
2181 @node Leaf Functions
2182 @subsection Handling Leaf Functions
2184 @cindex leaf functions
2185 @cindex functions, leaf
2186 On some machines, a leaf function (i.e., one which makes no calls) can run
2187 more efficiently if it does not make its own register window. Often this
2188 means it is required to receive its arguments in the registers where they
2189 are passed by the caller, instead of the registers where they would
2192 The special treatment for leaf functions generally applies only when
2193 other conditions are met; for example, often they may use only those
2194 registers for its own variables and temporaries. We use the term ``leaf
2195 function'' to mean a function that is suitable for this special
2196 handling, so that functions with no calls are not necessarily ``leaf
2199 GCC assigns register numbers before it knows whether the function is
2200 suitable for leaf function treatment. So it needs to renumber the
2201 registers in order to output a leaf function. The following macros
2204 @defmac LEAF_REGISTERS
2205 Name of a char vector, indexed by hard register number, which
2206 contains 1 for a register that is allowable in a candidate for leaf
2209 If leaf function treatment involves renumbering the registers, then the
2210 registers marked here should be the ones before renumbering---those that
2211 GCC would ordinarily allocate. The registers which will actually be
2212 used in the assembler code, after renumbering, should not be marked with 1
2215 Define this macro only if the target machine offers a way to optimize
2216 the treatment of leaf functions.
2219 @defmac LEAF_REG_REMAP (@var{regno})
2220 A C expression whose value is the register number to which @var{regno}
2221 should be renumbered, when a function is treated as a leaf function.
2223 If @var{regno} is a register number which should not appear in a leaf
2224 function before renumbering, then the expression should yield @minus{}1, which
2225 will cause the compiler to abort.
2227 Define this macro only if the target machine offers a way to optimize the
2228 treatment of leaf functions, and registers need to be renumbered to do
2232 @findex current_function_is_leaf
2233 @findex current_function_uses_only_leaf_regs
2234 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2235 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2236 specially. They can test the C variable @code{current_function_is_leaf}
2237 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2238 set prior to local register allocation and is valid for the remaining
2239 compiler passes. They can also test the C variable
2240 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2241 functions which only use leaf registers.
2242 @code{current_function_uses_only_leaf_regs} is valid after all passes
2243 that modify the instructions have been run and is only useful if
2244 @code{LEAF_REGISTERS} is defined.
2245 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2246 @c of the next paragraph?! --mew 2feb93
2248 @node Stack Registers
2249 @subsection Registers That Form a Stack
2251 There are special features to handle computers where some of the
2252 ``registers'' form a stack. Stack registers are normally written by
2253 pushing onto the stack, and are numbered relative to the top of the
2256 Currently, GCC can only handle one group of stack-like registers, and
2257 they must be consecutively numbered. Furthermore, the existing
2258 support for stack-like registers is specific to the 80387 floating
2259 point coprocessor. If you have a new architecture that uses
2260 stack-like registers, you will need to do substantial work on
2261 @file{reg-stack.c} and write your machine description to cooperate
2262 with it, as well as defining these macros.
2265 Define this if the machine has any stack-like registers.
2268 @defmac STACK_REG_COVER_CLASS
2269 This is a cover class containing the stack registers. Define this if
2270 the machine has any stack-like registers.
2273 @defmac FIRST_STACK_REG
2274 The number of the first stack-like register. This one is the top
2278 @defmac LAST_STACK_REG
2279 The number of the last stack-like register. This one is the bottom of
2283 @node Register Classes
2284 @section Register Classes
2285 @cindex register class definitions
2286 @cindex class definitions, register
2288 On many machines, the numbered registers are not all equivalent.
2289 For example, certain registers may not be allowed for indexed addressing;
2290 certain registers may not be allowed in some instructions. These machine
2291 restrictions are described to the compiler using @dfn{register classes}.
2293 You define a number of register classes, giving each one a name and saying
2294 which of the registers belong to it. Then you can specify register classes
2295 that are allowed as operands to particular instruction patterns.
2299 In general, each register will belong to several classes. In fact, one
2300 class must be named @code{ALL_REGS} and contain all the registers. Another
2301 class must be named @code{NO_REGS} and contain no registers. Often the
2302 union of two classes will be another class; however, this is not required.
2304 @findex GENERAL_REGS
2305 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2306 terribly special about the name, but the operand constraint letters
2307 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2308 the same as @code{ALL_REGS}, just define it as a macro which expands
2311 Order the classes so that if class @var{x} is contained in class @var{y}
2312 then @var{x} has a lower class number than @var{y}.
2314 The way classes other than @code{GENERAL_REGS} are specified in operand
2315 constraints is through machine-dependent operand constraint letters.
2316 You can define such letters to correspond to various classes, then use
2317 them in operand constraints.
2319 You should define a class for the union of two classes whenever some
2320 instruction allows both classes. For example, if an instruction allows
2321 either a floating point (coprocessor) register or a general register for a
2322 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2323 which includes both of them. Otherwise you will get suboptimal code,
2324 or even internal compiler errors when reload cannot find a register in the
2325 the class computed via @code{reg_class_subunion}.
2327 You must also specify certain redundant information about the register
2328 classes: for each class, which classes contain it and which ones are
2329 contained in it; for each pair of classes, the largest class contained
2332 When a value occupying several consecutive registers is expected in a
2333 certain class, all the registers used must belong to that class.
2334 Therefore, register classes cannot be used to enforce a requirement for
2335 a register pair to start with an even-numbered register. The way to
2336 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2338 Register classes used for input-operands of bitwise-and or shift
2339 instructions have a special requirement: each such class must have, for
2340 each fixed-point machine mode, a subclass whose registers can transfer that
2341 mode to or from memory. For example, on some machines, the operations for
2342 single-byte values (@code{QImode}) are limited to certain registers. When
2343 this is so, each register class that is used in a bitwise-and or shift
2344 instruction must have a subclass consisting of registers from which
2345 single-byte values can be loaded or stored. This is so that
2346 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2348 @deftp {Data type} {enum reg_class}
2349 An enumerated type that must be defined with all the register class names
2350 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2351 must be the last register class, followed by one more enumerated value,
2352 @code{LIM_REG_CLASSES}, which is not a register class but rather
2353 tells how many classes there are.
2355 Each register class has a number, which is the value of casting
2356 the class name to type @code{int}. The number serves as an index
2357 in many of the tables described below.
2360 @defmac N_REG_CLASSES
2361 The number of distinct register classes, defined as follows:
2364 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2368 @defmac REG_CLASS_NAMES
2369 An initializer containing the names of the register classes as C string
2370 constants. These names are used in writing some of the debugging dumps.
2373 @defmac REG_CLASS_CONTENTS
2374 An initializer containing the contents of the register classes, as integers
2375 which are bit masks. The @var{n}th integer specifies the contents of class
2376 @var{n}. The way the integer @var{mask} is interpreted is that
2377 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2379 When the machine has more than 32 registers, an integer does not suffice.
2380 Then the integers are replaced by sub-initializers, braced groupings containing
2381 several integers. Each sub-initializer must be suitable as an initializer
2382 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2383 In this situation, the first integer in each sub-initializer corresponds to
2384 registers 0 through 31, the second integer to registers 32 through 63, and
2388 @defmac REGNO_REG_CLASS (@var{regno})
2389 A C expression whose value is a register class containing hard register
2390 @var{regno}. In general there is more than one such class; choose a class
2391 which is @dfn{minimal}, meaning that no smaller class also contains the
2395 @defmac BASE_REG_CLASS
2396 A macro whose definition is the name of the class to which a valid
2397 base register must belong. A base register is one used in an address
2398 which is the register value plus a displacement.
2401 @defmac MODE_BASE_REG_CLASS (@var{mode})
2402 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2403 the selection of a base register in a mode dependent manner. If
2404 @var{mode} is VOIDmode then it should return the same value as
2405 @code{BASE_REG_CLASS}.
2408 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2409 A C expression whose value is the register class to which a valid
2410 base register must belong in order to be used in a base plus index
2411 register address. You should define this macro if base plus index
2412 addresses have different requirements than other base register uses.
2415 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2416 A C expression whose value is the register class to which a valid
2417 base register must belong. @var{outer_code} and @var{index_code} define the
2418 context in which the base register occurs. @var{outer_code} is the code of
2419 the immediately enclosing expression (@code{MEM} for the top level of an
2420 address, @code{ADDRESS} for something that occurs in an
2421 @code{address_operand}). @var{index_code} is the code of the corresponding
2422 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2425 @defmac INDEX_REG_CLASS
2426 A macro whose definition is the name of the class to which a valid
2427 index register must belong. An index register is one used in an
2428 address where its value is either multiplied by a scale factor or
2429 added to another register (as well as added to a displacement).
2432 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2433 A C expression which is nonzero if register number @var{num} is
2434 suitable for use as a base register in operand addresses.
2437 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2438 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2439 that expression may examine the mode of the memory reference in
2440 @var{mode}. You should define this macro if the mode of the memory
2441 reference affects whether a register may be used as a base register. If
2442 you define this macro, the compiler will use it instead of
2443 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2444 addresses that appear outside a @code{MEM}, i.e., as an
2445 @code{address_operand}.
2448 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2449 A C expression which is nonzero if register number @var{num} is suitable for
2450 use as a base register in base plus index operand addresses, accessing
2451 memory in mode @var{mode}. It may be either a suitable hard register or a
2452 pseudo register that has been allocated such a hard register. You should
2453 define this macro if base plus index addresses have different requirements
2454 than other base register uses.
2456 Use of this macro is deprecated; please use the more general
2457 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2460 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2461 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2462 that that expression may examine the context in which the register
2463 appears in the memory reference. @var{outer_code} is the code of the
2464 immediately enclosing expression (@code{MEM} if at the top level of the
2465 address, @code{ADDRESS} for something that occurs in an
2466 @code{address_operand}). @var{index_code} is the code of the
2467 corresponding index expression if @var{outer_code} is @code{PLUS};
2468 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2469 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2472 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2473 A C expression which is nonzero if register number @var{num} is
2474 suitable for use as an index register in operand addresses. It may be
2475 either a suitable hard register or a pseudo register that has been
2476 allocated such a hard register.
2478 The difference between an index register and a base register is that
2479 the index register may be scaled. If an address involves the sum of
2480 two registers, neither one of them scaled, then either one may be
2481 labeled the ``base'' and the other the ``index''; but whichever
2482 labeling is used must fit the machine's constraints of which registers
2483 may serve in each capacity. The compiler will try both labelings,
2484 looking for one that is valid, and will reload one or both registers
2485 only if neither labeling works.
2488 @hook TARGET_PREFERRED_RENAME_CLASS
2490 @hook TARGET_PREFERRED_RELOAD_CLASS
2491 A target hook that places additional restrictions on the register class
2492 to use when it is necessary to copy value @var{x} into a register in class
2493 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2494 another, smaller class.
2496 The default version of this hook always returns value of @code{rclass} argument.
2498 Sometimes returning a more restrictive class makes better code. For
2499 example, on the 68000, when @var{x} is an integer constant that is in range
2500 for a @samp{moveq} instruction, the value of this macro is always
2501 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2502 Requiring a data register guarantees that a @samp{moveq} will be used.
2504 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2505 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2506 loaded into some register class. By returning @code{NO_REGS} you can
2507 force @var{x} into a memory location. For example, rs6000 can load
2508 immediate values into general-purpose registers, but does not have an
2509 instruction for loading an immediate value into a floating-point
2510 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2511 @var{x} is a floating-point constant. If the constant can't be loaded
2512 into any kind of register, code generation will be better if
2513 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2514 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2516 If an insn has pseudos in it after register allocation, reload will go
2517 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2518 to find the best one. Returning @code{NO_REGS}, in this case, makes
2519 reload add a @code{!} in front of the constraint: the x86 back-end uses
2520 this feature to discourage usage of 387 registers when math is done in
2521 the SSE registers (and vice versa).
2524 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2525 A C expression that places additional restrictions on the register class
2526 to use when it is necessary to copy value @var{x} into a register in class
2527 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2528 another, smaller class. On many machines, the following definition is
2532 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2535 Sometimes returning a more restrictive class makes better code. For
2536 example, on the 68000, when @var{x} is an integer constant that is in range
2537 for a @samp{moveq} instruction, the value of this macro is always
2538 @code{DATA_REGS} as long as @var{class} includes the data registers.
2539 Requiring a data register guarantees that a @samp{moveq} will be used.
2541 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2542 @var{class} is if @var{x} is a legitimate constant which cannot be
2543 loaded into some register class. By returning @code{NO_REGS} you can
2544 force @var{x} into a memory location. For example, rs6000 can load
2545 immediate values into general-purpose registers, but does not have an
2546 instruction for loading an immediate value into a floating-point
2547 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2548 @var{x} is a floating-point constant. If the constant can't be loaded
2549 into any kind of register, code generation will be better if
2550 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2551 of using @code{PREFERRED_RELOAD_CLASS}.
2553 If an insn has pseudos in it after register allocation, reload will go
2554 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2555 to find the best one. Returning @code{NO_REGS}, in this case, makes
2556 reload add a @code{!} in front of the constraint: the x86 back-end uses
2557 this feature to discourage usage of 387 registers when math is done in
2558 the SSE registers (and vice versa).
2561 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2562 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2563 input reloads. If you don't define this macro, the default is to use
2564 @var{class}, unchanged.
2566 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2567 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2570 @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2571 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2574 The default version of this hook always returns value of @code{rclass}
2577 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2578 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2581 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2582 A C expression that places additional restrictions on the register class
2583 to use when it is necessary to be able to hold a value of mode
2584 @var{mode} in a reload register for which class @var{class} would
2587 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2588 there are certain modes that simply can't go in certain reload classes.
2590 The value is a register class; perhaps @var{class}, or perhaps another,
2593 Don't define this macro unless the target machine has limitations which
2594 require the macro to do something nontrivial.
2597 @hook TARGET_SECONDARY_RELOAD
2598 Many machines have some registers that cannot be copied directly to or
2599 from memory or even from other types of registers. An example is the
2600 @samp{MQ} register, which on most machines, can only be copied to or
2601 from general registers, but not memory. Below, we shall be using the
2602 term 'intermediate register' when a move operation cannot be performed
2603 directly, but has to be done by copying the source into the intermediate
2604 register first, and then copying the intermediate register to the
2605 destination. An intermediate register always has the same mode as
2606 source and destination. Since it holds the actual value being copied,
2607 reload might apply optimizations to re-use an intermediate register
2608 and eliding the copy from the source when it can determine that the
2609 intermediate register still holds the required value.
2611 Another kind of secondary reload is required on some machines which
2612 allow copying all registers to and from memory, but require a scratch
2613 register for stores to some memory locations (e.g., those with symbolic
2614 address on the RT, and those with certain symbolic address on the SPARC
2615 when compiling PIC)@. Scratch registers need not have the same mode
2616 as the value being copied, and usually hold a different value than
2617 that being copied. Special patterns in the md file are needed to
2618 describe how the copy is performed with the help of the scratch register;
2619 these patterns also describe the number, register class(es) and mode(s)
2620 of the scratch register(s).
2622 In some cases, both an intermediate and a scratch register are required.
2624 For input reloads, this target hook is called with nonzero @var{in_p},
2625 and @var{x} is an rtx that needs to be copied to a register of class
2626 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2627 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2628 needs to be copied to rtx @var{x} in @var{reload_mode}.
2630 If copying a register of @var{reload_class} from/to @var{x} requires
2631 an intermediate register, the hook @code{secondary_reload} should
2632 return the register class required for this intermediate register.
2633 If no intermediate register is required, it should return NO_REGS.
2634 If more than one intermediate register is required, describe the one
2635 that is closest in the copy chain to the reload register.
2637 If scratch registers are needed, you also have to describe how to
2638 perform the copy from/to the reload register to/from this
2639 closest intermediate register. Or if no intermediate register is
2640 required, but still a scratch register is needed, describe the
2641 copy from/to the reload register to/from the reload operand @var{x}.
2643 You do this by setting @code{sri->icode} to the instruction code of a pattern
2644 in the md file which performs the move. Operands 0 and 1 are the output
2645 and input of this copy, respectively. Operands from operand 2 onward are
2646 for scratch operands. These scratch operands must have a mode, and a
2647 single-register-class
2648 @c [later: or memory]
2651 When an intermediate register is used, the @code{secondary_reload}
2652 hook will be called again to determine how to copy the intermediate
2653 register to/from the reload operand @var{x}, so your hook must also
2654 have code to handle the register class of the intermediate operand.
2656 @c [For later: maybe we'll allow multi-alternative reload patterns -
2657 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2658 @c and match the constraints of input and output to determine the required
2659 @c alternative. A restriction would be that constraints used to match
2660 @c against reloads registers would have to be written as register class
2661 @c constraints, or we need a new target macro / hook that tells us if an
2662 @c arbitrary constraint can match an unknown register of a given class.
2663 @c Such a macro / hook would also be useful in other places.]
2666 @var{x} might be a pseudo-register or a @code{subreg} of a
2667 pseudo-register, which could either be in a hard register or in memory.
2668 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2669 in memory and the hard register number if it is in a register.
2671 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2672 currently not supported. For the time being, you will have to continue
2673 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2675 @code{copy_cost} also uses this target hook to find out how values are
2676 copied. If you want it to include some extra cost for the need to allocate
2677 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2678 Or if two dependent moves are supposed to have a lower cost than the sum
2679 of the individual moves due to expected fortuitous scheduling and/or special
2680 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2683 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2684 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2685 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2686 These macros are obsolete, new ports should use the target hook
2687 @code{TARGET_SECONDARY_RELOAD} instead.
2689 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2690 target hook. Older ports still define these macros to indicate to the
2691 reload phase that it may
2692 need to allocate at least one register for a reload in addition to the
2693 register to contain the data. Specifically, if copying @var{x} to a
2694 register @var{class} in @var{mode} requires an intermediate register,
2695 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2696 largest register class all of whose registers can be used as
2697 intermediate registers or scratch registers.
2699 If copying a register @var{class} in @var{mode} to @var{x} requires an
2700 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2701 was supposed to be defined be defined to return the largest register
2702 class required. If the
2703 requirements for input and output reloads were the same, the macro
2704 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2707 The values returned by these macros are often @code{GENERAL_REGS}.
2708 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2709 can be directly copied to or from a register of @var{class} in
2710 @var{mode} without requiring a scratch register. Do not define this
2711 macro if it would always return @code{NO_REGS}.
2713 If a scratch register is required (either with or without an
2714 intermediate register), you were supposed to define patterns for
2715 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2716 (@pxref{Standard Names}. These patterns, which were normally
2717 implemented with a @code{define_expand}, should be similar to the
2718 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2721 These patterns need constraints for the reload register and scratch
2723 contain a single register class. If the original reload register (whose
2724 class is @var{class}) can meet the constraint given in the pattern, the
2725 value returned by these macros is used for the class of the scratch
2726 register. Otherwise, two additional reload registers are required.
2727 Their classes are obtained from the constraints in the insn pattern.
2729 @var{x} might be a pseudo-register or a @code{subreg} of a
2730 pseudo-register, which could either be in a hard register or in memory.
2731 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2732 in memory and the hard register number if it is in a register.
2734 These macros should not be used in the case where a particular class of
2735 registers can only be copied to memory and not to another class of
2736 registers. In that case, secondary reload registers are not needed and
2737 would not be helpful. Instead, a stack location must be used to perform
2738 the copy and the @code{mov@var{m}} pattern should use memory as an
2739 intermediate storage. This case often occurs between floating-point and
2743 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2744 Certain machines have the property that some registers cannot be copied
2745 to some other registers without using memory. Define this macro on
2746 those machines to be a C expression that is nonzero if objects of mode
2747 @var{m} in registers of @var{class1} can only be copied to registers of
2748 class @var{class2} by storing a register of @var{class1} into memory
2749 and loading that memory location into a register of @var{class2}.
2751 Do not define this macro if its value would always be zero.
2754 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2755 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2756 allocates a stack slot for a memory location needed for register copies.
2757 If this macro is defined, the compiler instead uses the memory location
2758 defined by this macro.
2760 Do not define this macro if you do not define
2761 @code{SECONDARY_MEMORY_NEEDED}.
2764 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2765 When the compiler needs a secondary memory location to copy between two
2766 registers of mode @var{mode}, it normally allocates sufficient memory to
2767 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2768 load operations in a mode that many bits wide and whose class is the
2769 same as that of @var{mode}.
2771 This is right thing to do on most machines because it ensures that all
2772 bits of the register are copied and prevents accesses to the registers
2773 in a narrower mode, which some machines prohibit for floating-point
2776 However, this default behavior is not correct on some machines, such as
2777 the DEC Alpha, that store short integers in floating-point registers
2778 differently than in integer registers. On those machines, the default
2779 widening will not work correctly and you must define this macro to
2780 suppress that widening in some cases. See the file @file{alpha.h} for
2783 Do not define this macro if you do not define
2784 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2785 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2788 @hook TARGET_CLASS_LIKELY_SPILLED_P
2789 A target hook which returns @code{true} if pseudos that have been assigned
2790 to registers of class @var{rclass} would likely be spilled because
2791 registers of @var{rclass} are needed for spill registers.
2793 The default version of this target hook returns @code{true} if @var{rclass}
2794 has exactly one register and @code{false} otherwise. On most machines, this
2795 default should be used. Only use this target hook to some other expression
2796 if pseudos allocated by @file{local-alloc.c} end up in memory because their
2797 hard registers were needed for spill registers. If this target hook returns
2798 @code{false} for those classes, those pseudos will only be allocated by
2799 @file{global.c}, which knows how to reallocate the pseudo to another
2800 register. If there would not be another register available for reallocation,
2801 you should not change the implementation of this target hook since
2802 the only effect of such implementation would be to slow down register
2806 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2807 A C expression for the maximum number of consecutive registers
2808 of class @var{class} needed to hold a value of mode @var{mode}.
2810 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2811 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2812 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2813 @var{mode})} for all @var{regno} values in the class @var{class}.
2815 This macro helps control the handling of multiple-word values
2819 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2820 If defined, a C expression that returns nonzero for a @var{class} for which
2821 a change from mode @var{from} to mode @var{to} is invalid.
2823 For the example, loading 32-bit integer or floating-point objects into
2824 floating-point registers on the Alpha extends them to 64 bits.
2825 Therefore loading a 64-bit object and then storing it as a 32-bit object
2826 does not store the low-order 32 bits, as would be the case for a normal
2827 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2831 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2832 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2833 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2837 @node Old Constraints
2838 @section Obsolete Macros for Defining Constraints
2839 @cindex defining constraints, obsolete method
2840 @cindex constraints, defining, obsolete method
2842 Machine-specific constraints can be defined with these macros instead
2843 of the machine description constructs described in @ref{Define
2844 Constraints}. This mechanism is obsolete. New ports should not use
2845 it; old ports should convert to the new mechanism.
2847 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2848 For the constraint at the start of @var{str}, which starts with the letter
2849 @var{c}, return the length. This allows you to have register class /
2850 constant / extra constraints that are longer than a single letter;
2851 you don't need to define this macro if you can do with single-letter
2852 constraints only. The definition of this macro should use
2853 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2854 to handle specially.
2855 There are some sanity checks in genoutput.c that check the constraint lengths
2856 for the md file, so you can also use this macro to help you while you are
2857 transitioning from a byzantine single-letter-constraint scheme: when you
2858 return a negative length for a constraint you want to re-use, genoutput
2859 will complain about every instance where it is used in the md file.
2862 @defmac REG_CLASS_FROM_LETTER (@var{char})
2863 A C expression which defines the machine-dependent operand constraint
2864 letters for register classes. If @var{char} is such a letter, the
2865 value should be the register class corresponding to it. Otherwise,
2866 the value should be @code{NO_REGS}. The register letter @samp{r},
2867 corresponding to class @code{GENERAL_REGS}, will not be passed
2868 to this macro; you do not need to handle it.
2871 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2872 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2873 passed in @var{str}, so that you can use suffixes to distinguish between
2877 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2878 A C expression that defines the machine-dependent operand constraint
2879 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2880 particular ranges of integer values. If @var{c} is one of those
2881 letters, the expression should check that @var{value}, an integer, is in
2882 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2883 not one of those letters, the value should be 0 regardless of
2887 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2888 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2889 string passed in @var{str}, so that you can use suffixes to distinguish
2890 between different variants.
2893 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2894 A C expression that defines the machine-dependent operand constraint
2895 letters that specify particular ranges of @code{const_double} values
2896 (@samp{G} or @samp{H}).
2898 If @var{c} is one of those letters, the expression should check that
2899 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2900 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2901 letters, the value should be 0 regardless of @var{value}.
2903 @code{const_double} is used for all floating-point constants and for
2904 @code{DImode} fixed-point constants. A given letter can accept either
2905 or both kinds of values. It can use @code{GET_MODE} to distinguish
2906 between these kinds.
2909 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2910 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2911 string passed in @var{str}, so that you can use suffixes to distinguish
2912 between different variants.
2915 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2916 A C expression that defines the optional machine-dependent constraint
2917 letters that can be used to segregate specific types of operands, usually
2918 memory references, for the target machine. Any letter that is not
2919 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2920 @code{REG_CLASS_FROM_CONSTRAINT}
2921 may be used. Normally this macro will not be defined.
2923 If it is required for a particular target machine, it should return 1
2924 if @var{value} corresponds to the operand type represented by the
2925 constraint letter @var{c}. If @var{c} is not defined as an extra
2926 constraint, the value returned should be 0 regardless of @var{value}.
2928 For example, on the ROMP, load instructions cannot have their output
2929 in r0 if the memory reference contains a symbolic address. Constraint
2930 letter @samp{Q} is defined as representing a memory address that does
2931 @emph{not} contain a symbolic address. An alternative is specified with
2932 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2933 alternative specifies @samp{m} on the input and a register class that
2934 does not include r0 on the output.
2937 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2938 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2939 in @var{str}, so that you can use suffixes to distinguish between different
2943 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2944 A C expression that defines the optional machine-dependent constraint
2945 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2946 be treated like memory constraints by the reload pass.
2948 It should return 1 if the operand type represented by the constraint
2949 at the start of @var{str}, the first letter of which is the letter @var{c},
2950 comprises a subset of all memory references including
2951 all those whose address is simply a base register. This allows the reload
2952 pass to reload an operand, if it does not directly correspond to the operand
2953 type of @var{c}, by copying its address into a base register.
2955 For example, on the S/390, some instructions do not accept arbitrary
2956 memory references, but only those that do not make use of an index
2957 register. The constraint letter @samp{Q} is defined via
2958 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2959 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2960 a @samp{Q} constraint can handle any memory operand, because the
2961 reload pass knows it can be reloaded by copying the memory address
2962 into a base register if required. This is analogous to the way
2963 an @samp{o} constraint can handle any memory operand.
2966 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2967 A C expression that defines the optional machine-dependent constraint
2968 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2969 @code{EXTRA_CONSTRAINT_STR}, that should
2970 be treated like address constraints by the reload pass.
2972 It should return 1 if the operand type represented by the constraint
2973 at the start of @var{str}, which starts with the letter @var{c}, comprises
2974 a subset of all memory addresses including
2975 all those that consist of just a base register. This allows the reload
2976 pass to reload an operand, if it does not directly correspond to the operand
2977 type of @var{str}, by copying it into a base register.
2979 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2980 be used with the @code{address_operand} predicate. It is treated
2981 analogously to the @samp{p} constraint.
2984 @node Stack and Calling
2985 @section Stack Layout and Calling Conventions
2986 @cindex calling conventions
2988 @c prevent bad page break with this line
2989 This describes the stack layout and calling conventions.
2993 * Exception Handling::
2998 * Register Arguments::
3000 * Aggregate Return::
3005 * Stack Smashing Protection::
3009 @subsection Basic Stack Layout
3010 @cindex stack frame layout
3011 @cindex frame layout
3013 @c prevent bad page break with this line
3014 Here is the basic stack layout.
3016 @defmac STACK_GROWS_DOWNWARD
3017 Define this macro if pushing a word onto the stack moves the stack
3018 pointer to a smaller address.
3020 When we say, ``define this macro if @dots{}'', it means that the
3021 compiler checks this macro only with @code{#ifdef} so the precise
3022 definition used does not matter.
3025 @defmac STACK_PUSH_CODE
3026 This macro defines the operation used when something is pushed
3027 on the stack. In RTL, a push operation will be
3028 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3030 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3031 and @code{POST_INC}. Which of these is correct depends on
3032 the stack direction and on whether the stack pointer points
3033 to the last item on the stack or whether it points to the
3034 space for the next item on the stack.
3036 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3037 defined, which is almost always right, and @code{PRE_INC} otherwise,
3038 which is often wrong.
3041 @defmac FRAME_GROWS_DOWNWARD
3042 Define this macro to nonzero value if the addresses of local variable slots
3043 are at negative offsets from the frame pointer.
3046 @defmac ARGS_GROW_DOWNWARD
3047 Define this macro if successive arguments to a function occupy decreasing
3048 addresses on the stack.
3051 @defmac STARTING_FRAME_OFFSET
3052 Offset from the frame pointer to the first local variable slot to be allocated.
3054 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3055 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3056 Otherwise, it is found by adding the length of the first slot to the
3057 value @code{STARTING_FRAME_OFFSET}.
3058 @c i'm not sure if the above is still correct.. had to change it to get
3059 @c rid of an overfull. --mew 2feb93
3062 @defmac STACK_ALIGNMENT_NEEDED
3063 Define to zero to disable final alignment of the stack during reload.
3064 The nonzero default for this macro is suitable for most ports.
3066 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3067 is a register save block following the local block that doesn't require
3068 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3069 stack alignment and do it in the backend.
3072 @defmac STACK_POINTER_OFFSET
3073 Offset from the stack pointer register to the first location at which
3074 outgoing arguments are placed. If not specified, the default value of
3075 zero is used. This is the proper value for most machines.
3077 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3078 the first location at which outgoing arguments are placed.
3081 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3082 Offset from the argument pointer register to the first argument's
3083 address. On some machines it may depend on the data type of the
3086 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3087 the first argument's address.
3090 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3091 Offset from the stack pointer register to an item dynamically allocated
3092 on the stack, e.g., by @code{alloca}.
3094 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3095 length of the outgoing arguments. The default is correct for most
3096 machines. See @file{function.c} for details.
3099 @defmac INITIAL_FRAME_ADDRESS_RTX
3100 A C expression whose value is RTL representing the address of the initial
3101 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3102 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3103 default value will be used. Define this macro in order to make frame pointer
3104 elimination work in the presence of @code{__builtin_frame_address (count)} and
3105 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3108 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3109 A C expression whose value is RTL representing the address in a stack
3110 frame where the pointer to the caller's frame is stored. Assume that
3111 @var{frameaddr} is an RTL expression for the address of the stack frame
3114 If you don't define this macro, the default is to return the value
3115 of @var{frameaddr}---that is, the stack frame address is also the
3116 address of the stack word that points to the previous frame.
3119 @defmac SETUP_FRAME_ADDRESSES
3120 If defined, a C expression that produces the machine-specific code to
3121 setup the stack so that arbitrary frames can be accessed. For example,
3122 on the SPARC, we must flush all of the register windows to the stack
3123 before we can access arbitrary stack frames. You will seldom need to
3127 @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
3128 This target hook should return an rtx that is used to store
3129 the address of the current frame into the built in @code{setjmp} buffer.
3130 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3131 machines. One reason you may need to define this target hook is if
3132 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3135 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3136 A C expression whose value is RTL representing the value of the frame
3137 address for the current frame. @var{frameaddr} is the frame pointer
3138 of the current frame. This is used for __builtin_frame_address.
3139 You need only define this macro if the frame address is not the same
3140 as the frame pointer. Most machines do not need to define it.
3143 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3144 A C expression whose value is RTL representing the value of the return
3145 address for the frame @var{count} steps up from the current frame, after
3146 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3147 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3148 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3150 The value of the expression must always be the correct address when
3151 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3152 determine the return address of other frames.
3155 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3156 Define this if the return address of a particular stack frame is accessed
3157 from the frame pointer of the previous stack frame.
3160 @defmac INCOMING_RETURN_ADDR_RTX
3161 A C expression whose value is RTL representing the location of the
3162 incoming return address at the beginning of any function, before the
3163 prologue. This RTL is either a @code{REG}, indicating that the return
3164 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3167 You only need to define this macro if you want to support call frame
3168 debugging information like that provided by DWARF 2.
3170 If this RTL is a @code{REG}, you should also define
3171 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3174 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3175 A C expression whose value is an integer giving a DWARF 2 column
3176 number that may be used as an alternative return column. The column
3177 must not correspond to any gcc hard register (that is, it must not
3178 be in the range of @code{DWARF_FRAME_REGNUM}).
3180 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3181 general register, but an alternative column needs to be used for signal
3182 frames. Some targets have also used different frame return columns
3186 @defmac DWARF_ZERO_REG
3187 A C expression whose value is an integer giving a DWARF 2 register
3188 number that is considered to always have the value zero. This should
3189 only be defined if the target has an architected zero register, and
3190 someone decided it was a good idea to use that register number to
3191 terminate the stack backtrace. New ports should avoid this.
3194 @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
3195 This target hook allows the backend to emit frame-related insns that
3196 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3197 info engine will invoke it on insns of the form
3199 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3203 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3205 to let the backend emit the call frame instructions. @var{label} is
3206 the CFI label attached to the insn, @var{pattern} is the pattern of
3207 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3210 @defmac INCOMING_FRAME_SP_OFFSET
3211 A C expression whose value is an integer giving the offset, in bytes,
3212 from the value of the stack pointer register to the top of the stack
3213 frame at the beginning of any function, before the prologue. The top of
3214 the frame is defined to be the value of the stack pointer in the
3215 previous frame, just before the call instruction.
3217 You only need to define this macro if you want to support call frame
3218 debugging information like that provided by DWARF 2.
3221 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3222 A C expression whose value is an integer giving the offset, in bytes,
3223 from the argument pointer to the canonical frame address (cfa). The
3224 final value should coincide with that calculated by
3225 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3226 during virtual register instantiation.
3228 The default value for this macro is
3229 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3230 which is correct for most machines; in general, the arguments are found
3231 immediately before the stack frame. Note that this is not the case on
3232 some targets that save registers into the caller's frame, such as SPARC
3233 and rs6000, and so such targets need to define this macro.
3235 You only need to define this macro if the default is incorrect, and you
3236 want to support call frame debugging information like that provided by
3240 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3241 If defined, a C expression whose value is an integer giving the offset
3242 in bytes from the frame pointer to the canonical frame address (cfa).
3243 The final value should coincide with that calculated by
3244 @code{INCOMING_FRAME_SP_OFFSET}.
3246 Normally the CFA is calculated as an offset from the argument pointer,
3247 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3248 variable due to the ABI, this may not be possible. If this macro is
3249 defined, it implies that the virtual register instantiation should be
3250 based on the frame pointer instead of the argument pointer. Only one
3251 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3255 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3256 If defined, a C expression whose value is an integer giving the offset
3257 in bytes from the canonical frame address (cfa) to the frame base used
3258 in DWARF 2 debug information. The default is zero. A different value
3259 may reduce the size of debug information on some ports.
3262 @node Exception Handling
3263 @subsection Exception Handling Support
3264 @cindex exception handling
3266 @defmac EH_RETURN_DATA_REGNO (@var{N})
3267 A C expression whose value is the @var{N}th register number used for
3268 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3269 @var{N} registers are usable.
3271 The exception handling library routines communicate with the exception
3272 handlers via a set of agreed upon registers. Ideally these registers
3273 should be call-clobbered; it is possible to use call-saved registers,
3274 but may negatively impact code size. The target must support at least
3275 2 data registers, but should define 4 if there are enough free registers.
3277 You must define this macro if you want to support call frame exception
3278 handling like that provided by DWARF 2.
3281 @defmac EH_RETURN_STACKADJ_RTX
3282 A C expression whose value is RTL representing a location in which
3283 to store a stack adjustment to be applied before function return.
3284 This is used to unwind the stack to an exception handler's call frame.
3285 It will be assigned zero on code paths that return normally.
3287 Typically this is a call-clobbered hard register that is otherwise
3288 untouched by the epilogue, but could also be a stack slot.
3290 Do not define this macro if the stack pointer is saved and restored
3291 by the regular prolog and epilog code in the call frame itself; in
3292 this case, the exception handling library routines will update the
3293 stack location to be restored in place. Otherwise, you must define
3294 this macro if you want to support call frame exception handling like
3295 that provided by DWARF 2.
3298 @defmac EH_RETURN_HANDLER_RTX
3299 A C expression whose value is RTL representing a location in which
3300 to store the address of an exception handler to which we should
3301 return. It will not be assigned on code paths that return normally.
3303 Typically this is the location in the call frame at which the normal
3304 return address is stored. For targets that return by popping an
3305 address off the stack, this might be a memory address just below
3306 the @emph{target} call frame rather than inside the current call
3307 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3308 been assigned, so it may be used to calculate the location of the
3311 Some targets have more complex requirements than storing to an
3312 address calculable during initial code generation. In that case
3313 the @code{eh_return} instruction pattern should be used instead.
3315 If you want to support call frame exception handling, you must
3316 define either this macro or the @code{eh_return} instruction pattern.
3319 @defmac RETURN_ADDR_OFFSET
3320 If defined, an integer-valued C expression for which rtl will be generated
3321 to add it to the exception handler address before it is searched in the
3322 exception handling tables, and to subtract it again from the address before
3323 using it to return to the exception handler.
3326 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3327 This macro chooses the encoding of pointers embedded in the exception
3328 handling sections. If at all possible, this should be defined such
3329 that the exception handling section will not require dynamic relocations,
3330 and so may be read-only.
3332 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3333 @var{global} is true if the symbol may be affected by dynamic relocations.
3334 The macro should return a combination of the @code{DW_EH_PE_*} defines
3335 as found in @file{dwarf2.h}.
3337 If this macro is not defined, pointers will not be encoded but
3338 represented directly.
3341 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3342 This macro allows the target to emit whatever special magic is required
3343 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3344 Generic code takes care of pc-relative and indirect encodings; this must
3345 be defined if the target uses text-relative or data-relative encodings.
3347 This is a C statement that branches to @var{done} if the format was
3348 handled. @var{encoding} is the format chosen, @var{size} is the number
3349 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3353 @defmac MD_UNWIND_SUPPORT
3354 A string specifying a file to be #include'd in unwind-dw2.c. The file
3355 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3358 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3359 This macro allows the target to add CPU and operating system specific
3360 code to the call-frame unwinder for use when there is no unwind data
3361 available. The most common reason to implement this macro is to unwind
3362 through signal frames.
3364 This macro is called from @code{uw_frame_state_for} in
3365 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3366 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3367 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3368 for the address of the code being executed and @code{context->cfa} for
3369 the stack pointer value. If the frame can be decoded, the register
3370 save addresses should be updated in @var{fs} and the macro should
3371 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3372 the macro should evaluate to @code{_URC_END_OF_STACK}.
3374 For proper signal handling in Java this macro is accompanied by
3375 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3378 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3379 This macro allows the target to add operating system specific code to the
3380 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3381 usually used for signal or interrupt frames.
3383 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3384 @var{context} is an @code{_Unwind_Context};
3385 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3386 for the abi and context in the @code{.unwabi} directive. If the
3387 @code{.unwabi} directive can be handled, the register save addresses should
3388 be updated in @var{fs}.
3391 @defmac TARGET_USES_WEAK_UNWIND_INFO
3392 A C expression that evaluates to true if the target requires unwind
3393 info to be given comdat linkage. Define it to be @code{1} if comdat
3394 linkage is necessary. The default is @code{0}.
3397 @node Stack Checking
3398 @subsection Specifying How Stack Checking is Done
3400 GCC will check that stack references are within the boundaries of the
3401 stack, if the option @option{-fstack-check} is specified, in one of
3406 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3407 will assume that you have arranged for full stack checking to be done
3408 at appropriate places in the configuration files. GCC will not do
3409 other special processing.
3412 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3413 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3414 that you have arranged for static stack checking (checking of the
3415 static stack frame of functions) to be done at appropriate places
3416 in the configuration files. GCC will only emit code to do dynamic
3417 stack checking (checking on dynamic stack allocations) using the third
3421 If neither of the above are true, GCC will generate code to periodically
3422 ``probe'' the stack pointer using the values of the macros defined below.
3425 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3426 GCC will change its allocation strategy for large objects if the option
3427 @option{-fstack-check} is specified: they will always be allocated
3428 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3430 @defmac STACK_CHECK_BUILTIN
3431 A nonzero value if stack checking is done by the configuration files in a
3432 machine-dependent manner. You should define this macro if stack checking
3433 is required by the ABI of your machine or if you would like to do stack
3434 checking in some more efficient way than the generic approach. The default
3435 value of this macro is zero.
3438 @defmac STACK_CHECK_STATIC_BUILTIN
3439 A nonzero value if static stack checking is done by the configuration files
3440 in a machine-dependent manner. You should define this macro if you would
3441 like to do static stack checking in some more efficient way than the generic
3442 approach. The default value of this macro is zero.
3445 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3446 An integer specifying the interval at which GCC must generate stack probe
3447 instructions, defined as 2 raised to this integer. You will normally
3448 define this macro so that the interval be no larger than the size of
3449 the ``guard pages'' at the end of a stack area. The default value
3450 of 12 (4096-byte interval) is suitable for most systems.
3453 @defmac STACK_CHECK_MOVING_SP
3454 An integer which is nonzero if GCC should move the stack pointer page by page
3455 when doing probes. This can be necessary on systems where the stack pointer
3456 contains the bottom address of the memory area accessible to the executing
3457 thread at any point in time. In this situation an alternate signal stack
3458 is required in order to be able to recover from a stack overflow. The
3459 default value of this macro is zero.
3462 @defmac STACK_CHECK_PROTECT
3463 The number of bytes of stack needed to recover from a stack overflow, for
3464 languages where such a recovery is supported. The default value of 75 words
3465 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3466 8192 bytes with other exception handling mechanisms should be adequate for
3470 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3471 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3472 in the opposite case.
3474 @defmac STACK_CHECK_MAX_FRAME_SIZE
3475 The maximum size of a stack frame, in bytes. GCC will generate probe
3476 instructions in non-leaf functions to ensure at least this many bytes of
3477 stack are available. If a stack frame is larger than this size, stack
3478 checking will not be reliable and GCC will issue a warning. The
3479 default is chosen so that GCC only generates one instruction on most
3480 systems. You should normally not change the default value of this macro.
3483 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3484 GCC uses this value to generate the above warning message. It
3485 represents the amount of fixed frame used by a function, not including
3486 space for any callee-saved registers, temporaries and user variables.
3487 You need only specify an upper bound for this amount and will normally
3488 use the default of four words.
3491 @defmac STACK_CHECK_MAX_VAR_SIZE
3492 The maximum size, in bytes, of an object that GCC will place in the
3493 fixed area of the stack frame when the user specifies
3494 @option{-fstack-check}.
3495 GCC computed the default from the values of the above macros and you will
3496 normally not need to override that default.
3500 @node Frame Registers
3501 @subsection Registers That Address the Stack Frame
3503 @c prevent bad page break with this line
3504 This discusses registers that address the stack frame.
3506 @defmac STACK_POINTER_REGNUM
3507 The register number of the stack pointer register, which must also be a
3508 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3509 the hardware determines which register this is.
3512 @defmac FRAME_POINTER_REGNUM
3513 The register number of the frame pointer register, which is used to
3514 access automatic variables in the stack frame. On some machines, the
3515 hardware determines which register this is. On other machines, you can
3516 choose any register you wish for this purpose.
3519 @defmac HARD_FRAME_POINTER_REGNUM
3520 On some machines the offset between the frame pointer and starting
3521 offset of the automatic variables is not known until after register
3522 allocation has been done (for example, because the saved registers are
3523 between these two locations). On those machines, define
3524 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3525 be used internally until the offset is known, and define
3526 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3527 used for the frame pointer.
3529 You should define this macro only in the very rare circumstances when it
3530 is not possible to calculate the offset between the frame pointer and
3531 the automatic variables until after register allocation has been
3532 completed. When this macro is defined, you must also indicate in your
3533 definition of @code{ELIMINABLE_REGS} how to eliminate
3534 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3535 or @code{STACK_POINTER_REGNUM}.
3537 Do not define this macro if it would be the same as
3538 @code{FRAME_POINTER_REGNUM}.
3541 @defmac ARG_POINTER_REGNUM
3542 The register number of the arg pointer register, which is used to access
3543 the function's argument list. On some machines, this is the same as the
3544 frame pointer register. On some machines, the hardware determines which
3545 register this is. On other machines, you can choose any register you
3546 wish for this purpose. If this is not the same register as the frame
3547 pointer register, then you must mark it as a fixed register according to
3548 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3549 (@pxref{Elimination}).
3552 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3553 Define this to a preprocessor constant that is nonzero if
3554 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3555 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3556 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3557 definition is not suitable for use in preprocessor conditionals.
3560 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3561 Define this to a preprocessor constant that is nonzero if
3562 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3563 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3564 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3565 definition is not suitable for use in preprocessor conditionals.
3568 @defmac RETURN_ADDRESS_POINTER_REGNUM
3569 The register number of the return address pointer register, which is used to
3570 access the current function's return address from the stack. On some
3571 machines, the return address is not at a fixed offset from the frame
3572 pointer or stack pointer or argument pointer. This register can be defined
3573 to point to the return address on the stack, and then be converted by
3574 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3576 Do not define this macro unless there is no other way to get the return
3577 address from the stack.
3580 @defmac STATIC_CHAIN_REGNUM
3581 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3582 Register numbers used for passing a function's static chain pointer. If
3583 register windows are used, the register number as seen by the called
3584 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3585 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3586 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3589 The static chain register need not be a fixed register.
3591 If the static chain is passed in memory, these macros should not be
3592 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3595 @hook TARGET_STATIC_CHAIN
3596 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3597 targets that may use different static chain locations for different
3598 nested functions. This may be required if the target has function
3599 attributes that affect the calling conventions of the function and
3600 those calling conventions use different static chain locations.
3602 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3604 If the static chain is passed in memory, this hook should be used to
3605 provide rtx giving @code{mem} expressions that denote where they are stored.
3606 Often the @code{mem} expression as seen by the caller will be at an offset
3607 from the stack pointer and the @code{mem} expression as seen by the callee
3608 will be at an offset from the frame pointer.
3609 @findex stack_pointer_rtx
3610 @findex frame_pointer_rtx
3611 @findex arg_pointer_rtx
3612 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3613 @code{arg_pointer_rtx} will have been initialized and should be used
3614 to refer to those items.
3617 @defmac DWARF_FRAME_REGISTERS
3618 This macro specifies the maximum number of hard registers that can be
3619 saved in a call frame. This is used to size data structures used in
3620 DWARF2 exception handling.
3622 Prior to GCC 3.0, this macro was needed in order to establish a stable
3623 exception handling ABI in the face of adding new hard registers for ISA
3624 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3625 in the number of hard registers. Nevertheless, this macro can still be
3626 used to reduce the runtime memory requirements of the exception handling
3627 routines, which can be substantial if the ISA contains a lot of
3628 registers that are not call-saved.
3630 If this macro is not defined, it defaults to
3631 @code{FIRST_PSEUDO_REGISTER}.
3634 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3636 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3637 for backward compatibility in pre GCC 3.0 compiled code.
3639 If this macro is not defined, it defaults to
3640 @code{DWARF_FRAME_REGISTERS}.
3643 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3645 Define this macro if the target's representation for dwarf registers
3646 is different than the internal representation for unwind column.
3647 Given a dwarf register, this macro should return the internal unwind
3648 column number to use instead.
3650 See the PowerPC's SPE target for an example.
3653 @defmac DWARF_FRAME_REGNUM (@var{regno})
3655 Define this macro if the target's representation for dwarf registers
3656 used in .eh_frame or .debug_frame is different from that used in other
3657 debug info sections. Given a GCC hard register number, this macro
3658 should return the .eh_frame register number. The default is
3659 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3663 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3665 Define this macro to map register numbers held in the call frame info
3666 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3667 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3668 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3669 return @code{@var{regno}}.
3674 @subsection Eliminating Frame Pointer and Arg Pointer
3676 @c prevent bad page break with this line
3677 This is about eliminating the frame pointer and arg pointer.
3679 @hook TARGET_FRAME_POINTER_REQUIRED
3680 This target hook should return @code{true} if a function must have and use
3681 a frame pointer. This target hook is called in the reload pass. If its return
3682 value is @code{true} the function will have a frame pointer.
3684 This target hook can in principle examine the current function and decide
3685 according to the facts, but on most machines the constant @code{false} or the
3686 constant @code{true} suffices. Use @code{false} when the machine allows code
3687 to be generated with no frame pointer, and doing so saves some time or space.
3688 Use @code{true} when there is no possible advantage to avoiding a frame
3691 In certain cases, the compiler does not know how to produce valid code
3692 without a frame pointer. The compiler recognizes those cases and
3693 automatically gives the function a frame pointer regardless of what
3694 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3697 In a function that does not require a frame pointer, the frame pointer
3698 register can be allocated for ordinary usage, unless you mark it as a
3699 fixed register. See @code{FIXED_REGISTERS} for more information.
3701 Default return value is @code{false}.
3704 @findex get_frame_size
3705 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3706 A C statement to store in the variable @var{depth-var} the difference
3707 between the frame pointer and the stack pointer values immediately after
3708 the function prologue. The value would be computed from information
3709 such as the result of @code{get_frame_size ()} and the tables of
3710 registers @code{regs_ever_live} and @code{call_used_regs}.
3712 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3713 need not be defined. Otherwise, it must be defined even if
3714 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3715 case, you may set @var{depth-var} to anything.
3718 @defmac ELIMINABLE_REGS
3719 If defined, this macro specifies a table of register pairs used to
3720 eliminate unneeded registers that point into the stack frame. If it is not
3721 defined, the only elimination attempted by the compiler is to replace
3722 references to the frame pointer with references to the stack pointer.
3724 The definition of this macro is a list of structure initializations, each
3725 of which specifies an original and replacement register.
3727 On some machines, the position of the argument pointer is not known until
3728 the compilation is completed. In such a case, a separate hard register
3729 must be used for the argument pointer. This register can be eliminated by
3730 replacing it with either the frame pointer or the argument pointer,
3731 depending on whether or not the frame pointer has been eliminated.
3733 In this case, you might specify:
3735 #define ELIMINABLE_REGS \
3736 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3737 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3738 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3741 Note that the elimination of the argument pointer with the stack pointer is
3742 specified first since that is the preferred elimination.
3745 @hook TARGET_CAN_ELIMINATE
3746 This target hook should returns @code{true} if the compiler is allowed to
3747 try to replace register number @var{from_reg} with register number
3748 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3749 is defined, and will usually be @code{true}, since most of the cases
3750 preventing register elimination are things that the compiler already
3753 Default return value is @code{true}.
3756 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3757 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3758 specifies the initial difference between the specified pair of
3759 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3763 @node Stack Arguments
3764 @subsection Passing Function Arguments on the Stack
3765 @cindex arguments on stack
3766 @cindex stack arguments
3768 The macros in this section control how arguments are passed
3769 on the stack. See the following section for other macros that
3770 control passing certain arguments in registers.
3772 @hook TARGET_PROMOTE_PROTOTYPES
3773 This target hook returns @code{true} if an argument declared in a
3774 prototype as an integral type smaller than @code{int} should actually be
3775 passed as an @code{int}. In addition to avoiding errors in certain
3776 cases of mismatch, it also makes for better code on certain machines.
3777 The default is to not promote prototypes.
3781 A C expression. If nonzero, push insns will be used to pass
3783 If the target machine does not have a push instruction, set it to zero.
3784 That directs GCC to use an alternate strategy: to
3785 allocate the entire argument block and then store the arguments into
3786 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3789 @defmac PUSH_ARGS_REVERSED
3790 A C expression. If nonzero, function arguments will be evaluated from
3791 last to first, rather than from first to last. If this macro is not
3792 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3793 and args grow in opposite directions, and 0 otherwise.
3796 @defmac PUSH_ROUNDING (@var{npushed})
3797 A C expression that is the number of bytes actually pushed onto the
3798 stack when an instruction attempts to push @var{npushed} bytes.
3800 On some machines, the definition
3803 #define PUSH_ROUNDING(BYTES) (BYTES)
3807 will suffice. But on other machines, instructions that appear
3808 to push one byte actually push two bytes in an attempt to maintain
3809 alignment. Then the definition should be
3812 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3815 If the value of this macro has a type, it should be an unsigned type.
3818 @findex current_function_outgoing_args_size
3819 @defmac ACCUMULATE_OUTGOING_ARGS
3820 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3821 will be computed and placed into the variable
3822 @code{current_function_outgoing_args_size}. No space will be pushed
3823 onto the stack for each call; instead, the function prologue should
3824 increase the stack frame size by this amount.
3826 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3830 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3831 Define this macro if functions should assume that stack space has been
3832 allocated for arguments even when their values are passed in
3835 The value of this macro is the size, in bytes, of the area reserved for
3836 arguments passed in registers for the function represented by @var{fndecl},
3837 which can be zero if GCC is calling a library function.
3838 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3841 This space can be allocated by the caller, or be a part of the
3842 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3845 @c above is overfull. not sure what to do. --mew 5feb93 did
3846 @c something, not sure if it looks good. --mew 10feb93
3848 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3849 Define this to a nonzero value if it is the responsibility of the
3850 caller to allocate the area reserved for arguments passed in registers
3851 when calling a function of @var{fntype}. @var{fntype} may be NULL
3852 if the function called is a library function.
3854 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3855 whether the space for these arguments counts in the value of
3856 @code{current_function_outgoing_args_size}.
3859 @defmac STACK_PARMS_IN_REG_PARM_AREA
3860 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3861 stack parameters don't skip the area specified by it.
3862 @c i changed this, makes more sens and it should have taken care of the
3863 @c overfull.. not as specific, tho. --mew 5feb93
3865 Normally, when a parameter is not passed in registers, it is placed on the
3866 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3867 suppresses this behavior and causes the parameter to be passed on the
3868 stack in its natural location.
3871 @hook TARGET_RETURN_POPS_ARGS
3872 This target hook returns the number of bytes of its own arguments that
3873 a function pops on returning, or 0 if the function pops no arguments
3874 and the caller must therefore pop them all after the function returns.
3876 @var{fundecl} is a C variable whose value is a tree node that describes
3877 the function in question. Normally it is a node of type
3878 @code{FUNCTION_DECL} that describes the declaration of the function.
3879 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3881 @var{funtype} is a C variable whose value is a tree node that
3882 describes the function in question. Normally it is a node of type
3883 @code{FUNCTION_TYPE} that describes the data type of the function.
3884 From this it is possible to obtain the data types of the value and
3885 arguments (if known).
3887 When a call to a library function is being considered, @var{fundecl}
3888 will contain an identifier node for the library function. Thus, if
3889 you need to distinguish among various library functions, you can do so
3890 by their names. Note that ``library function'' in this context means
3891 a function used to perform arithmetic, whose name is known specially
3892 in the compiler and was not mentioned in the C code being compiled.
3894 @var{size} is the number of bytes of arguments passed on the
3895 stack. If a variable number of bytes is passed, it is zero, and
3896 argument popping will always be the responsibility of the calling function.
3898 On the VAX, all functions always pop their arguments, so the definition
3899 of this macro is @var{size}. On the 68000, using the standard
3900 calling convention, no functions pop their arguments, so the value of
3901 the macro is always 0 in this case. But an alternative calling
3902 convention is available in which functions that take a fixed number of
3903 arguments pop them but other functions (such as @code{printf}) pop
3904 nothing (the caller pops all). When this convention is in use,
3905 @var{funtype} is examined to determine whether a function takes a fixed
3906 number of arguments.
3909 @defmac CALL_POPS_ARGS (@var{cum})
3910 A C expression that should indicate the number of bytes a call sequence
3911 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3912 when compiling a function call.
3914 @var{cum} is the variable in which all arguments to the called function
3915 have been accumulated.
3917 On certain architectures, such as the SH5, a call trampoline is used
3918 that pops certain registers off the stack, depending on the arguments
3919 that have been passed to the function. Since this is a property of the
3920 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3924 @node Register Arguments
3925 @subsection Passing Arguments in Registers
3926 @cindex arguments in registers
3927 @cindex registers arguments
3929 This section describes the macros which let you control how various
3930 types of arguments are passed in registers or how they are arranged in
3933 @hook TARGET_FUNCTION_ARG
3934 Return an RTX indicating whether a function argument is passed in a
3935 register and if so, which register.
3937 The arguments are @var{ca}, which summarizes all the previous
3938 arguments; @var{mode}, the machine mode of the argument; @var{type},
3939 the data type of the argument as a tree node or 0 if that is not known
3940 (which happens for C support library functions); and @var{named},
3941 which is @code{true} for an ordinary argument and @code{false} for
3942 nameless arguments that correspond to @samp{@dots{}} in the called
3943 function's prototype. @var{type} can be an incomplete type if a
3944 syntax error has previously occurred.
3946 The return value is usually either a @code{reg} RTX for the hard
3947 register in which to pass the argument, or zero to pass the argument
3950 The value of the expression can also be a @code{parallel} RTX@. This is
3951 used when an argument is passed in multiple locations. The mode of the
3952 @code{parallel} should be the mode of the entire argument. The
3953 @code{parallel} holds any number of @code{expr_list} pairs; each one
3954 describes where part of the argument is passed. In each
3955 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3956 register in which to pass this part of the argument, and the mode of the
3957 register RTX indicates how large this part of the argument is. The
3958 second operand of the @code{expr_list} is a @code{const_int} which gives
3959 the offset in bytes into the entire argument of where this part starts.
3960 As a special exception the first @code{expr_list} in the @code{parallel}
3961 RTX may have a first operand of zero. This indicates that the entire
3962 argument is also stored on the stack.
3964 The last time this hook is called, it is called with @code{MODE ==
3965 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3966 pattern as operands 2 and 3 respectively.
3968 @cindex @file{stdarg.h} and register arguments
3969 The usual way to make the ISO library @file{stdarg.h} work on a
3970 machine where some arguments are usually passed in registers, is to
3971 cause nameless arguments to be passed on the stack instead. This is
3972 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
3973 @var{named} is @code{false}.
3975 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
3976 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
3977 You may use the hook @code{targetm.calls.must_pass_in_stack}
3978 in the definition of this macro to determine if this argument is of a
3979 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3980 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
3981 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3982 defined, the argument will be computed in the stack and then loaded into
3986 @hook TARGET_MUST_PASS_IN_STACK
3987 This target hook should return @code{true} if we should not pass @var{type}
3988 solely in registers. The file @file{expr.h} defines a
3989 definition that is usually appropriate, refer to @file{expr.h} for additional
3993 @hook TARGET_FUNCTION_INCOMING_ARG
3994 Define this hook if the target machine has ``register windows'', so
3995 that the register in which a function sees an arguments is not
3996 necessarily the same as the one in which the caller passed the
3999 For such machines, @code{TARGET_FUNCTION_ARG} computes the register in
4000 which the caller passes the value, and
4001 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4002 fashion to tell the function being called where the arguments will
4005 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4006 @code{TARGET_FUNCTION_ARG} serves both purposes.
4009 @hook TARGET_ARG_PARTIAL_BYTES
4010 This target hook returns the number of bytes at the beginning of an
4011 argument that must be put in registers. The value must be zero for
4012 arguments that are passed entirely in registers or that are entirely
4013 pushed on the stack.
4015 On some machines, certain arguments must be passed partially in
4016 registers and partially in memory. On these machines, typically the
4017 first few words of arguments are passed in registers, and the rest
4018 on the stack. If a multi-word argument (a @code{double} or a
4019 structure) crosses that boundary, its first few words must be passed
4020 in registers and the rest must be pushed. This macro tells the
4021 compiler when this occurs, and how many bytes should go in registers.
4023 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4024 register to be used by the caller for this argument; likewise
4025 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4028 @hook TARGET_PASS_BY_REFERENCE
4029 This target hook should return @code{true} if an argument at the
4030 position indicated by @var{cum} should be passed by reference. This
4031 predicate is queried after target independent reasons for being
4032 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4034 If the hook returns true, a copy of that argument is made in memory and a
4035 pointer to the argument is passed instead of the argument itself.
4036 The pointer is passed in whatever way is appropriate for passing a pointer
4040 @hook TARGET_CALLEE_COPIES
4041 The function argument described by the parameters to this hook is
4042 known to be passed by reference. The hook should return true if the
4043 function argument should be copied by the callee instead of copied
4046 For any argument for which the hook returns true, if it can be
4047 determined that the argument is not modified, then a copy need
4050 The default version of this hook always returns false.
4053 @defmac CUMULATIVE_ARGS
4054 A C type for declaring a variable that is used as the first argument
4055 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4056 target machines, the type @code{int} suffices and can hold the number
4057 of bytes of argument so far.
4059 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4060 arguments that have been passed on the stack. The compiler has other
4061 variables to keep track of that. For target machines on which all
4062 arguments are passed on the stack, there is no need to store anything in
4063 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4064 should not be empty, so use @code{int}.
4067 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4068 If defined, this macro is called before generating any code for a
4069 function, but after the @var{cfun} descriptor for the function has been
4070 created. The back end may use this macro to update @var{cfun} to
4071 reflect an ABI other than that which would normally be used by default.
4072 If the compiler is generating code for a compiler-generated function,
4073 @var{fndecl} may be @code{NULL}.
4076 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4077 A C statement (sans semicolon) for initializing the variable
4078 @var{cum} for the state at the beginning of the argument list. The
4079 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4080 is the tree node for the data type of the function which will receive
4081 the args, or 0 if the args are to a compiler support library function.
4082 For direct calls that are not libcalls, @var{fndecl} contain the
4083 declaration node of the function. @var{fndecl} is also set when
4084 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4085 being compiled. @var{n_named_args} is set to the number of named
4086 arguments, including a structure return address if it is passed as a
4087 parameter, when making a call. When processing incoming arguments,
4088 @var{n_named_args} is set to @minus{}1.
4090 When processing a call to a compiler support library function,
4091 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4092 contains the name of the function, as a string. @var{libname} is 0 when
4093 an ordinary C function call is being processed. Thus, each time this
4094 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4095 never both of them at once.
4098 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4099 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4100 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4101 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4102 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4103 0)} is used instead.
4106 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4107 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4108 finding the arguments for the function being compiled. If this macro is
4109 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4111 The value passed for @var{libname} is always 0, since library routines
4112 with special calling conventions are never compiled with GCC@. The
4113 argument @var{libname} exists for symmetry with
4114 @code{INIT_CUMULATIVE_ARGS}.
4115 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4116 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4119 @hook TARGET_FUNCTION_ARG_ADVANCE
4120 This hook updates the summarizer variable pointed to by @var{ca} to
4121 advance past an argument in the argument list. The values @var{mode},
4122 @var{type} and @var{named} describe that argument. Once this is done,
4123 the variable @var{cum} is suitable for analyzing the @emph{following}
4124 argument with @code{TARGET_FUNCTION_ARG}, etc.
4126 This hook need not do anything if the argument in question was passed
4127 on the stack. The compiler knows how to track the amount of stack space
4128 used for arguments without any special help.
4131 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4132 If defined, a C expression that is the number of bytes to add to the
4133 offset of the argument passed in memory. This is needed for the SPU,
4134 which passes @code{char} and @code{short} arguments in the preferred
4135 slot that is in the middle of the quad word instead of starting at the
4139 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4140 If defined, a C expression which determines whether, and in which direction,
4141 to pad out an argument with extra space. The value should be of type
4142 @code{enum direction}: either @code{upward} to pad above the argument,
4143 @code{downward} to pad below, or @code{none} to inhibit padding.
4145 The @emph{amount} of padding is always just enough to reach the next
4146 multiple of @code{TARGET_FUNCTION_ARG_BOUNDARY}; this macro does not
4149 This macro has a default definition which is right for most systems.
4150 For little-endian machines, the default is to pad upward. For
4151 big-endian machines, the default is to pad downward for an argument of
4152 constant size shorter than an @code{int}, and upward otherwise.
4155 @defmac PAD_VARARGS_DOWN
4156 If defined, a C expression which determines whether the default
4157 implementation of va_arg will attempt to pad down before reading the
4158 next argument, if that argument is smaller than its aligned space as
4159 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4160 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4163 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4164 Specify padding for the last element of a block move between registers and
4165 memory. @var{first} is nonzero if this is the only element. Defining this
4166 macro allows better control of register function parameters on big-endian
4167 machines, without using @code{PARALLEL} rtl. In particular,
4168 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4169 registers, as there is no longer a "wrong" part of a register; For example,
4170 a three byte aggregate may be passed in the high part of a register if so
4174 @hook TARGET_FUNCTION_ARG_BOUNDARY
4175 This hook returns the alignment boundary, in bits, of an argument
4176 with the specified mode and type. The default hook returns
4177 @code{PARM_BOUNDARY} for all arguments.
4180 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4181 A C expression that is nonzero if @var{regno} is the number of a hard
4182 register in which function arguments are sometimes passed. This does
4183 @emph{not} include implicit arguments such as the static chain and
4184 the structure-value address. On many machines, no registers can be
4185 used for this purpose since all function arguments are pushed on the
4189 @hook TARGET_SPLIT_COMPLEX_ARG
4190 This hook should return true if parameter of type @var{type} are passed
4191 as two scalar parameters. By default, GCC will attempt to pack complex
4192 arguments into the target's word size. Some ABIs require complex arguments
4193 to be split and treated as their individual components. For example, on
4194 AIX64, complex floats should be passed in a pair of floating point
4195 registers, even though a complex float would fit in one 64-bit floating
4198 The default value of this hook is @code{NULL}, which is treated as always
4202 @hook TARGET_BUILD_BUILTIN_VA_LIST
4203 This hook returns a type node for @code{va_list} for the target.
4204 The default version of the hook returns @code{void*}.
4207 @hook TARGET_ENUM_VA_LIST_P
4208 This target hook is used in function @code{c_common_nodes_and_builtins}
4209 to iterate through the target specific builtin types for va_list. The
4210 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4211 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4213 The arguments @var{pname} and @var{ptree} are used to store the result of
4214 this macro and are set to the name of the va_list builtin type and its
4216 If the return value of this macro is zero, then there is no more element.
4217 Otherwise the @var{IDX} should be increased for the next call of this
4218 macro to iterate through all types.
4221 @hook TARGET_FN_ABI_VA_LIST
4222 This hook returns the va_list type of the calling convention specified by
4224 The default version of this hook returns @code{va_list_type_node}.
4227 @hook TARGET_CANONICAL_VA_LIST_TYPE
4228 This hook returns the va_list type of the calling convention specified by the
4229 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4233 @hook TARGET_GIMPLIFY_VA_ARG_EXPR
4234 This hook performs target-specific gimplification of
4235 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4236 arguments to @code{va_arg}; the latter two are as in
4237 @code{gimplify.c:gimplify_expr}.
4240 @hook TARGET_VALID_POINTER_MODE
4241 Define this to return nonzero if the port can handle pointers
4242 with machine mode @var{mode}. The default version of this
4243 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4246 @hook TARGET_REF_MAY_ALIAS_ERRNO
4248 @hook TARGET_SCALAR_MODE_SUPPORTED_P
4249 Define this to return nonzero if the port is prepared to handle
4250 insns involving scalar mode @var{mode}. For a scalar mode to be
4251 considered supported, all the basic arithmetic and comparisons
4254 The default version of this hook returns true for any mode
4255 required to handle the basic C types (as defined by the port).
4256 Included here are the double-word arithmetic supported by the
4257 code in @file{optabs.c}.
4260 @hook TARGET_VECTOR_MODE_SUPPORTED_P
4261 Define this to return nonzero if the port is prepared to handle
4262 insns involving vector mode @var{mode}. At the very least, it
4263 must have move patterns for this mode.
4266 @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
4267 Define this to return nonzero for machine modes for which the port has
4268 small register classes. If this target hook returns nonzero for a given
4269 @var{mode}, the compiler will try to minimize the lifetime of registers
4270 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4271 In this case, the hook is expected to return nonzero if it returns nonzero
4274 On some machines, it is risky to let hard registers live across arbitrary
4275 insns. Typically, these machines have instructions that require values
4276 to be in specific registers (like an accumulator), and reload will fail
4277 if the required hard register is used for another purpose across such an
4280 Passes before reload do not know which hard registers will be used
4281 in an instruction, but the machine modes of the registers set or used in
4282 the instruction are already known. And for some machines, register
4283 classes are small for, say, integer registers but not for floating point
4284 registers. For example, the AMD x86-64 architecture requires specific
4285 registers for the legacy x86 integer instructions, but there are many
4286 SSE registers for floating point operations. On such targets, a good
4287 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4288 machine modes but zero for the SSE register classes.
4290 The default version of this hook returns false for any mode. It is always
4291 safe to redefine this hook to return with a nonzero value. But if you
4292 unnecessarily define it, you will reduce the amount of optimizations
4293 that can be performed in some cases. If you do not define this hook
4294 to return a nonzero value when it is required, the compiler will run out
4295 of spill registers and print a fatal error message.
4298 @hook TARGET_FLAGS_REGNUM
4301 @subsection How Scalar Function Values Are Returned
4302 @cindex return values in registers
4303 @cindex values, returned by functions
4304 @cindex scalars, returned as values
4306 This section discusses the macros that control returning scalars as
4307 values---values that can fit in registers.
4309 @hook TARGET_FUNCTION_VALUE
4311 Define this to return an RTX representing the place where a function
4312 returns or receives a value of data type @var{ret_type}, a tree node
4313 representing a data type. @var{fn_decl_or_type} is a tree node
4314 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4315 function being called. If @var{outgoing} is false, the hook should
4316 compute the register in which the caller will see the return value.
4317 Otherwise, the hook should return an RTX representing the place where
4318 a function returns a value.
4320 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4321 (Actually, on most machines, scalar values are returned in the same
4322 place regardless of mode.) The value of the expression is usually a
4323 @code{reg} RTX for the hard register where the return value is stored.
4324 The value can also be a @code{parallel} RTX, if the return value is in
4325 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4326 @code{parallel} form. Note that the callee will populate every
4327 location specified in the @code{parallel}, but if the first element of
4328 the @code{parallel} contains the whole return value, callers will use
4329 that element as the canonical location and ignore the others. The m68k
4330 port uses this type of @code{parallel} to return pointers in both
4331 @samp{%a0} (the canonical location) and @samp{%d0}.
4333 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4334 the same promotion rules specified in @code{PROMOTE_MODE} if
4335 @var{valtype} is a scalar type.
4337 If the precise function being called is known, @var{func} is a tree
4338 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4339 pointer. This makes it possible to use a different value-returning
4340 convention for specific functions when all their calls are
4343 Some target machines have ``register windows'' so that the register in
4344 which a function returns its value is not the same as the one in which
4345 the caller sees the value. For such machines, you should return
4346 different RTX depending on @var{outgoing}.
4348 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4349 aggregate data types, because these are returned in another way. See
4350 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4353 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4354 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4355 a new target instead.
4358 @defmac LIBCALL_VALUE (@var{mode})
4359 A C expression to create an RTX representing the place where a library
4360 function returns a value of mode @var{mode}.
4362 Note that ``library function'' in this context means a compiler
4363 support routine, used to perform arithmetic, whose name is known
4364 specially by the compiler and was not mentioned in the C code being
4368 @hook TARGET_LIBCALL_VALUE
4369 Define this hook if the back-end needs to know the name of the libcall
4370 function in order to determine where the result should be returned.
4372 The mode of the result is given by @var{mode} and the name of the called
4373 library function is given by @var{fun}. The hook should return an RTX
4374 representing the place where the library function result will be returned.
4376 If this hook is not defined, then LIBCALL_VALUE will be used.
4379 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4380 A C expression that is nonzero if @var{regno} is the number of a hard
4381 register in which the values of called function may come back.
4383 A register whose use for returning values is limited to serving as the
4384 second of a pair (for a value of type @code{double}, say) need not be
4385 recognized by this macro. So for most machines, this definition
4389 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4392 If the machine has register windows, so that the caller and the called
4393 function use different registers for the return value, this macro
4394 should recognize only the caller's register numbers.
4396 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4397 for a new target instead.
4400 @hook TARGET_FUNCTION_VALUE_REGNO_P
4401 A target hook that return @code{true} if @var{regno} is the number of a hard
4402 register in which the values of called function may come back.
4404 A register whose use for returning values is limited to serving as the
4405 second of a pair (for a value of type @code{double}, say) need not be
4406 recognized by this target hook.
4408 If the machine has register windows, so that the caller and the called
4409 function use different registers for the return value, this target hook
4410 should recognize only the caller's register numbers.
4412 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4415 @defmac APPLY_RESULT_SIZE
4416 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4417 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4418 saving and restoring an arbitrary return value.
4421 @hook TARGET_RETURN_IN_MSB
4422 This hook should return true if values of type @var{type} are returned
4423 at the most significant end of a register (in other words, if they are
4424 padded at the least significant end). You can assume that @var{type}
4425 is returned in a register; the caller is required to check this.
4427 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4428 be able to hold the complete return value. For example, if a 1-, 2-
4429 or 3-byte structure is returned at the most significant end of a
4430 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4434 @node Aggregate Return
4435 @subsection How Large Values Are Returned
4436 @cindex aggregates as return values
4437 @cindex large return values
4438 @cindex returning aggregate values
4439 @cindex structure value address
4441 When a function value's mode is @code{BLKmode} (and in some other
4442 cases), the value is not returned according to
4443 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4444 caller passes the address of a block of memory in which the value
4445 should be stored. This address is called the @dfn{structure value
4448 This section describes how to control returning structure values in
4451 @hook TARGET_RETURN_IN_MEMORY
4452 This target hook should return a nonzero value to say to return the
4453 function value in memory, just as large structures are always returned.
4454 Here @var{type} will be the data type of the value, and @var{fntype}
4455 will be the type of the function doing the returning, or @code{NULL} for
4458 Note that values of mode @code{BLKmode} must be explicitly handled
4459 by this function. Also, the option @option{-fpcc-struct-return}
4460 takes effect regardless of this macro. On most systems, it is
4461 possible to leave the hook undefined; this causes a default
4462 definition to be used, whose value is the constant 1 for @code{BLKmode}
4463 values, and 0 otherwise.
4465 Do not use this hook to indicate that structures and unions should always
4466 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4470 @defmac DEFAULT_PCC_STRUCT_RETURN
4471 Define this macro to be 1 if all structure and union return values must be
4472 in memory. Since this results in slower code, this should be defined
4473 only if needed for compatibility with other compilers or with an ABI@.
4474 If you define this macro to be 0, then the conventions used for structure
4475 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4478 If not defined, this defaults to the value 1.
4481 @hook TARGET_STRUCT_VALUE_RTX
4482 This target hook should return the location of the structure value
4483 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4484 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4485 be @code{NULL}, for libcalls. You do not need to define this target
4486 hook if the address is always passed as an ``invisible'' first
4489 On some architectures the place where the structure value address
4490 is found by the called function is not the same place that the
4491 caller put it. This can be due to register windows, or it could
4492 be because the function prologue moves it to a different place.
4493 @var{incoming} is @code{1} or @code{2} when the location is needed in
4494 the context of the called function, and @code{0} in the context of
4497 If @var{incoming} is nonzero and the address is to be found on the
4498 stack, return a @code{mem} which refers to the frame pointer. If
4499 @var{incoming} is @code{2}, the result is being used to fetch the
4500 structure value address at the beginning of a function. If you need
4501 to emit adjusting code, you should do it at this point.
4504 @defmac PCC_STATIC_STRUCT_RETURN
4505 Define this macro if the usual system convention on the target machine
4506 for returning structures and unions is for the called function to return
4507 the address of a static variable containing the value.
4509 Do not define this if the usual system convention is for the caller to
4510 pass an address to the subroutine.
4512 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4513 nothing when you use @option{-freg-struct-return} mode.
4516 @hook TARGET_GET_RAW_RESULT_MODE
4518 @hook TARGET_GET_RAW_ARG_MODE
4521 @subsection Caller-Saves Register Allocation
4523 If you enable it, GCC can save registers around function calls. This
4524 makes it possible to use call-clobbered registers to hold variables that
4525 must live across calls.
4527 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4528 A C expression to determine whether it is worthwhile to consider placing
4529 a pseudo-register in a call-clobbered hard register and saving and
4530 restoring it around each function call. The expression should be 1 when
4531 this is worth doing, and 0 otherwise.
4533 If you don't define this macro, a default is used which is good on most
4534 machines: @code{4 * @var{calls} < @var{refs}}.
4537 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4538 A C expression specifying which mode is required for saving @var{nregs}
4539 of a pseudo-register in call-clobbered hard register @var{regno}. If
4540 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4541 returned. For most machines this macro need not be defined since GCC
4542 will select the smallest suitable mode.
4545 @node Function Entry
4546 @subsection Function Entry and Exit
4547 @cindex function entry and exit
4551 This section describes the macros that output function entry
4552 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4554 @hook TARGET_ASM_FUNCTION_PROLOGUE
4555 If defined, a function that outputs the assembler code for entry to a
4556 function. The prologue is responsible for setting up the stack frame,
4557 initializing the frame pointer register, saving registers that must be
4558 saved, and allocating @var{size} additional bytes of storage for the
4559 local variables. @var{size} is an integer. @var{file} is a stdio
4560 stream to which the assembler code should be output.
4562 The label for the beginning of the function need not be output by this
4563 macro. That has already been done when the macro is run.
4565 @findex regs_ever_live
4566 To determine which registers to save, the macro can refer to the array
4567 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4568 @var{r} is used anywhere within the function. This implies the function
4569 prologue should save register @var{r}, provided it is not one of the
4570 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4571 @code{regs_ever_live}.)
4573 On machines that have ``register windows'', the function entry code does
4574 not save on the stack the registers that are in the windows, even if
4575 they are supposed to be preserved by function calls; instead it takes
4576 appropriate steps to ``push'' the register stack, if any non-call-used
4577 registers are used in the function.
4579 @findex frame_pointer_needed
4580 On machines where functions may or may not have frame-pointers, the
4581 function entry code must vary accordingly; it must set up the frame
4582 pointer if one is wanted, and not otherwise. To determine whether a
4583 frame pointer is in wanted, the macro can refer to the variable
4584 @code{frame_pointer_needed}. The variable's value will be 1 at run
4585 time in a function that needs a frame pointer. @xref{Elimination}.
4587 The function entry code is responsible for allocating any stack space
4588 required for the function. This stack space consists of the regions
4589 listed below. In most cases, these regions are allocated in the
4590 order listed, with the last listed region closest to the top of the
4591 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4592 the highest address if it is not defined). You can use a different order
4593 for a machine if doing so is more convenient or required for
4594 compatibility reasons. Except in cases where required by standard
4595 or by a debugger, there is no reason why the stack layout used by GCC
4596 need agree with that used by other compilers for a machine.
4599 @hook TARGET_ASM_FUNCTION_END_PROLOGUE
4600 If defined, a function that outputs assembler code at the end of a
4601 prologue. This should be used when the function prologue is being
4602 emitted as RTL, and you have some extra assembler that needs to be
4603 emitted. @xref{prologue instruction pattern}.
4606 @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
4607 If defined, a function that outputs assembler code at the start of an
4608 epilogue. This should be used when the function epilogue is being
4609 emitted as RTL, and you have some extra assembler that needs to be
4610 emitted. @xref{epilogue instruction pattern}.
4613 @hook TARGET_ASM_FUNCTION_EPILOGUE
4614 If defined, a function that outputs the assembler code for exit from a
4615 function. The epilogue is responsible for restoring the saved
4616 registers and stack pointer to their values when the function was
4617 called, and returning control to the caller. This macro takes the
4618 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4619 registers to restore are determined from @code{regs_ever_live} and
4620 @code{CALL_USED_REGISTERS} in the same way.
4622 On some machines, there is a single instruction that does all the work
4623 of returning from the function. On these machines, give that
4624 instruction the name @samp{return} and do not define the macro
4625 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4627 Do not define a pattern named @samp{return} if you want the
4628 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4629 switches to control whether return instructions or epilogues are used,
4630 define a @samp{return} pattern with a validity condition that tests the
4631 target switches appropriately. If the @samp{return} pattern's validity
4632 condition is false, epilogues will be used.
4634 On machines where functions may or may not have frame-pointers, the
4635 function exit code must vary accordingly. Sometimes the code for these
4636 two cases is completely different. To determine whether a frame pointer
4637 is wanted, the macro can refer to the variable
4638 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4639 a function that needs a frame pointer.
4641 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4642 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4643 The C variable @code{current_function_is_leaf} is nonzero for such a
4644 function. @xref{Leaf Functions}.
4646 On some machines, some functions pop their arguments on exit while
4647 others leave that for the caller to do. For example, the 68020 when
4648 given @option{-mrtd} pops arguments in functions that take a fixed
4649 number of arguments.
4651 @findex current_function_pops_args
4652 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4653 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4654 needs to know what was decided. The number of bytes of the current
4655 function's arguments that this function should pop is available in
4656 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4661 @findex current_function_pretend_args_size
4662 A region of @code{current_function_pretend_args_size} bytes of
4663 uninitialized space just underneath the first argument arriving on the
4664 stack. (This may not be at the very start of the allocated stack region
4665 if the calling sequence has pushed anything else since pushing the stack
4666 arguments. But usually, on such machines, nothing else has been pushed
4667 yet, because the function prologue itself does all the pushing.) This
4668 region is used on machines where an argument may be passed partly in
4669 registers and partly in memory, and, in some cases to support the
4670 features in @code{<stdarg.h>}.
4673 An area of memory used to save certain registers used by the function.
4674 The size of this area, which may also include space for such things as
4675 the return address and pointers to previous stack frames, is
4676 machine-specific and usually depends on which registers have been used
4677 in the function. Machines with register windows often do not require
4681 A region of at least @var{size} bytes, possibly rounded up to an allocation
4682 boundary, to contain the local variables of the function. On some machines,
4683 this region and the save area may occur in the opposite order, with the
4684 save area closer to the top of the stack.
4687 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4688 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4689 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4690 argument lists of the function. @xref{Stack Arguments}.
4693 @defmac EXIT_IGNORE_STACK
4694 Define this macro as a C expression that is nonzero if the return
4695 instruction or the function epilogue ignores the value of the stack
4696 pointer; in other words, if it is safe to delete an instruction to
4697 adjust the stack pointer before a return from the function. The
4700 Note that this macro's value is relevant only for functions for which
4701 frame pointers are maintained. It is never safe to delete a final
4702 stack adjustment in a function that has no frame pointer, and the
4703 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4706 @defmac EPILOGUE_USES (@var{regno})
4707 Define this macro as a C expression that is nonzero for registers that are
4708 used by the epilogue or the @samp{return} pattern. The stack and frame
4709 pointer registers are already assumed to be used as needed.
4712 @defmac EH_USES (@var{regno})
4713 Define this macro as a C expression that is nonzero for registers that are
4714 used by the exception handling mechanism, and so should be considered live
4715 on entry to an exception edge.
4718 @defmac DELAY_SLOTS_FOR_EPILOGUE
4719 Define this macro if the function epilogue contains delay slots to which
4720 instructions from the rest of the function can be ``moved''. The
4721 definition should be a C expression whose value is an integer
4722 representing the number of delay slots there.
4725 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4726 A C expression that returns 1 if @var{insn} can be placed in delay
4727 slot number @var{n} of the epilogue.
4729 The argument @var{n} is an integer which identifies the delay slot now
4730 being considered (since different slots may have different rules of
4731 eligibility). It is never negative and is always less than the number
4732 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4733 If you reject a particular insn for a given delay slot, in principle, it
4734 may be reconsidered for a subsequent delay slot. Also, other insns may
4735 (at least in principle) be considered for the so far unfilled delay
4738 @findex current_function_epilogue_delay_list
4739 @findex final_scan_insn
4740 The insns accepted to fill the epilogue delay slots are put in an RTL
4741 list made with @code{insn_list} objects, stored in the variable
4742 @code{current_function_epilogue_delay_list}. The insn for the first
4743 delay slot comes first in the list. Your definition of the macro
4744 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4745 outputting the insns in this list, usually by calling
4746 @code{final_scan_insn}.
4748 You need not define this macro if you did not define
4749 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4752 @hook TARGET_ASM_OUTPUT_MI_THUNK
4753 A function that outputs the assembler code for a thunk
4754 function, used to implement C++ virtual function calls with multiple
4755 inheritance. The thunk acts as a wrapper around a virtual function,
4756 adjusting the implicit object parameter before handing control off to
4759 First, emit code to add the integer @var{delta} to the location that
4760 contains the incoming first argument. Assume that this argument
4761 contains a pointer, and is the one used to pass the @code{this} pointer
4762 in C++. This is the incoming argument @emph{before} the function prologue,
4763 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4764 all other incoming arguments.
4766 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4767 made after adding @code{delta}. In particular, if @var{p} is the
4768 adjusted pointer, the following adjustment should be made:
4771 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4774 After the additions, emit code to jump to @var{function}, which is a
4775 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4776 not touch the return address. Hence returning from @var{FUNCTION} will
4777 return to whoever called the current @samp{thunk}.
4779 The effect must be as if @var{function} had been called directly with
4780 the adjusted first argument. This macro is responsible for emitting all
4781 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4782 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4784 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4785 have already been extracted from it.) It might possibly be useful on
4786 some targets, but probably not.
4788 If you do not define this macro, the target-independent code in the C++
4789 front end will generate a less efficient heavyweight thunk that calls
4790 @var{function} instead of jumping to it. The generic approach does
4791 not support varargs.
4794 @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
4795 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4796 to output the assembler code for the thunk function specified by the
4797 arguments it is passed, and false otherwise. In the latter case, the
4798 generic approach will be used by the C++ front end, with the limitations
4803 @subsection Generating Code for Profiling
4804 @cindex profiling, code generation
4806 These macros will help you generate code for profiling.
4808 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4809 A C statement or compound statement to output to @var{file} some
4810 assembler code to call the profiling subroutine @code{mcount}.
4813 The details of how @code{mcount} expects to be called are determined by
4814 your operating system environment, not by GCC@. To figure them out,
4815 compile a small program for profiling using the system's installed C
4816 compiler and look at the assembler code that results.
4818 Older implementations of @code{mcount} expect the address of a counter
4819 variable to be loaded into some register. The name of this variable is
4820 @samp{LP} followed by the number @var{labelno}, so you would generate
4821 the name using @samp{LP%d} in a @code{fprintf}.
4824 @defmac PROFILE_HOOK
4825 A C statement or compound statement to output to @var{file} some assembly
4826 code to call the profiling subroutine @code{mcount} even the target does
4827 not support profiling.
4830 @defmac NO_PROFILE_COUNTERS
4831 Define this macro to be an expression with a nonzero value if the
4832 @code{mcount} subroutine on your system does not need a counter variable
4833 allocated for each function. This is true for almost all modern
4834 implementations. If you define this macro, you must not use the
4835 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4838 @defmac PROFILE_BEFORE_PROLOGUE
4839 Define this macro if the code for function profiling should come before
4840 the function prologue. Normally, the profiling code comes after.
4844 @subsection Permitting tail calls
4847 @hook TARGET_FUNCTION_OK_FOR_SIBCALL
4848 True if it is ok to do sibling call optimization for the specified
4849 call expression @var{exp}. @var{decl} will be the called function,
4850 or @code{NULL} if this is an indirect call.
4852 It is not uncommon for limitations of calling conventions to prevent
4853 tail calls to functions outside the current unit of translation, or
4854 during PIC compilation. The hook is used to enforce these restrictions,
4855 as the @code{sibcall} md pattern can not fail, or fall over to a
4856 ``normal'' call. The criteria for successful sibling call optimization
4857 may vary greatly between different architectures.
4860 @hook TARGET_EXTRA_LIVE_ON_ENTRY
4861 Add any hard registers to @var{regs} that are live on entry to the
4862 function. This hook only needs to be defined to provide registers that
4863 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4864 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4865 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4866 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4869 @node Stack Smashing Protection
4870 @subsection Stack smashing protection
4871 @cindex stack smashing protection
4873 @hook TARGET_STACK_PROTECT_GUARD
4874 This hook returns a @code{DECL} node for the external variable to use
4875 for the stack protection guard. This variable is initialized by the
4876 runtime to some random value and is used to initialize the guard value
4877 that is placed at the top of the local stack frame. The type of this
4878 variable must be @code{ptr_type_node}.
4880 The default version of this hook creates a variable called
4881 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4884 @hook TARGET_STACK_PROTECT_FAIL
4885 This hook returns a tree expression that alerts the runtime that the
4886 stack protect guard variable has been modified. This expression should
4887 involve a call to a @code{noreturn} function.
4889 The default version of this hook invokes a function called
4890 @samp{__stack_chk_fail}, taking no arguments. This function is
4891 normally defined in @file{libgcc2.c}.
4894 @hook TARGET_SUPPORTS_SPLIT_STACK
4897 @section Implementing the Varargs Macros
4898 @cindex varargs implementation
4900 GCC comes with an implementation of @code{<varargs.h>} and
4901 @code{<stdarg.h>} that work without change on machines that pass arguments
4902 on the stack. Other machines require their own implementations of
4903 varargs, and the two machine independent header files must have
4904 conditionals to include it.
4906 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4907 the calling convention for @code{va_start}. The traditional
4908 implementation takes just one argument, which is the variable in which
4909 to store the argument pointer. The ISO implementation of
4910 @code{va_start} takes an additional second argument. The user is
4911 supposed to write the last named argument of the function here.
4913 However, @code{va_start} should not use this argument. The way to find
4914 the end of the named arguments is with the built-in functions described
4917 @defmac __builtin_saveregs ()
4918 Use this built-in function to save the argument registers in memory so
4919 that the varargs mechanism can access them. Both ISO and traditional
4920 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4921 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4923 On some machines, @code{__builtin_saveregs} is open-coded under the
4924 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4925 other machines, it calls a routine written in assembler language,
4926 found in @file{libgcc2.c}.
4928 Code generated for the call to @code{__builtin_saveregs} appears at the
4929 beginning of the function, as opposed to where the call to
4930 @code{__builtin_saveregs} is written, regardless of what the code is.
4931 This is because the registers must be saved before the function starts
4932 to use them for its own purposes.
4933 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4937 @defmac __builtin_next_arg (@var{lastarg})
4938 This builtin returns the address of the first anonymous stack
4939 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4940 returns the address of the location above the first anonymous stack
4941 argument. Use it in @code{va_start} to initialize the pointer for
4942 fetching arguments from the stack. Also use it in @code{va_start} to
4943 verify that the second parameter @var{lastarg} is the last named argument
4944 of the current function.
4947 @defmac __builtin_classify_type (@var{object})
4948 Since each machine has its own conventions for which data types are
4949 passed in which kind of register, your implementation of @code{va_arg}
4950 has to embody these conventions. The easiest way to categorize the
4951 specified data type is to use @code{__builtin_classify_type} together
4952 with @code{sizeof} and @code{__alignof__}.
4954 @code{__builtin_classify_type} ignores the value of @var{object},
4955 considering only its data type. It returns an integer describing what
4956 kind of type that is---integer, floating, pointer, structure, and so on.
4958 The file @file{typeclass.h} defines an enumeration that you can use to
4959 interpret the values of @code{__builtin_classify_type}.
4962 These machine description macros help implement varargs:
4964 @hook TARGET_EXPAND_BUILTIN_SAVEREGS
4965 If defined, this hook produces the machine-specific code for a call to
4966 @code{__builtin_saveregs}. This code will be moved to the very
4967 beginning of the function, before any parameter access are made. The
4968 return value of this function should be an RTX that contains the value
4969 to use as the return of @code{__builtin_saveregs}.
4972 @hook TARGET_SETUP_INCOMING_VARARGS
4973 This target hook offers an alternative to using
4974 @code{__builtin_saveregs} and defining the hook
4975 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4976 register arguments into the stack so that all the arguments appear to
4977 have been passed consecutively on the stack. Once this is done, you can
4978 use the standard implementation of varargs that works for machines that
4979 pass all their arguments on the stack.
4981 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4982 structure, containing the values that are obtained after processing the
4983 named arguments. The arguments @var{mode} and @var{type} describe the
4984 last named argument---its machine mode and its data type as a tree node.
4986 The target hook should do two things: first, push onto the stack all the
4987 argument registers @emph{not} used for the named arguments, and second,
4988 store the size of the data thus pushed into the @code{int}-valued
4989 variable pointed to by @var{pretend_args_size}. The value that you
4990 store here will serve as additional offset for setting up the stack
4993 Because you must generate code to push the anonymous arguments at
4994 compile time without knowing their data types,
4995 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4996 have just a single category of argument register and use it uniformly
4999 If the argument @var{second_time} is nonzero, it means that the
5000 arguments of the function are being analyzed for the second time. This
5001 happens for an inline function, which is not actually compiled until the
5002 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5003 not generate any instructions in this case.
5006 @hook TARGET_STRICT_ARGUMENT_NAMING
5007 Define this hook to return @code{true} if the location where a function
5008 argument is passed depends on whether or not it is a named argument.
5010 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5011 is set for varargs and stdarg functions. If this hook returns
5012 @code{true}, the @var{named} argument is always true for named
5013 arguments, and false for unnamed arguments. If it returns @code{false},
5014 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5015 then all arguments are treated as named. Otherwise, all named arguments
5016 except the last are treated as named.
5018 You need not define this hook if it always returns @code{false}.
5021 @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
5022 If you need to conditionally change ABIs so that one works with
5023 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5024 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5025 defined, then define this hook to return @code{true} if
5026 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5027 Otherwise, you should not define this hook.
5031 @section Trampolines for Nested Functions
5032 @cindex trampolines for nested functions
5033 @cindex nested functions, trampolines for
5035 A @dfn{trampoline} is a small piece of code that is created at run time
5036 when the address of a nested function is taken. It normally resides on
5037 the stack, in the stack frame of the containing function. These macros
5038 tell GCC how to generate code to allocate and initialize a
5041 The instructions in the trampoline must do two things: load a constant
5042 address into the static chain register, and jump to the real address of
5043 the nested function. On CISC machines such as the m68k, this requires
5044 two instructions, a move immediate and a jump. Then the two addresses
5045 exist in the trampoline as word-long immediate operands. On RISC
5046 machines, it is often necessary to load each address into a register in
5047 two parts. Then pieces of each address form separate immediate
5050 The code generated to initialize the trampoline must store the variable
5051 parts---the static chain value and the function address---into the
5052 immediate operands of the instructions. On a CISC machine, this is
5053 simply a matter of copying each address to a memory reference at the
5054 proper offset from the start of the trampoline. On a RISC machine, it
5055 may be necessary to take out pieces of the address and store them
5058 @hook TARGET_ASM_TRAMPOLINE_TEMPLATE
5059 This hook is called by @code{assemble_trampoline_template} to output,
5060 on the stream @var{f}, assembler code for a block of data that contains
5061 the constant parts of a trampoline. This code should not include a
5062 label---the label is taken care of automatically.
5064 If you do not define this hook, it means no template is needed
5065 for the target. Do not define this hook on systems where the block move
5066 code to copy the trampoline into place would be larger than the code
5067 to generate it on the spot.
5070 @defmac TRAMPOLINE_SECTION
5071 Return the section into which the trampoline template is to be placed
5072 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5075 @defmac TRAMPOLINE_SIZE
5076 A C expression for the size in bytes of the trampoline, as an integer.
5079 @defmac TRAMPOLINE_ALIGNMENT
5080 Alignment required for trampolines, in bits.
5082 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5083 is used for aligning trampolines.
5086 @hook TARGET_TRAMPOLINE_INIT
5087 This hook is called to initialize a trampoline.
5088 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5089 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5090 RTX for the static chain value that should be passed to the function
5093 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5094 first thing this hook should do is emit a block move into @var{m_tramp}
5095 from the memory block returned by @code{assemble_trampoline_template}.
5096 Note that the block move need only cover the constant parts of the
5097 trampoline. If the target isolates the variable parts of the trampoline
5098 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5100 If the target requires any other actions, such as flushing caches or
5101 enabling stack execution, these actions should be performed after
5102 initializing the trampoline proper.
5105 @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
5106 This hook should perform any machine-specific adjustment in
5107 the address of the trampoline. Its argument contains the address of the
5108 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5109 the address to be used for a function call should be different from the
5110 address at which the template was stored, the different address should
5111 be returned; otherwise @var{addr} should be returned unchanged.
5112 If this hook is not defined, @var{addr} will be used for function calls.
5115 Implementing trampolines is difficult on many machines because they have
5116 separate instruction and data caches. Writing into a stack location
5117 fails to clear the memory in the instruction cache, so when the program
5118 jumps to that location, it executes the old contents.
5120 Here are two possible solutions. One is to clear the relevant parts of
5121 the instruction cache whenever a trampoline is set up. The other is to
5122 make all trampolines identical, by having them jump to a standard
5123 subroutine. The former technique makes trampoline execution faster; the
5124 latter makes initialization faster.
5126 To clear the instruction cache when a trampoline is initialized, define
5127 the following macro.
5129 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5130 If defined, expands to a C expression clearing the @emph{instruction
5131 cache} in the specified interval. The definition of this macro would
5132 typically be a series of @code{asm} statements. Both @var{beg} and
5133 @var{end} are both pointer expressions.
5136 The operating system may also require the stack to be made executable
5137 before calling the trampoline. To implement this requirement, define
5138 the following macro.
5140 @defmac ENABLE_EXECUTE_STACK
5141 Define this macro if certain operations must be performed before executing
5142 code located on the stack. The macro should expand to a series of C
5143 file-scope constructs (e.g.@: functions) and provide a unique entry point
5144 named @code{__enable_execute_stack}. The target is responsible for
5145 emitting calls to the entry point in the code, for example from the
5146 @code{TARGET_TRAMPOLINE_INIT} hook.
5149 To use a standard subroutine, define the following macro. In addition,
5150 you must make sure that the instructions in a trampoline fill an entire
5151 cache line with identical instructions, or else ensure that the
5152 beginning of the trampoline code is always aligned at the same point in
5153 its cache line. Look in @file{m68k.h} as a guide.
5155 @defmac TRANSFER_FROM_TRAMPOLINE
5156 Define this macro if trampolines need a special subroutine to do their
5157 work. The macro should expand to a series of @code{asm} statements
5158 which will be compiled with GCC@. They go in a library function named
5159 @code{__transfer_from_trampoline}.
5161 If you need to avoid executing the ordinary prologue code of a compiled
5162 C function when you jump to the subroutine, you can do so by placing a
5163 special label of your own in the assembler code. Use one @code{asm}
5164 statement to generate an assembler label, and another to make the label
5165 global. Then trampolines can use that label to jump directly to your
5166 special assembler code.
5170 @section Implicit Calls to Library Routines
5171 @cindex library subroutine names
5172 @cindex @file{libgcc.a}
5174 @c prevent bad page break with this line
5175 Here is an explanation of implicit calls to library routines.
5177 @defmac DECLARE_LIBRARY_RENAMES
5178 This macro, if defined, should expand to a piece of C code that will get
5179 expanded when compiling functions for libgcc.a. It can be used to
5180 provide alternate names for GCC's internal library functions if there
5181 are ABI-mandated names that the compiler should provide.
5184 @findex set_optab_libfunc
5185 @findex init_one_libfunc
5186 @hook TARGET_INIT_LIBFUNCS
5187 This hook should declare additional library routines or rename
5188 existing ones, using the functions @code{set_optab_libfunc} and
5189 @code{init_one_libfunc} defined in @file{optabs.c}.
5190 @code{init_optabs} calls this macro after initializing all the normal
5193 The default is to do nothing. Most ports don't need to define this hook.
5196 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5197 This macro should return @code{true} if the library routine that
5198 implements the floating point comparison operator @var{comparison} in
5199 mode @var{mode} will return a boolean, and @var{false} if it will
5202 GCC's own floating point libraries return tristates from the
5203 comparison operators, so the default returns false always. Most ports
5204 don't need to define this macro.
5207 @defmac TARGET_LIB_INT_CMP_BIASED
5208 This macro should evaluate to @code{true} if the integer comparison
5209 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5210 operand is smaller than the second, 1 to indicate that they are equal,
5211 and 2 to indicate that the first operand is greater than the second.
5212 If this macro evaluates to @code{false} the comparison functions return
5213 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5214 in @file{libgcc.a}, you do not need to define this macro.
5217 @cindex @code{EDOM}, implicit usage
5220 The value of @code{EDOM} on the target machine, as a C integer constant
5221 expression. If you don't define this macro, GCC does not attempt to
5222 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5223 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5226 If you do not define @code{TARGET_EDOM}, then compiled code reports
5227 domain errors by calling the library function and letting it report the
5228 error. If mathematical functions on your system use @code{matherr} when
5229 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5230 that @code{matherr} is used normally.
5233 @cindex @code{errno}, implicit usage
5234 @defmac GEN_ERRNO_RTX
5235 Define this macro as a C expression to create an rtl expression that
5236 refers to the global ``variable'' @code{errno}. (On certain systems,
5237 @code{errno} may not actually be a variable.) If you don't define this
5238 macro, a reasonable default is used.
5241 @cindex C99 math functions, implicit usage
5242 @defmac TARGET_C99_FUNCTIONS
5243 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5244 @code{sinf} and similarly for other functions defined by C99 standard. The
5245 default is zero because a number of existing systems lack support for these
5246 functions in their runtime so this macro needs to be redefined to one on
5247 systems that do support the C99 runtime.
5250 @cindex sincos math function, implicit usage
5251 @defmac TARGET_HAS_SINCOS
5252 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5253 and @code{cos} with the same argument to a call to @code{sincos}. The
5254 default is zero. The target has to provide the following functions:
5256 void sincos(double x, double *sin, double *cos);
5257 void sincosf(float x, float *sin, float *cos);
5258 void sincosl(long double x, long double *sin, long double *cos);
5262 @defmac NEXT_OBJC_RUNTIME
5263 Define this macro to generate code for Objective-C message sending using
5264 the calling convention of the NeXT system. This calling convention
5265 involves passing the object, the selector and the method arguments all
5266 at once to the method-lookup library function.
5268 The default calling convention passes just the object and the selector
5269 to the lookup function, which returns a pointer to the method.
5272 @node Addressing Modes
5273 @section Addressing Modes
5274 @cindex addressing modes
5276 @c prevent bad page break with this line
5277 This is about addressing modes.
5279 @defmac HAVE_PRE_INCREMENT
5280 @defmacx HAVE_PRE_DECREMENT
5281 @defmacx HAVE_POST_INCREMENT
5282 @defmacx HAVE_POST_DECREMENT
5283 A C expression that is nonzero if the machine supports pre-increment,
5284 pre-decrement, post-increment, or post-decrement addressing respectively.
5287 @defmac HAVE_PRE_MODIFY_DISP
5288 @defmacx HAVE_POST_MODIFY_DISP
5289 A C expression that is nonzero if the machine supports pre- or
5290 post-address side-effect generation involving constants other than
5291 the size of the memory operand.
5294 @defmac HAVE_PRE_MODIFY_REG
5295 @defmacx HAVE_POST_MODIFY_REG
5296 A C expression that is nonzero if the machine supports pre- or
5297 post-address side-effect generation involving a register displacement.
5300 @defmac CONSTANT_ADDRESS_P (@var{x})
5301 A C expression that is 1 if the RTX @var{x} is a constant which
5302 is a valid address. On most machines the default definition of
5303 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5304 is acceptable, but a few machines are more restrictive as to which
5305 constant addresses are supported.
5308 @defmac CONSTANT_P (@var{x})
5309 @code{CONSTANT_P}, which is defined by target-independent code,
5310 accepts integer-values expressions whose values are not explicitly
5311 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5312 expressions and @code{const} arithmetic expressions, in addition to
5313 @code{const_int} and @code{const_double} expressions.
5316 @defmac MAX_REGS_PER_ADDRESS
5317 A number, the maximum number of registers that can appear in a valid
5318 memory address. Note that it is up to you to specify a value equal to
5319 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5323 @hook TARGET_LEGITIMATE_ADDRESS_P
5324 A function that returns whether @var{x} (an RTX) is a legitimate memory
5325 address on the target machine for a memory operand of mode @var{mode}.
5327 Legitimate addresses are defined in two variants: a strict variant and a
5328 non-strict one. The @var{strict} parameter chooses which variant is
5329 desired by the caller.
5331 The strict variant is used in the reload pass. It must be defined so
5332 that any pseudo-register that has not been allocated a hard register is
5333 considered a memory reference. This is because in contexts where some
5334 kind of register is required, a pseudo-register with no hard register
5335 must be rejected. For non-hard registers, the strict variant should look
5336 up the @code{reg_renumber} array; it should then proceed using the hard
5337 register number in the array, or treat the pseudo as a memory reference
5338 if the array holds @code{-1}.
5340 The non-strict variant is used in other passes. It must be defined to
5341 accept all pseudo-registers in every context where some kind of
5342 register is required.
5344 Normally, constant addresses which are the sum of a @code{symbol_ref}
5345 and an integer are stored inside a @code{const} RTX to mark them as
5346 constant. Therefore, there is no need to recognize such sums
5347 specifically as legitimate addresses. Normally you would simply
5348 recognize any @code{const} as legitimate.
5350 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5351 sums that are not marked with @code{const}. It assumes that a naked
5352 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5353 naked constant sums as illegitimate addresses, so that none of them will
5354 be given to @code{PRINT_OPERAND_ADDRESS}.
5356 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5357 On some machines, whether a symbolic address is legitimate depends on
5358 the section that the address refers to. On these machines, define the
5359 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5360 into the @code{symbol_ref}, and then check for it here. When you see a
5361 @code{const}, you will have to look inside it to find the
5362 @code{symbol_ref} in order to determine the section. @xref{Assembler
5365 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5366 Some ports are still using a deprecated legacy substitute for
5367 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5371 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5375 and should @code{goto @var{label}} if the address @var{x} is a valid
5376 address on the target machine for a memory operand of mode @var{mode}.
5378 @findex REG_OK_STRICT
5379 Compiler source files that want to use the strict variant of this
5380 macro define the macro @code{REG_OK_STRICT}. You should use an
5381 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5382 that case and the non-strict variant otherwise.
5384 Using the hook is usually simpler because it limits the number of
5385 files that are recompiled when changes are made.
5388 @defmac TARGET_MEM_CONSTRAINT
5389 A single character to be used instead of the default @code{'m'}
5390 character for general memory addresses. This defines the constraint
5391 letter which matches the memory addresses accepted by
5392 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5393 support new address formats in your back end without changing the
5394 semantics of the @code{'m'} constraint. This is necessary in order to
5395 preserve functionality of inline assembly constructs using the
5396 @code{'m'} constraint.
5399 @defmac FIND_BASE_TERM (@var{x})
5400 A C expression to determine the base term of address @var{x},
5401 or to provide a simplified version of @var{x} from which @file{alias.c}
5402 can easily find the base term. This macro is used in only two places:
5403 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5405 It is always safe for this macro to not be defined. It exists so
5406 that alias analysis can understand machine-dependent addresses.
5408 The typical use of this macro is to handle addresses containing
5409 a label_ref or symbol_ref within an UNSPEC@.
5412 @hook TARGET_LEGITIMIZE_ADDRESS
5413 This hook is given an invalid memory address @var{x} for an
5414 operand of mode @var{mode} and should try to return a valid memory
5417 @findex break_out_memory_refs
5418 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5419 and @var{oldx} will be the operand that was given to that function to produce
5422 The code of the hook should not alter the substructure of
5423 @var{x}. If it transforms @var{x} into a more legitimate form, it
5424 should return the new @var{x}.
5426 It is not necessary for this hook to come up with a legitimate address.
5427 The compiler has standard ways of doing so in all cases. In fact, it
5428 is safe to omit this hook or make it return @var{x} if it cannot find
5429 a valid way to legitimize the address. But often a machine-dependent
5430 strategy can generate better code.
5433 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5434 A C compound statement that attempts to replace @var{x}, which is an address
5435 that needs reloading, with a valid memory address for an operand of mode
5436 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5437 It is not necessary to define this macro, but it might be useful for
5438 performance reasons.
5440 For example, on the i386, it is sometimes possible to use a single
5441 reload register instead of two by reloading a sum of two pseudo
5442 registers into a register. On the other hand, for number of RISC
5443 processors offsets are limited so that often an intermediate address
5444 needs to be generated in order to address a stack slot. By defining
5445 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5446 generated for adjacent some stack slots can be made identical, and thus
5449 @emph{Note}: This macro should be used with caution. It is necessary
5450 to know something of how reload works in order to effectively use this,
5451 and it is quite easy to produce macros that build in too much knowledge
5452 of reload internals.
5454 @emph{Note}: This macro must be able to reload an address created by a
5455 previous invocation of this macro. If it fails to handle such addresses
5456 then the compiler may generate incorrect code or abort.
5459 The macro definition should use @code{push_reload} to indicate parts that
5460 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5461 suitable to be passed unaltered to @code{push_reload}.
5463 The code generated by this macro must not alter the substructure of
5464 @var{x}. If it transforms @var{x} into a more legitimate form, it
5465 should assign @var{x} (which will always be a C variable) a new value.
5466 This also applies to parts that you change indirectly by calling
5469 @findex strict_memory_address_p
5470 The macro definition may use @code{strict_memory_address_p} to test if
5471 the address has become legitimate.
5474 If you want to change only a part of @var{x}, one standard way of doing
5475 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5476 single level of rtl. Thus, if the part to be changed is not at the
5477 top level, you'll need to replace first the top level.
5478 It is not necessary for this macro to come up with a legitimate
5479 address; but often a machine-dependent strategy can generate better code.
5482 @hook TARGET_MODE_DEPENDENT_ADDRESS_P
5483 This hook returns @code{true} if memory address @var{addr} can have
5484 different meanings depending on the machine mode of the memory
5485 reference it is used for or if the address is valid for some modes
5488 Autoincrement and autodecrement addresses typically have mode-dependent
5489 effects because the amount of the increment or decrement is the size
5490 of the operand being addressed. Some machines have other mode-dependent
5491 addresses. Many RISC machines have no mode-dependent addresses.
5493 You may assume that @var{addr} is a valid address for the machine.
5495 The default version of this hook returns @code{false}.
5498 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5499 A C statement or compound statement with a conditional @code{goto
5500 @var{label};} executed if memory address @var{x} (an RTX) can have
5501 different meanings depending on the machine mode of the memory
5502 reference it is used for or if the address is valid for some modes
5505 Autoincrement and autodecrement addresses typically have mode-dependent
5506 effects because the amount of the increment or decrement is the size
5507 of the operand being addressed. Some machines have other mode-dependent
5508 addresses. Many RISC machines have no mode-dependent addresses.
5510 You may assume that @var{addr} is a valid address for the machine.
5512 These are obsolete macros, replaced by the
5513 @code{TARGET_MODE_DEPENDENT_ADDRESS_P} target hook.
5516 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5517 A C expression that is nonzero if @var{x} is a legitimate constant for
5518 an immediate operand on the target machine. You can assume that
5519 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5520 @samp{1} is a suitable definition for this macro on machines where
5521 anything @code{CONSTANT_P} is valid.
5524 @hook TARGET_DELEGITIMIZE_ADDRESS
5525 This hook is used to undo the possibly obfuscating effects of the
5526 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5527 macros. Some backend implementations of these macros wrap symbol
5528 references inside an @code{UNSPEC} rtx to represent PIC or similar
5529 addressing modes. This target hook allows GCC's optimizers to understand
5530 the semantics of these opaque @code{UNSPEC}s by converting them back
5531 into their original form.
5534 @hook TARGET_CANNOT_FORCE_CONST_MEM
5535 This hook should return true if @var{x} is of a form that cannot (or
5536 should not) be spilled to the constant pool. The default version of
5537 this hook returns false.
5539 The primary reason to define this hook is to prevent reload from
5540 deciding that a non-legitimate constant would be better reloaded
5541 from the constant pool instead of spilling and reloading a register
5542 holding the constant. This restriction is often true of addresses
5543 of TLS symbols for various targets.
5546 @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
5547 This hook should return true if pool entries for constant @var{x} can
5548 be placed in an @code{object_block} structure. @var{mode} is the mode
5551 The default version returns false for all constants.
5554 @hook TARGET_BUILTIN_RECIPROCAL
5555 This hook should return the DECL of a function that implements reciprocal of
5556 the builtin function with builtin function code @var{fn}, or
5557 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5558 when @var{fn} is a code of a machine-dependent builtin function. When
5559 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5560 of a square root function are performed, and only reciprocals of @code{sqrt}
5564 @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
5565 This hook should return the DECL of a function @var{f} that given an
5566 address @var{addr} as an argument returns a mask @var{m} that can be
5567 used to extract from two vectors the relevant data that resides in
5568 @var{addr} in case @var{addr} is not properly aligned.
5570 The autovectorizer, when vectorizing a load operation from an address
5571 @var{addr} that may be unaligned, will generate two vector loads from
5572 the two aligned addresses around @var{addr}. It then generates a
5573 @code{REALIGN_LOAD} operation to extract the relevant data from the
5574 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5575 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5576 the third argument, @var{OFF}, defines how the data will be extracted
5577 from these two vectors: if @var{OFF} is 0, then the returned vector is
5578 @var{v2}; otherwise, the returned vector is composed from the last
5579 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5580 @var{OFF} elements of @var{v2}.
5582 If this hook is defined, the autovectorizer will generate a call
5583 to @var{f} (using the DECL tree that this hook returns) and will
5584 use the return value of @var{f} as the argument @var{OFF} to
5585 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5586 should comply with the semantics expected by @code{REALIGN_LOAD}
5588 If this hook is not defined, then @var{addr} will be used as
5589 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5590 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5593 @hook TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN
5594 This hook should return the DECL of a function @var{f} that implements
5595 widening multiplication of the even elements of two input vectors of type @var{x}.
5597 If this hook is defined, the autovectorizer will use it along with the
5598 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5599 widening multiplication in cases that the order of the results does not have to be
5600 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5601 @code{widen_mult_hi/lo} idioms will be used.
5604 @hook TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD
5605 This hook should return the DECL of a function @var{f} that implements
5606 widening multiplication of the odd elements of two input vectors of type @var{x}.
5608 If this hook is defined, the autovectorizer will use it along with the
5609 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5610 widening multiplication in cases that the order of the results does not have to be
5611 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5612 @code{widen_mult_hi/lo} idioms will be used.
5615 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
5616 Returns cost of different scalar or vector statements for vectorization cost model.
5617 For vector memory operations the cost may depend on type (@var{vectype}) and
5618 misalignment value (@var{misalign}).
5621 @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
5622 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5625 @hook TARGET_VECTORIZE_BUILTIN_VEC_PERM
5626 Target builtin that implements vector permute.
5629 @hook TARGET_VECTORIZE_BUILTIN_VEC_PERM_OK
5630 Return true if a vector created for @code{builtin_vec_perm} is valid.
5633 @hook TARGET_VECTORIZE_BUILTIN_CONVERSION
5634 This hook should return the DECL of a function that implements conversion of the
5635 input vector of type @var{src_type} to type @var{dest_type}.
5636 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5637 specifies how the conversion is to be applied
5638 (truncation, rounding, etc.).
5640 If this hook is defined, the autovectorizer will use the
5641 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5642 conversion. Otherwise, it will return @code{NULL_TREE}.
5645 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
5646 This hook should return the decl of a function that implements the
5647 vectorized variant of the builtin function with builtin function code
5648 @var{code} or @code{NULL_TREE} if such a function is not available.
5649 The value of @var{fndecl} is the builtin function declaration. The
5650 return type of the vectorized function shall be of vector type
5651 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5654 @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
5655 This hook should return true if the target supports misaligned vector
5656 store/load of a specific factor denoted in the @var{misalignment}
5657 parameter. The vector store/load should be of machine mode @var{mode} and
5658 the elements in the vectors should be of type @var{type}. @var{is_packed}
5659 parameter is true if the memory access is defined in a packed struct.
5662 @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
5663 This hook should return the preferred mode for vectorizing scalar
5664 mode @var{mode}. The default is
5665 equal to @code{word_mode}, because the vectorizer can do some
5666 transformations even in absence of specialized @acronym{SIMD} hardware.
5669 @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES
5670 This hook should return a mask of sizes that should be iterated over
5671 after trying to autovectorize using the vector size derived from the
5672 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5673 The default is zero which means to not iterate over other vector sizes.
5676 @node Anchored Addresses
5677 @section Anchored Addresses
5678 @cindex anchored addresses
5679 @cindex @option{-fsection-anchors}
5681 GCC usually addresses every static object as a separate entity.
5682 For example, if we have:
5686 int foo (void) @{ return a + b + c; @}
5689 the code for @code{foo} will usually calculate three separate symbolic
5690 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5691 it would be better to calculate just one symbolic address and access
5692 the three variables relative to it. The equivalent pseudocode would
5698 register int *xr = &x;
5699 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5703 (which isn't valid C). We refer to shared addresses like @code{x} as
5704 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5706 The hooks below describe the target properties that GCC needs to know
5707 in order to make effective use of section anchors. It won't use
5708 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5709 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5711 @hook TARGET_MIN_ANCHOR_OFFSET
5712 The minimum offset that should be applied to a section anchor.
5713 On most targets, it should be the smallest offset that can be
5714 applied to a base register while still giving a legitimate address
5715 for every mode. The default value is 0.
5718 @hook TARGET_MAX_ANCHOR_OFFSET
5719 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5720 offset that should be applied to section anchors. The default
5724 @hook TARGET_ASM_OUTPUT_ANCHOR
5725 Write the assembly code to define section anchor @var{x}, which is a
5726 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5727 The hook is called with the assembly output position set to the beginning
5728 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5730 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5731 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5732 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5733 is @code{NULL}, which disables the use of section anchors altogether.
5736 @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
5737 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5738 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5739 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5741 The default version is correct for most targets, but you might need to
5742 intercept this hook to handle things like target-specific attributes
5743 or target-specific sections.
5746 @node Condition Code
5747 @section Condition Code Status
5748 @cindex condition code status
5750 The macros in this section can be split in two families, according to the
5751 two ways of representing condition codes in GCC.
5753 The first representation is the so called @code{(cc0)} representation
5754 (@pxref{Jump Patterns}), where all instructions can have an implicit
5755 clobber of the condition codes. The second is the condition code
5756 register representation, which provides better schedulability for
5757 architectures that do have a condition code register, but on which
5758 most instructions do not affect it. The latter category includes
5761 The implicit clobbering poses a strong restriction on the placement of
5762 the definition and use of the condition code, which need to be in adjacent
5763 insns for machines using @code{(cc0)}. This can prevent important
5764 optimizations on some machines. For example, on the IBM RS/6000, there
5765 is a delay for taken branches unless the condition code register is set
5766 three instructions earlier than the conditional branch. The instruction
5767 scheduler cannot perform this optimization if it is not permitted to
5768 separate the definition and use of the condition code register.
5770 For this reason, it is possible and suggested to use a register to
5771 represent the condition code for new ports. If there is a specific
5772 condition code register in the machine, use a hard register. If the
5773 condition code or comparison result can be placed in any general register,
5774 or if there are multiple condition registers, use a pseudo register.
5775 Registers used to store the condition code value will usually have a mode
5776 that is in class @code{MODE_CC}.
5778 Alternatively, you can use @code{BImode} if the comparison operator is
5779 specified already in the compare instruction. In this case, you are not
5780 interested in most macros in this section.
5783 * CC0 Condition Codes:: Old style representation of condition codes.
5784 * MODE_CC Condition Codes:: Modern representation of condition codes.
5785 * Cond Exec Macros:: Macros to control conditional execution.
5788 @node CC0 Condition Codes
5789 @subsection Representation of condition codes using @code{(cc0)}
5793 The file @file{conditions.h} defines a variable @code{cc_status} to
5794 describe how the condition code was computed (in case the interpretation of
5795 the condition code depends on the instruction that it was set by). This
5796 variable contains the RTL expressions on which the condition code is
5797 currently based, and several standard flags.
5799 Sometimes additional machine-specific flags must be defined in the machine
5800 description header file. It can also add additional machine-specific
5801 information by defining @code{CC_STATUS_MDEP}.
5803 @defmac CC_STATUS_MDEP
5804 C code for a data type which is used for declaring the @code{mdep}
5805 component of @code{cc_status}. It defaults to @code{int}.
5807 This macro is not used on machines that do not use @code{cc0}.
5810 @defmac CC_STATUS_MDEP_INIT
5811 A C expression to initialize the @code{mdep} field to ``empty''.
5812 The default definition does nothing, since most machines don't use
5813 the field anyway. If you want to use the field, you should probably
5814 define this macro to initialize it.
5816 This macro is not used on machines that do not use @code{cc0}.
5819 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5820 A C compound statement to set the components of @code{cc_status}
5821 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5822 this macro's responsibility to recognize insns that set the condition
5823 code as a byproduct of other activity as well as those that explicitly
5826 This macro is not used on machines that do not use @code{cc0}.
5828 If there are insns that do not set the condition code but do alter
5829 other machine registers, this macro must check to see whether they
5830 invalidate the expressions that the condition code is recorded as
5831 reflecting. For example, on the 68000, insns that store in address
5832 registers do not set the condition code, which means that usually
5833 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5834 insns. But suppose that the previous insn set the condition code
5835 based on location @samp{a4@@(102)} and the current insn stores a new
5836 value in @samp{a4}. Although the condition code is not changed by
5837 this, it will no longer be true that it reflects the contents of
5838 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5839 @code{cc_status} in this case to say that nothing is known about the
5840 condition code value.
5842 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5843 with the results of peephole optimization: insns whose patterns are
5844 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5845 constants which are just the operands. The RTL structure of these
5846 insns is not sufficient to indicate what the insns actually do. What
5847 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5848 @code{CC_STATUS_INIT}.
5850 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5851 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5852 @samp{cc}. This avoids having detailed information about patterns in
5853 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5856 @node MODE_CC Condition Codes
5857 @subsection Representation of condition codes using registers
5861 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5862 On many machines, the condition code may be produced by other instructions
5863 than compares, for example the branch can use directly the condition
5864 code set by a subtract instruction. However, on some machines
5865 when the condition code is set this way some bits (such as the overflow
5866 bit) are not set in the same way as a test instruction, so that a different
5867 branch instruction must be used for some conditional branches. When
5868 this happens, use the machine mode of the condition code register to
5869 record different formats of the condition code register. Modes can
5870 also be used to record which compare instruction (e.g. a signed or an
5871 unsigned comparison) produced the condition codes.
5873 If other modes than @code{CCmode} are required, add them to
5874 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
5875 a mode given an operand of a compare. This is needed because the modes
5876 have to be chosen not only during RTL generation but also, for example,
5877 by instruction combination. The result of @code{SELECT_CC_MODE} should
5878 be consistent with the mode used in the patterns; for example to support
5879 the case of the add on the SPARC discussed above, we have the pattern
5883 [(set (reg:CC_NOOV 0)
5885 (plus:SI (match_operand:SI 0 "register_operand" "%r")
5886 (match_operand:SI 1 "arith_operand" "rI"))
5893 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
5894 for comparisons whose argument is a @code{plus}:
5897 #define SELECT_CC_MODE(OP,X,Y) \
5898 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5899 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5900 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5901 || GET_CODE (X) == NEG) \
5902 ? CC_NOOVmode : CCmode))
5905 Another reason to use modes is to retain information on which operands
5906 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
5909 You should define this macro if and only if you define extra CC modes
5910 in @file{@var{machine}-modes.def}.
5913 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5914 On some machines not all possible comparisons are defined, but you can
5915 convert an invalid comparison into a valid one. For example, the Alpha
5916 does not have a @code{GT} comparison, but you can use an @code{LT}
5917 comparison instead and swap the order of the operands.
5919 On such machines, define this macro to be a C statement to do any
5920 required conversions. @var{code} is the initial comparison code
5921 and @var{op0} and @var{op1} are the left and right operands of the
5922 comparison, respectively. You should modify @var{code}, @var{op0}, and
5923 @var{op1} as required.
5925 GCC will not assume that the comparison resulting from this macro is
5926 valid but will see if the resulting insn matches a pattern in the
5929 You need not define this macro if it would never change the comparison
5933 @defmac REVERSIBLE_CC_MODE (@var{mode})
5934 A C expression whose value is one if it is always safe to reverse a
5935 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5936 can ever return @var{mode} for a floating-point inequality comparison,
5937 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5939 You need not define this macro if it would always returns zero or if the
5940 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5941 For example, here is the definition used on the SPARC, where floating-point
5942 inequality comparisons are always given @code{CCFPEmode}:
5945 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5949 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5950 A C expression whose value is reversed condition code of the @var{code} for
5951 comparison done in CC_MODE @var{mode}. The macro is used only in case
5952 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5953 machine has some non-standard way how to reverse certain conditionals. For
5954 instance in case all floating point conditions are non-trapping, compiler may
5955 freely convert unordered compares to ordered one. Then definition may look
5959 #define REVERSE_CONDITION(CODE, MODE) \
5960 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5961 : reverse_condition_maybe_unordered (CODE))
5965 @hook TARGET_FIXED_CONDITION_CODE_REGS
5966 On targets which do not use @code{(cc0)}, and which use a hard
5967 register rather than a pseudo-register to hold condition codes, the
5968 regular CSE passes are often not able to identify cases in which the
5969 hard register is set to a common value. Use this hook to enable a
5970 small pass which optimizes such cases. This hook should return true
5971 to enable this pass, and it should set the integers to which its
5972 arguments point to the hard register numbers used for condition codes.
5973 When there is only one such register, as is true on most systems, the
5974 integer pointed to by @var{p2} should be set to
5975 @code{INVALID_REGNUM}.
5977 The default version of this hook returns false.
5980 @hook TARGET_CC_MODES_COMPATIBLE
5981 On targets which use multiple condition code modes in class
5982 @code{MODE_CC}, it is sometimes the case that a comparison can be
5983 validly done in more than one mode. On such a system, define this
5984 target hook to take two mode arguments and to return a mode in which
5985 both comparisons may be validly done. If there is no such mode,
5986 return @code{VOIDmode}.
5988 The default version of this hook checks whether the modes are the
5989 same. If they are, it returns that mode. If they are different, it
5990 returns @code{VOIDmode}.
5993 @node Cond Exec Macros
5994 @subsection Macros to control conditional execution
5995 @findex conditional execution
5998 There is one macro that may need to be defined for targets
5999 supporting conditional execution, independent of how they
6000 represent conditional branches.
6002 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6003 A C expression that returns true if the conditional execution predicate
6004 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6005 versa. Define this to return 0 if the target has conditional execution
6006 predicates that cannot be reversed safely. There is no need to validate
6007 that the arguments of op1 and op2 are the same, this is done separately.
6008 If no expansion is specified, this macro is defined as follows:
6011 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6012 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6017 @section Describing Relative Costs of Operations
6018 @cindex costs of instructions
6019 @cindex relative costs
6020 @cindex speed of instructions
6022 These macros let you describe the relative speed of various operations
6023 on the target machine.
6025 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6026 A C expression for the cost of moving data of mode @var{mode} from a
6027 register in class @var{from} to one in class @var{to}. The classes are
6028 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6029 value of 2 is the default; other values are interpreted relative to
6032 It is not required that the cost always equal 2 when @var{from} is the
6033 same as @var{to}; on some machines it is expensive to move between
6034 registers if they are not general registers.
6036 If reload sees an insn consisting of a single @code{set} between two
6037 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6038 classes returns a value of 2, reload does not check to ensure that the
6039 constraints of the insn are met. Setting a cost of other than 2 will
6040 allow reload to verify that the constraints are met. You should do this
6041 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6043 These macros are obsolete, new ports should use the target hook
6044 @code{TARGET_REGISTER_MOVE_COST} instead.
6047 @hook TARGET_REGISTER_MOVE_COST
6048 This target hook should return the cost of moving data of mode @var{mode}
6049 from a register in class @var{from} to one in class @var{to}. The classes
6050 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6051 A value of 2 is the default; other values are interpreted relative to
6054 It is not required that the cost always equal 2 when @var{from} is the
6055 same as @var{to}; on some machines it is expensive to move between
6056 registers if they are not general registers.
6058 If reload sees an insn consisting of a single @code{set} between two
6059 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6060 classes returns a value of 2, reload does not check to ensure that the
6061 constraints of the insn are met. Setting a cost of other than 2 will
6062 allow reload to verify that the constraints are met. You should do this
6063 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6065 The default version of this function returns 2.
6068 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6069 A C expression for the cost of moving data of mode @var{mode} between a
6070 register of class @var{class} and memory; @var{in} is zero if the value
6071 is to be written to memory, nonzero if it is to be read in. This cost
6072 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6073 registers and memory is more expensive than between two registers, you
6074 should define this macro to express the relative cost.
6076 If you do not define this macro, GCC uses a default cost of 4 plus
6077 the cost of copying via a secondary reload register, if one is
6078 needed. If your machine requires a secondary reload register to copy
6079 between memory and a register of @var{class} but the reload mechanism is
6080 more complex than copying via an intermediate, define this macro to
6081 reflect the actual cost of the move.
6083 GCC defines the function @code{memory_move_secondary_cost} if
6084 secondary reloads are needed. It computes the costs due to copying via
6085 a secondary register. If your machine copies from memory using a
6086 secondary register in the conventional way but the default base value of
6087 4 is not correct for your machine, define this macro to add some other
6088 value to the result of that function. The arguments to that function
6089 are the same as to this macro.
6091 These macros are obsolete, new ports should use the target hook
6092 @code{TARGET_MEMORY_MOVE_COST} instead.
6095 @hook TARGET_MEMORY_MOVE_COST
6096 This target hook should return the cost of moving data of mode @var{mode}
6097 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6098 if the value is to be written to memory, @code{true} if it is to be read in.
6099 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6100 If moving between registers and memory is more expensive than between two
6101 registers, you should add this target hook to express the relative cost.
6103 If you do not add this target hook, GCC uses a default cost of 4 plus
6104 the cost of copying via a secondary reload register, if one is
6105 needed. If your machine requires a secondary reload register to copy
6106 between memory and a register of @var{rclass} but the reload mechanism is
6107 more complex than copying via an intermediate, use this target hook to
6108 reflect the actual cost of the move.
6110 GCC defines the function @code{memory_move_secondary_cost} if
6111 secondary reloads are needed. It computes the costs due to copying via
6112 a secondary register. If your machine copies from memory using a
6113 secondary register in the conventional way but the default base value of
6114 4 is not correct for your machine, use this target hook to add some other
6115 value to the result of that function. The arguments to that function
6116 are the same as to this target hook.
6119 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6120 A C expression for the cost of a branch instruction. A value of 1 is
6121 the default; other values are interpreted relative to that. Parameter
6122 @var{speed_p} is true when the branch in question should be optimized
6123 for speed. When it is false, @code{BRANCH_COST} should return a value
6124 optimal for code size rather than performance. @var{predictable_p} is
6125 true for well-predicted branches. On many architectures the
6126 @code{BRANCH_COST} can be reduced then.
6129 Here are additional macros which do not specify precise relative costs,
6130 but only that certain actions are more expensive than GCC would
6133 @defmac SLOW_BYTE_ACCESS
6134 Define this macro as a C expression which is nonzero if accessing less
6135 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6136 faster than accessing a word of memory, i.e., if such access
6137 require more than one instruction or if there is no difference in cost
6138 between byte and (aligned) word loads.
6140 When this macro is not defined, the compiler will access a field by
6141 finding the smallest containing object; when it is defined, a fullword
6142 load will be used if alignment permits. Unless bytes accesses are
6143 faster than word accesses, using word accesses is preferable since it
6144 may eliminate subsequent memory access if subsequent accesses occur to
6145 other fields in the same word of the structure, but to different bytes.
6148 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6149 Define this macro to be the value 1 if memory accesses described by the
6150 @var{mode} and @var{alignment} parameters have a cost many times greater
6151 than aligned accesses, for example if they are emulated in a trap
6154 When this macro is nonzero, the compiler will act as if
6155 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6156 moves. This can cause significantly more instructions to be produced.
6157 Therefore, do not set this macro nonzero if unaligned accesses only add a
6158 cycle or two to the time for a memory access.
6160 If the value of this macro is always zero, it need not be defined. If
6161 this macro is defined, it should produce a nonzero value when
6162 @code{STRICT_ALIGNMENT} is nonzero.
6165 @defmac MOVE_RATIO (@var{speed})
6166 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6167 which a sequence of insns should be generated instead of a
6168 string move insn or a library call. Increasing the value will always
6169 make code faster, but eventually incurs high cost in increased code size.
6171 Note that on machines where the corresponding move insn is a
6172 @code{define_expand} that emits a sequence of insns, this macro counts
6173 the number of such sequences.
6175 The parameter @var{speed} is true if the code is currently being
6176 optimized for speed rather than size.
6178 If you don't define this, a reasonable default is used.
6181 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6182 A C expression used to determine whether @code{move_by_pieces} will be used to
6183 copy a chunk of memory, or whether some other block move mechanism
6184 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6185 than @code{MOVE_RATIO}.
6188 @defmac MOVE_MAX_PIECES
6189 A C expression used by @code{move_by_pieces} to determine the largest unit
6190 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6193 @defmac CLEAR_RATIO (@var{speed})
6194 The threshold of number of scalar move insns, @emph{below} which a sequence
6195 of insns should be generated to clear memory instead of a string clear insn
6196 or a library call. Increasing the value will always make code faster, but
6197 eventually incurs high cost in increased code size.
6199 The parameter @var{speed} is true if the code is currently being
6200 optimized for speed rather than size.
6202 If you don't define this, a reasonable default is used.
6205 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6206 A C expression used to determine whether @code{clear_by_pieces} will be used
6207 to clear a chunk of memory, or whether some other block clear mechanism
6208 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6209 than @code{CLEAR_RATIO}.
6212 @defmac SET_RATIO (@var{speed})
6213 The threshold of number of scalar move insns, @emph{below} which a sequence
6214 of insns should be generated to set memory to a constant value, instead of
6215 a block set insn or a library call.
6216 Increasing the value will always make code faster, but
6217 eventually incurs high cost in increased code size.
6219 The parameter @var{speed} is true if the code is currently being
6220 optimized for speed rather than size.
6222 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6225 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6226 A C expression used to determine whether @code{store_by_pieces} will be
6227 used to set a chunk of memory to a constant value, or whether some
6228 other mechanism will be used. Used by @code{__builtin_memset} when
6229 storing values other than constant zero.
6230 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6231 than @code{SET_RATIO}.
6234 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6235 A C expression used to determine whether @code{store_by_pieces} will be
6236 used to set a chunk of memory to a constant string value, or whether some
6237 other mechanism will be used. Used by @code{__builtin_strcpy} when
6238 called with a constant source string.
6239 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6240 than @code{MOVE_RATIO}.
6243 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6244 A C expression used to determine whether a load postincrement is a good
6245 thing to use for a given mode. Defaults to the value of
6246 @code{HAVE_POST_INCREMENT}.
6249 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6250 A C expression used to determine whether a load postdecrement is a good
6251 thing to use for a given mode. Defaults to the value of
6252 @code{HAVE_POST_DECREMENT}.
6255 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6256 A C expression used to determine whether a load preincrement is a good
6257 thing to use for a given mode. Defaults to the value of
6258 @code{HAVE_PRE_INCREMENT}.
6261 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6262 A C expression used to determine whether a load predecrement is a good
6263 thing to use for a given mode. Defaults to the value of
6264 @code{HAVE_PRE_DECREMENT}.
6267 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6268 A C expression used to determine whether a store postincrement is a good
6269 thing to use for a given mode. Defaults to the value of
6270 @code{HAVE_POST_INCREMENT}.
6273 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6274 A C expression used to determine whether a store postdecrement is a good
6275 thing to use for a given mode. Defaults to the value of
6276 @code{HAVE_POST_DECREMENT}.
6279 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6280 This macro is used to determine whether a store preincrement is a good
6281 thing to use for a given mode. Defaults to the value of
6282 @code{HAVE_PRE_INCREMENT}.
6285 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6286 This macro is used to determine whether a store predecrement is a good
6287 thing to use for a given mode. Defaults to the value of
6288 @code{HAVE_PRE_DECREMENT}.
6291 @defmac NO_FUNCTION_CSE
6292 Define this macro if it is as good or better to call a constant
6293 function address than to call an address kept in a register.
6296 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6297 Define this macro if a non-short-circuit operation produced by
6298 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6299 @code{BRANCH_COST} is greater than or equal to the value 2.
6302 @hook TARGET_RTX_COSTS
6303 This target hook describes the relative costs of RTL expressions.
6305 The cost may depend on the precise form of the expression, which is
6306 available for examination in @var{x}, and the rtx code of the expression
6307 in which it is contained, found in @var{outer_code}. @var{code} is the
6308 expression code---redundant, since it can be obtained with
6309 @code{GET_CODE (@var{x})}.
6311 In implementing this hook, you can use the construct
6312 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6315 On entry to the hook, @code{*@var{total}} contains a default estimate
6316 for the cost of the expression. The hook should modify this value as
6317 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6318 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6319 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6321 When optimizing for code size, i.e.@: when @code{speed} is
6322 false, this target hook should be used to estimate the relative
6323 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6325 The hook returns true when all subexpressions of @var{x} have been
6326 processed, and false when @code{rtx_cost} should recurse.
6329 @hook TARGET_ADDRESS_COST
6330 This hook computes the cost of an addressing mode that contains
6331 @var{address}. If not defined, the cost is computed from
6332 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6334 For most CISC machines, the default cost is a good approximation of the
6335 true cost of the addressing mode. However, on RISC machines, all
6336 instructions normally have the same length and execution time. Hence
6337 all addresses will have equal costs.
6339 In cases where more than one form of an address is known, the form with
6340 the lowest cost will be used. If multiple forms have the same, lowest,
6341 cost, the one that is the most complex will be used.
6343 For example, suppose an address that is equal to the sum of a register
6344 and a constant is used twice in the same basic block. When this macro
6345 is not defined, the address will be computed in a register and memory
6346 references will be indirect through that register. On machines where
6347 the cost of the addressing mode containing the sum is no higher than
6348 that of a simple indirect reference, this will produce an additional
6349 instruction and possibly require an additional register. Proper
6350 specification of this macro eliminates this overhead for such machines.
6352 This hook is never called with an invalid address.
6354 On machines where an address involving more than one register is as
6355 cheap as an address computation involving only one register, defining
6356 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6357 be live over a region of code where only one would have been if
6358 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6359 should be considered in the definition of this macro. Equivalent costs
6360 should probably only be given to addresses with different numbers of
6361 registers on machines with lots of registers.
6365 @section Adjusting the Instruction Scheduler
6367 The instruction scheduler may need a fair amount of machine-specific
6368 adjustment in order to produce good code. GCC provides several target
6369 hooks for this purpose. It is usually enough to define just a few of
6370 them: try the first ones in this list first.
6372 @hook TARGET_SCHED_ISSUE_RATE
6373 This hook returns the maximum number of instructions that can ever
6374 issue at the same time on the target machine. The default is one.
6375 Although the insn scheduler can define itself the possibility of issue
6376 an insn on the same cycle, the value can serve as an additional
6377 constraint to issue insns on the same simulated processor cycle (see
6378 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6379 This value must be constant over the entire compilation. If you need
6380 it to vary depending on what the instructions are, you must use
6381 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6384 @hook TARGET_SCHED_VARIABLE_ISSUE
6385 This hook is executed by the scheduler after it has scheduled an insn
6386 from the ready list. It should return the number of insns which can
6387 still be issued in the current cycle. The default is
6388 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6389 @code{USE}, which normally are not counted against the issue rate.
6390 You should define this hook if some insns take more machine resources
6391 than others, so that fewer insns can follow them in the same cycle.
6392 @var{file} is either a null pointer, or a stdio stream to write any
6393 debug output to. @var{verbose} is the verbose level provided by
6394 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6398 @hook TARGET_SCHED_ADJUST_COST
6399 This function corrects the value of @var{cost} based on the
6400 relationship between @var{insn} and @var{dep_insn} through the
6401 dependence @var{link}. It should return the new value. The default
6402 is to make no adjustment to @var{cost}. This can be used for example
6403 to specify to the scheduler using the traditional pipeline description
6404 that an output- or anti-dependence does not incur the same cost as a
6405 data-dependence. If the scheduler using the automaton based pipeline
6406 description, the cost of anti-dependence is zero and the cost of
6407 output-dependence is maximum of one and the difference of latency
6408 times of the first and the second insns. If these values are not
6409 acceptable, you could use the hook to modify them too. See also
6410 @pxref{Processor pipeline description}.
6413 @hook TARGET_SCHED_ADJUST_PRIORITY
6414 This hook adjusts the integer scheduling priority @var{priority} of
6415 @var{insn}. It should return the new priority. Increase the priority to
6416 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6417 later. Do not define this hook if you do not need to adjust the
6418 scheduling priorities of insns.
6421 @hook TARGET_SCHED_REORDER
6422 This hook is executed by the scheduler after it has scheduled the ready
6423 list, to allow the machine description to reorder it (for example to
6424 combine two small instructions together on @samp{VLIW} machines).
6425 @var{file} is either a null pointer, or a stdio stream to write any
6426 debug output to. @var{verbose} is the verbose level provided by
6427 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6428 list of instructions that are ready to be scheduled. @var{n_readyp} is
6429 a pointer to the number of elements in the ready list. The scheduler
6430 reads the ready list in reverse order, starting with
6431 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6432 is the timer tick of the scheduler. You may modify the ready list and
6433 the number of ready insns. The return value is the number of insns that
6434 can issue this cycle; normally this is just @code{issue_rate}. See also
6435 @samp{TARGET_SCHED_REORDER2}.
6438 @hook TARGET_SCHED_REORDER2
6439 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6440 function is called whenever the scheduler starts a new cycle. This one
6441 is called once per iteration over a cycle, immediately after
6442 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6443 return the number of insns to be scheduled in the same cycle. Defining
6444 this hook can be useful if there are frequent situations where
6445 scheduling one insn causes other insns to become ready in the same
6446 cycle. These other insns can then be taken into account properly.
6449 @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
6450 This hook is called after evaluation forward dependencies of insns in
6451 chain given by two parameter values (@var{head} and @var{tail}
6452 correspondingly) but before insns scheduling of the insn chain. For
6453 example, it can be used for better insn classification if it requires
6454 analysis of dependencies. This hook can use backward and forward
6455 dependencies of the insn scheduler because they are already
6459 @hook TARGET_SCHED_INIT
6460 This hook is executed by the scheduler at the beginning of each block of
6461 instructions that are to be scheduled. @var{file} is either a null
6462 pointer, or a stdio stream to write any debug output to. @var{verbose}
6463 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6464 @var{max_ready} is the maximum number of insns in the current scheduling
6465 region that can be live at the same time. This can be used to allocate
6466 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6469 @hook TARGET_SCHED_FINISH
6470 This hook is executed by the scheduler at the end of each block of
6471 instructions that are to be scheduled. It can be used to perform
6472 cleanup of any actions done by the other scheduling hooks. @var{file}
6473 is either a null pointer, or a stdio stream to write any debug output
6474 to. @var{verbose} is the verbose level provided by
6475 @option{-fsched-verbose-@var{n}}.
6478 @hook TARGET_SCHED_INIT_GLOBAL
6479 This hook is executed by the scheduler after function level initializations.
6480 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6481 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6482 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6485 @hook TARGET_SCHED_FINISH_GLOBAL
6486 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6487 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6488 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6491 @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
6492 The hook returns an RTL insn. The automaton state used in the
6493 pipeline hazard recognizer is changed as if the insn were scheduled
6494 when the new simulated processor cycle starts. Usage of the hook may
6495 simplify the automaton pipeline description for some @acronym{VLIW}
6496 processors. If the hook is defined, it is used only for the automaton
6497 based pipeline description. The default is not to change the state
6498 when the new simulated processor cycle starts.
6501 @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
6502 The hook can be used to initialize data used by the previous hook.
6505 @hook TARGET_SCHED_DFA_POST_CYCLE_INSN
6506 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6507 to changed the state as if the insn were scheduled when the new
6508 simulated processor cycle finishes.
6511 @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
6512 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6513 used to initialize data used by the previous hook.
6516 @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
6517 The hook to notify target that the current simulated cycle is about to finish.
6518 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6519 to change the state in more complicated situations - e.g., when advancing
6520 state on a single insn is not enough.
6523 @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
6524 The hook to notify target that new simulated cycle has just started.
6525 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6526 to change the state in more complicated situations - e.g., when advancing
6527 state on a single insn is not enough.
6530 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
6531 This hook controls better choosing an insn from the ready insn queue
6532 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6533 chooses the first insn from the queue. If the hook returns a positive
6534 value, an additional scheduler code tries all permutations of
6535 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6536 subsequent ready insns to choose an insn whose issue will result in
6537 maximal number of issued insns on the same cycle. For the
6538 @acronym{VLIW} processor, the code could actually solve the problem of
6539 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6540 rules of @acronym{VLIW} packing are described in the automaton.
6542 This code also could be used for superscalar @acronym{RISC}
6543 processors. Let us consider a superscalar @acronym{RISC} processor
6544 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6545 @var{B}, some insns can be executed only in pipelines @var{B} or
6546 @var{C}, and one insn can be executed in pipeline @var{B}. The
6547 processor may issue the 1st insn into @var{A} and the 2nd one into
6548 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6549 until the next cycle. If the scheduler issues the 3rd insn the first,
6550 the processor could issue all 3 insns per cycle.
6552 Actually this code demonstrates advantages of the automaton based
6553 pipeline hazard recognizer. We try quickly and easy many insn
6554 schedules to choose the best one.
6556 The default is no multipass scheduling.
6559 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
6561 This hook controls what insns from the ready insn queue will be
6562 considered for the multipass insn scheduling. If the hook returns
6563 zero for @var{insn}, the insn will be not chosen to
6566 The default is that any ready insns can be chosen to be issued.
6569 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
6570 This hook prepares the target backend for a new round of multipass
6574 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
6575 This hook is called when multipass scheduling evaluates instruction INSN.
6578 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
6579 This is called when multipass scheduling backtracks from evaluation of
6583 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
6584 This hook notifies the target about the result of the concluded current
6585 round of multipass scheduling.
6588 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
6589 This hook initializes target-specific data used in multipass scheduling.
6592 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
6593 This hook finalizes target-specific data used in multipass scheduling.
6596 @hook TARGET_SCHED_DFA_NEW_CYCLE
6597 This hook is called by the insn scheduler before issuing @var{insn}
6598 on cycle @var{clock}. If the hook returns nonzero,
6599 @var{insn} is not issued on this processor cycle. Instead,
6600 the processor cycle is advanced. If *@var{sort_p}
6601 is zero, the insn ready queue is not sorted on the new cycle
6602 start as usually. @var{dump} and @var{verbose} specify the file and
6603 verbosity level to use for debugging output.
6604 @var{last_clock} and @var{clock} are, respectively, the
6605 processor cycle on which the previous insn has been issued,
6606 and the current processor cycle.
6609 @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
6610 This hook is used to define which dependences are considered costly by
6611 the target, so costly that it is not advisable to schedule the insns that
6612 are involved in the dependence too close to one another. The parameters
6613 to this hook are as follows: The first parameter @var{_dep} is the dependence
6614 being evaluated. The second parameter @var{cost} is the cost of the
6615 dependence as estimated by the scheduler, and the third
6616 parameter @var{distance} is the distance in cycles between the two insns.
6617 The hook returns @code{true} if considering the distance between the two
6618 insns the dependence between them is considered costly by the target,
6619 and @code{false} otherwise.
6621 Defining this hook can be useful in multiple-issue out-of-order machines,
6622 where (a) it's practically hopeless to predict the actual data/resource
6623 delays, however: (b) there's a better chance to predict the actual grouping
6624 that will be formed, and (c) correctly emulating the grouping can be very
6625 important. In such targets one may want to allow issuing dependent insns
6626 closer to one another---i.e., closer than the dependence distance; however,
6627 not in cases of ``costly dependences'', which this hooks allows to define.
6630 @hook TARGET_SCHED_H_I_D_EXTENDED
6631 This hook is called by the insn scheduler after emitting a new instruction to
6632 the instruction stream. The hook notifies a target backend to extend its
6633 per instruction data structures.
6636 @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
6637 Return a pointer to a store large enough to hold target scheduling context.
6640 @hook TARGET_SCHED_INIT_SCHED_CONTEXT
6641 Initialize store pointed to by @var{tc} to hold target scheduling context.
6642 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6643 beginning of the block. Otherwise, copy the current context into @var{tc}.
6646 @hook TARGET_SCHED_SET_SCHED_CONTEXT
6647 Copy target scheduling context pointed to by @var{tc} to the current context.
6650 @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
6651 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6654 @hook TARGET_SCHED_FREE_SCHED_CONTEXT
6655 Deallocate a store for target scheduling context pointed to by @var{tc}.
6658 @hook TARGET_SCHED_SPECULATE_INSN
6659 This hook is called by the insn scheduler when @var{insn} has only
6660 speculative dependencies and therefore can be scheduled speculatively.
6661 The hook is used to check if the pattern of @var{insn} has a speculative
6662 version and, in case of successful check, to generate that speculative
6663 pattern. The hook should return 1, if the instruction has a speculative form,
6664 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6665 speculation. If the return value equals 1 then @var{new_pat} is assigned
6666 the generated speculative pattern.
6669 @hook TARGET_SCHED_NEEDS_BLOCK_P
6670 This hook is called by the insn scheduler during generation of recovery code
6671 for @var{insn}. It should return @code{true}, if the corresponding check
6672 instruction should branch to recovery code, or @code{false} otherwise.
6675 @hook TARGET_SCHED_GEN_SPEC_CHECK
6676 This hook is called by the insn scheduler to generate a pattern for recovery
6677 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6678 speculative instruction for which the check should be generated.
6679 @var{label} is either a label of a basic block, where recovery code should
6680 be emitted, or a null pointer, when requested check doesn't branch to
6681 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6682 a pattern for a branchy check corresponding to a simple check denoted by
6683 @var{insn} should be generated. In this case @var{label} can't be null.
6686 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC
6687 This hook is used as a workaround for
6688 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6689 called on the first instruction of the ready list. The hook is used to
6690 discard speculative instructions that stand first in the ready list from
6691 being scheduled on the current cycle. If the hook returns @code{false},
6692 @var{insn} will not be chosen to be issued.
6693 For non-speculative instructions,
6694 the hook should always return @code{true}. For example, in the ia64 backend
6695 the hook is used to cancel data speculative insns when the ALAT table
6699 @hook TARGET_SCHED_SET_SCHED_FLAGS
6700 This hook is used by the insn scheduler to find out what features should be
6702 The structure *@var{spec_info} should be filled in by the target.
6703 The structure describes speculation types that can be used in the scheduler.
6706 @hook TARGET_SCHED_SMS_RES_MII
6707 This hook is called by the swing modulo scheduler to calculate a
6708 resource-based lower bound which is based on the resources available in
6709 the machine and the resources required by each instruction. The target
6710 backend can use @var{g} to calculate such bound. A very simple lower
6711 bound will be used in case this hook is not implemented: the total number
6712 of instructions divided by the issue rate.
6715 @hook TARGET_SCHED_DISPATCH
6716 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6717 is supported in hardware and the condition specified in the parameter is true.
6720 @hook TARGET_SCHED_DISPATCH_DO
6721 This hook is called by Haifa Scheduler. It performs the operation specified
6722 in its second parameter.
6726 @section Dividing the Output into Sections (Texts, Data, @dots{})
6727 @c the above section title is WAY too long. maybe cut the part between
6728 @c the (...)? --mew 10feb93
6730 An object file is divided into sections containing different types of
6731 data. In the most common case, there are three sections: the @dfn{text
6732 section}, which holds instructions and read-only data; the @dfn{data
6733 section}, which holds initialized writable data; and the @dfn{bss
6734 section}, which holds uninitialized data. Some systems have other kinds
6737 @file{varasm.c} provides several well-known sections, such as
6738 @code{text_section}, @code{data_section} and @code{bss_section}.
6739 The normal way of controlling a @code{@var{foo}_section} variable
6740 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6741 as described below. The macros are only read once, when @file{varasm.c}
6742 initializes itself, so their values must be run-time constants.
6743 They may however depend on command-line flags.
6745 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6746 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6747 to be string literals.
6749 Some assemblers require a different string to be written every time a
6750 section is selected. If your assembler falls into this category, you
6751 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6752 @code{get_unnamed_section} to set up the sections.
6754 You must always create a @code{text_section}, either by defining
6755 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6756 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6757 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6758 create a distinct @code{readonly_data_section}, the default is to
6759 reuse @code{text_section}.
6761 All the other @file{varasm.c} sections are optional, and are null
6762 if the target does not provide them.
6764 @defmac TEXT_SECTION_ASM_OP
6765 A C expression whose value is a string, including spacing, containing the
6766 assembler operation that should precede instructions and read-only data.
6767 Normally @code{"\t.text"} is right.
6770 @defmac HOT_TEXT_SECTION_NAME
6771 If defined, a C string constant for the name of the section containing most
6772 frequently executed functions of the program. If not defined, GCC will provide
6773 a default definition if the target supports named sections.
6776 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6777 If defined, a C string constant for the name of the section containing unlikely
6778 executed functions in the program.
6781 @defmac DATA_SECTION_ASM_OP
6782 A C expression whose value is a string, including spacing, containing the
6783 assembler operation to identify the following data as writable initialized
6784 data. Normally @code{"\t.data"} is right.
6787 @defmac SDATA_SECTION_ASM_OP
6788 If defined, a C expression whose value is a string, including spacing,
6789 containing the assembler operation to identify the following data as
6790 initialized, writable small data.
6793 @defmac READONLY_DATA_SECTION_ASM_OP
6794 A C expression whose value is a string, including spacing, containing the
6795 assembler operation to identify the following data as read-only initialized
6799 @defmac BSS_SECTION_ASM_OP
6800 If defined, a C expression whose value is a string, including spacing,
6801 containing the assembler operation to identify the following data as
6802 uninitialized global data. If not defined, and neither
6803 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6804 uninitialized global data will be output in the data section if
6805 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6809 @defmac SBSS_SECTION_ASM_OP
6810 If defined, a C expression whose value is a string, including spacing,
6811 containing the assembler operation to identify the following data as
6812 uninitialized, writable small data.
6815 @defmac TLS_COMMON_ASM_OP
6816 If defined, a C expression whose value is a string containing the
6817 assembler operation to identify the following data as thread-local
6818 common data. The default is @code{".tls_common"}.
6821 @defmac TLS_SECTION_ASM_FLAG
6822 If defined, a C expression whose value is a character constant
6823 containing the flag used to mark a section as a TLS section. The
6824 default is @code{'T'}.
6827 @defmac INIT_SECTION_ASM_OP
6828 If defined, a C expression whose value is a string, including spacing,
6829 containing the assembler operation to identify the following data as
6830 initialization code. If not defined, GCC will assume such a section does
6831 not exist. This section has no corresponding @code{init_section}
6832 variable; it is used entirely in runtime code.
6835 @defmac FINI_SECTION_ASM_OP
6836 If defined, a C expression whose value is a string, including spacing,
6837 containing the assembler operation to identify the following data as
6838 finalization code. If not defined, GCC will assume such a section does
6839 not exist. This section has no corresponding @code{fini_section}
6840 variable; it is used entirely in runtime code.
6843 @defmac INIT_ARRAY_SECTION_ASM_OP
6844 If defined, a C expression whose value is a string, including spacing,
6845 containing the assembler operation to identify the following data as
6846 part of the @code{.init_array} (or equivalent) section. If not
6847 defined, GCC will assume such a section does not exist. Do not define
6848 both this macro and @code{INIT_SECTION_ASM_OP}.
6851 @defmac FINI_ARRAY_SECTION_ASM_OP
6852 If defined, a C expression whose value is a string, including spacing,
6853 containing the assembler operation to identify the following data as
6854 part of the @code{.fini_array} (or equivalent) section. If not
6855 defined, GCC will assume such a section does not exist. Do not define
6856 both this macro and @code{FINI_SECTION_ASM_OP}.
6859 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6860 If defined, an ASM statement that switches to a different section
6861 via @var{section_op}, calls @var{function}, and switches back to
6862 the text section. This is used in @file{crtstuff.c} if
6863 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6864 to initialization and finalization functions from the init and fini
6865 sections. By default, this macro uses a simple function call. Some
6866 ports need hand-crafted assembly code to avoid dependencies on
6867 registers initialized in the function prologue or to ensure that
6868 constant pools don't end up too far way in the text section.
6871 @defmac TARGET_LIBGCC_SDATA_SECTION
6872 If defined, a string which names the section into which small
6873 variables defined in crtstuff and libgcc should go. This is useful
6874 when the target has options for optimizing access to small data, and
6875 you want the crtstuff and libgcc routines to be conservative in what
6876 they expect of your application yet liberal in what your application
6877 expects. For example, for targets with a @code{.sdata} section (like
6878 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6879 require small data support from your application, but use this macro
6880 to put small data into @code{.sdata} so that your application can
6881 access these variables whether it uses small data or not.
6884 @defmac FORCE_CODE_SECTION_ALIGN
6885 If defined, an ASM statement that aligns a code section to some
6886 arbitrary boundary. This is used to force all fragments of the
6887 @code{.init} and @code{.fini} sections to have to same alignment
6888 and thus prevent the linker from having to add any padding.
6891 @defmac JUMP_TABLES_IN_TEXT_SECTION
6892 Define this macro to be an expression with a nonzero value if jump
6893 tables (for @code{tablejump} insns) should be output in the text
6894 section, along with the assembler instructions. Otherwise, the
6895 readonly data section is used.
6897 This macro is irrelevant if there is no separate readonly data section.
6900 @hook TARGET_ASM_INIT_SECTIONS
6901 Define this hook if you need to do something special to set up the
6902 @file{varasm.c} sections, or if your target has some special sections
6903 of its own that you need to create.
6905 GCC calls this hook after processing the command line, but before writing
6906 any assembly code, and before calling any of the section-returning hooks
6910 @hook TARGET_ASM_RELOC_RW_MASK
6911 Return a mask describing how relocations should be treated when
6912 selecting sections. Bit 1 should be set if global relocations
6913 should be placed in a read-write section; bit 0 should be set if
6914 local relocations should be placed in a read-write section.
6916 The default version of this function returns 3 when @option{-fpic}
6917 is in effect, and 0 otherwise. The hook is typically redefined
6918 when the target cannot support (some kinds of) dynamic relocations
6919 in read-only sections even in executables.
6922 @hook TARGET_ASM_SELECT_SECTION
6923 Return the section into which @var{exp} should be placed. You can
6924 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6925 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6926 requires link-time relocations. Bit 0 is set when variable contains
6927 local relocations only, while bit 1 is set for global relocations.
6928 @var{align} is the constant alignment in bits.
6930 The default version of this function takes care of putting read-only
6931 variables in @code{readonly_data_section}.
6933 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6936 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6937 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6938 for @code{FUNCTION_DECL}s as well as for variables and constants.
6940 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6941 function has been determined to be likely to be called, and nonzero if
6942 it is unlikely to be called.
6945 @hook TARGET_ASM_UNIQUE_SECTION
6946 Build up a unique section name, expressed as a @code{STRING_CST} node,
6947 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6948 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6949 the initial value of @var{exp} requires link-time relocations.
6951 The default version of this function appends the symbol name to the
6952 ELF section name that would normally be used for the symbol. For
6953 example, the function @code{foo} would be placed in @code{.text.foo}.
6954 Whatever the actual target object format, this is often good enough.
6957 @hook TARGET_ASM_FUNCTION_RODATA_SECTION
6958 Return the readonly data section associated with
6959 @samp{DECL_SECTION_NAME (@var{decl})}.
6960 The default version of this function selects @code{.gnu.linkonce.r.name} if
6961 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
6962 if function is in @code{.text.name}, and the normal readonly-data section
6966 @hook TARGET_ASM_SELECT_RTX_SECTION
6967 Return the section into which a constant @var{x}, of mode @var{mode},
6968 should be placed. You can assume that @var{x} is some kind of
6969 constant in RTL@. The argument @var{mode} is redundant except in the
6970 case of a @code{const_int} rtx. @var{align} is the constant alignment
6973 The default version of this function takes care of putting symbolic
6974 constants in @code{flag_pic} mode in @code{data_section} and everything
6975 else in @code{readonly_data_section}.
6978 @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
6979 Define this hook if you need to postprocess the assembler name generated
6980 by target-independent code. The @var{id} provided to this hook will be
6981 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
6982 or the mangled name of the @var{decl} in C++). The return value of the
6983 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
6984 your target system. The default implementation of this hook just
6985 returns the @var{id} provided.
6988 @hook TARGET_ENCODE_SECTION_INFO
6989 Define this hook if references to a symbol or a constant must be
6990 treated differently depending on something about the variable or
6991 function named by the symbol (such as what section it is in).
6993 The hook is executed immediately after rtl has been created for
6994 @var{decl}, which may be a variable or function declaration or
6995 an entry in the constant pool. In either case, @var{rtl} is the
6996 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6997 in this hook; that field may not have been initialized yet.
6999 In the case of a constant, it is safe to assume that the rtl is
7000 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7001 will also have this form, but that is not guaranteed. Global
7002 register variables, for instance, will have a @code{reg} for their
7003 rtl. (Normally the right thing to do with such unusual rtl is
7006 The @var{new_decl_p} argument will be true if this is the first time
7007 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7008 be false for subsequent invocations, which will happen for duplicate
7009 declarations. Whether or not anything must be done for the duplicate
7010 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7011 @var{new_decl_p} is always true when the hook is called for a constant.
7013 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7014 The usual thing for this hook to do is to record flags in the
7015 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7016 Historically, the name string was modified if it was necessary to
7017 encode more than one bit of information, but this practice is now
7018 discouraged; use @code{SYMBOL_REF_FLAGS}.
7020 The default definition of this hook, @code{default_encode_section_info}
7021 in @file{varasm.c}, sets a number of commonly-useful bits in
7022 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7023 before overriding it.
7026 @hook TARGET_STRIP_NAME_ENCODING
7027 Decode @var{name} and return the real name part, sans
7028 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7032 @hook TARGET_IN_SMALL_DATA_P
7033 Returns true if @var{exp} should be placed into a ``small data'' section.
7034 The default version of this hook always returns false.
7037 @hook TARGET_HAVE_SRODATA_SECTION
7038 Contains the value true if the target places read-only
7039 ``small data'' into a separate section. The default value is false.
7042 @hook TARGET_PROFILE_BEFORE_PROLOGUE
7044 @hook TARGET_BINDS_LOCAL_P
7045 Returns true if @var{exp} names an object for which name resolution
7046 rules must resolve to the current ``module'' (dynamic shared library
7047 or executable image).
7049 The default version of this hook implements the name resolution rules
7050 for ELF, which has a looser model of global name binding than other
7051 currently supported object file formats.
7054 @hook TARGET_HAVE_TLS
7055 Contains the value true if the target supports thread-local storage.
7056 The default value is false.
7061 @section Position Independent Code
7062 @cindex position independent code
7065 This section describes macros that help implement generation of position
7066 independent code. Simply defining these macros is not enough to
7067 generate valid PIC; you must also add support to the hook
7068 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7069 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7070 must modify the definition of @samp{movsi} to do something appropriate
7071 when the source operand contains a symbolic address. You may also
7072 need to alter the handling of switch statements so that they use
7074 @c i rearranged the order of the macros above to try to force one of
7075 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7077 @defmac PIC_OFFSET_TABLE_REGNUM
7078 The register number of the register used to address a table of static
7079 data addresses in memory. In some cases this register is defined by a
7080 processor's ``application binary interface'' (ABI)@. When this macro
7081 is defined, RTL is generated for this register once, as with the stack
7082 pointer and frame pointer registers. If this macro is not defined, it
7083 is up to the machine-dependent files to allocate such a register (if
7084 necessary). Note that this register must be fixed when in use (e.g.@:
7085 when @code{flag_pic} is true).
7088 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7089 A C expression that is nonzero if the register defined by
7090 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7091 the default is zero. Do not define
7092 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7095 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7096 A C expression that is nonzero if @var{x} is a legitimate immediate
7097 operand on the target machine when generating position independent code.
7098 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7099 check this. You can also assume @var{flag_pic} is true, so you need not
7100 check it either. You need not define this macro if all constants
7101 (including @code{SYMBOL_REF}) can be immediate operands when generating
7102 position independent code.
7105 @node Assembler Format
7106 @section Defining the Output Assembler Language
7108 This section describes macros whose principal purpose is to describe how
7109 to write instructions in assembler language---rather than what the
7113 * File Framework:: Structural information for the assembler file.
7114 * Data Output:: Output of constants (numbers, strings, addresses).
7115 * Uninitialized Data:: Output of uninitialized variables.
7116 * Label Output:: Output and generation of labels.
7117 * Initialization:: General principles of initialization
7118 and termination routines.
7119 * Macros for Initialization::
7120 Specific macros that control the handling of
7121 initialization and termination routines.
7122 * Instruction Output:: Output of actual instructions.
7123 * Dispatch Tables:: Output of jump tables.
7124 * Exception Region Output:: Output of exception region code.
7125 * Alignment Output:: Pseudo ops for alignment and skipping data.
7128 @node File Framework
7129 @subsection The Overall Framework of an Assembler File
7130 @cindex assembler format
7131 @cindex output of assembler code
7133 @c prevent bad page break with this line
7134 This describes the overall framework of an assembly file.
7136 @findex default_file_start
7137 @hook TARGET_ASM_FILE_START
7138 Output to @code{asm_out_file} any text which the assembler expects to
7139 find at the beginning of a file. The default behavior is controlled
7140 by two flags, documented below. Unless your target's assembler is
7141 quite unusual, if you override the default, you should call
7142 @code{default_file_start} at some point in your target hook. This
7143 lets other target files rely on these variables.
7146 @hook TARGET_ASM_FILE_START_APP_OFF
7147 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7148 printed as the very first line in the assembly file, unless
7149 @option{-fverbose-asm} is in effect. (If that macro has been defined
7150 to the empty string, this variable has no effect.) With the normal
7151 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7152 assembler that it need not bother stripping comments or extra
7153 whitespace from its input. This allows it to work a bit faster.
7155 The default is false. You should not set it to true unless you have
7156 verified that your port does not generate any extra whitespace or
7157 comments that will cause GAS to issue errors in NO_APP mode.
7160 @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
7161 If this flag is true, @code{output_file_directive} will be called
7162 for the primary source file, immediately after printing
7163 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7164 this to be done. The default is false.
7167 @hook TARGET_ASM_FILE_END
7168 Output to @code{asm_out_file} any text which the assembler expects
7169 to find at the end of a file. The default is to output nothing.
7172 @deftypefun void file_end_indicate_exec_stack ()
7173 Some systems use a common convention, the @samp{.note.GNU-stack}
7174 special section, to indicate whether or not an object file relies on
7175 the stack being executable. If your system uses this convention, you
7176 should define @code{TARGET_ASM_FILE_END} to this function. If you
7177 need to do other things in that hook, have your hook function call
7181 @hook TARGET_ASM_LTO_START
7182 Output to @code{asm_out_file} any text which the assembler expects
7183 to find at the start of an LTO section. The default is to output
7187 @hook TARGET_ASM_LTO_END
7188 Output to @code{asm_out_file} any text which the assembler expects
7189 to find at the end of an LTO section. The default is to output
7193 @hook TARGET_ASM_CODE_END
7194 Output to @code{asm_out_file} any text which is needed before emitting
7195 unwind info and debug info at the end of a file. Some targets emit
7196 here PIC setup thunks that cannot be emitted at the end of file,
7197 because they couldn't have unwind info then. The default is to output
7201 @defmac ASM_COMMENT_START
7202 A C string constant describing how to begin a comment in the target
7203 assembler language. The compiler assumes that the comment will end at
7204 the end of the line.
7208 A C string constant for text to be output before each @code{asm}
7209 statement or group of consecutive ones. Normally this is
7210 @code{"#APP"}, which is a comment that has no effect on most
7211 assemblers but tells the GNU assembler that it must check the lines
7212 that follow for all valid assembler constructs.
7216 A C string constant for text to be output after each @code{asm}
7217 statement or group of consecutive ones. Normally this is
7218 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7219 time-saving assumptions that are valid for ordinary compiler output.
7222 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7223 A C statement to output COFF information or DWARF debugging information
7224 which indicates that filename @var{name} is the current source file to
7225 the stdio stream @var{stream}.
7227 This macro need not be defined if the standard form of output
7228 for the file format in use is appropriate.
7231 @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
7233 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7234 A C statement to output the string @var{string} to the stdio stream
7235 @var{stream}. If you do not call the function @code{output_quoted_string}
7236 in your config files, GCC will only call it to output filenames to
7237 the assembler source. So you can use it to canonicalize the format
7238 of the filename using this macro.
7241 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7242 A C statement to output something to the assembler file to handle a
7243 @samp{#ident} directive containing the text @var{string}. If this
7244 macro is not defined, nothing is output for a @samp{#ident} directive.
7247 @hook TARGET_ASM_NAMED_SECTION
7248 Output assembly directives to switch to section @var{name}. The section
7249 should have attributes as specified by @var{flags}, which is a bit mask
7250 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7251 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7252 this section is associated.
7255 @hook TARGET_ASM_FUNCTION_SECTION
7256 Return preferred text (sub)section for function @var{decl}.
7257 Main purpose of this function is to separate cold, normal and hot
7258 functions. @var{startup} is true when function is known to be used only
7259 at startup (from static constructors or it is @code{main()}).
7260 @var{exit} is true when function is known to be used only at exit
7261 (from static destructors).
7262 Return NULL if function should go to default text section.
7265 @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
7267 @hook TARGET_HAVE_NAMED_SECTIONS
7268 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7269 It must not be modified by command-line option processing.
7272 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7273 @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7274 This flag is true if we can create zeroed data by switching to a BSS
7275 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7276 This is true on most ELF targets.
7279 @hook TARGET_SECTION_TYPE_FLAGS
7280 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7281 based on a variable or function decl, a section name, and whether or not the
7282 declaration's initializer may contain runtime relocations. @var{decl} may be
7283 null, in which case read-write data should be assumed.
7285 The default version of this function handles choosing code vs data,
7286 read-only vs read-write data, and @code{flag_pic}. You should only
7287 need to override this if your target has special flags that might be
7288 set via @code{__attribute__}.
7291 @hook TARGET_ASM_RECORD_GCC_SWITCHES
7292 Provides the target with the ability to record the gcc command line
7293 switches that have been passed to the compiler, and options that are
7294 enabled. The @var{type} argument specifies what is being recorded.
7295 It can take the following values:
7298 @item SWITCH_TYPE_PASSED
7299 @var{text} is a command line switch that has been set by the user.
7301 @item SWITCH_TYPE_ENABLED
7302 @var{text} is an option which has been enabled. This might be as a
7303 direct result of a command line switch, or because it is enabled by
7304 default or because it has been enabled as a side effect of a different
7305 command line switch. For example, the @option{-O2} switch enables
7306 various different individual optimization passes.
7308 @item SWITCH_TYPE_DESCRIPTIVE
7309 @var{text} is either NULL or some descriptive text which should be
7310 ignored. If @var{text} is NULL then it is being used to warn the
7311 target hook that either recording is starting or ending. The first
7312 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7313 warning is for start up and the second time the warning is for
7314 wind down. This feature is to allow the target hook to make any
7315 necessary preparations before it starts to record switches and to
7316 perform any necessary tidying up after it has finished recording
7319 @item SWITCH_TYPE_LINE_START
7320 This option can be ignored by this target hook.
7322 @item SWITCH_TYPE_LINE_END
7323 This option can be ignored by this target hook.
7326 The hook's return value must be zero. Other return values may be
7327 supported in the future.
7329 By default this hook is set to NULL, but an example implementation is
7330 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7331 it records the switches as ASCII text inside a new, string mergeable
7332 section in the assembler output file. The name of the new section is
7333 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7337 @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7338 This is the name of the section that will be created by the example
7339 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7345 @subsection Output of Data
7348 @hook TARGET_ASM_BYTE_OP
7349 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7350 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7351 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7352 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7353 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7354 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7355 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7356 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7357 These hooks specify assembly directives for creating certain kinds
7358 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7359 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7360 aligned two-byte object, and so on. Any of the hooks may be
7361 @code{NULL}, indicating that no suitable directive is available.
7363 The compiler will print these strings at the start of a new line,
7364 followed immediately by the object's initial value. In most cases,
7365 the string should contain a tab, a pseudo-op, and then another tab.
7368 @hook TARGET_ASM_INTEGER
7369 The @code{assemble_integer} function uses this hook to output an
7370 integer object. @var{x} is the object's value, @var{size} is its size
7371 in bytes and @var{aligned_p} indicates whether it is aligned. The
7372 function should return @code{true} if it was able to output the
7373 object. If it returns false, @code{assemble_integer} will try to
7374 split the object into smaller parts.
7376 The default implementation of this hook will use the
7377 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7378 when the relevant string is @code{NULL}.
7381 @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
7382 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7383 can't deal with, and output assembly code to @var{file} corresponding to
7384 the pattern @var{x}. This may be used to allow machine-dependent
7385 @code{UNSPEC}s to appear within constants.
7387 If target hook fails to recognize a pattern, it must return @code{false},
7388 so that a standard error message is printed. If it prints an error message
7389 itself, by calling, for example, @code{output_operand_lossage}, it may just
7393 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
7394 A C statement to recognize @var{rtx} patterns that
7395 @code{output_addr_const} can't deal with, and output assembly code to
7396 @var{stream} corresponding to the pattern @var{x}. This may be used to
7397 allow machine-dependent @code{UNSPEC}s to appear within constants.
7399 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
7400 @code{goto fail}, so that a standard error message is printed. If it
7401 prints an error message itself, by calling, for example,
7402 @code{output_operand_lossage}, it may just complete normally.
7405 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7406 A C statement to output to the stdio stream @var{stream} an assembler
7407 instruction to assemble a string constant containing the @var{len}
7408 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7409 @code{char *} and @var{len} a C expression of type @code{int}.
7411 If the assembler has a @code{.ascii} pseudo-op as found in the
7412 Berkeley Unix assembler, do not define the macro
7413 @code{ASM_OUTPUT_ASCII}.
7416 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7417 A C statement to output word @var{n} of a function descriptor for
7418 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7419 is defined, and is otherwise unused.
7422 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7423 You may define this macro as a C expression. You should define the
7424 expression to have a nonzero value if GCC should output the constant
7425 pool for a function before the code for the function, or a zero value if
7426 GCC should output the constant pool after the function. If you do
7427 not define this macro, the usual case, GCC will output the constant
7428 pool before the function.
7431 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7432 A C statement to output assembler commands to define the start of the
7433 constant pool for a function. @var{funname} is a string giving
7434 the name of the function. Should the return type of the function
7435 be required, it can be obtained via @var{fundecl}. @var{size}
7436 is the size, in bytes, of the constant pool that will be written
7437 immediately after this call.
7439 If no constant-pool prefix is required, the usual case, this macro need
7443 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7444 A C statement (with or without semicolon) to output a constant in the
7445 constant pool, if it needs special treatment. (This macro need not do
7446 anything for RTL expressions that can be output normally.)
7448 The argument @var{file} is the standard I/O stream to output the
7449 assembler code on. @var{x} is the RTL expression for the constant to
7450 output, and @var{mode} is the machine mode (in case @var{x} is a
7451 @samp{const_int}). @var{align} is the required alignment for the value
7452 @var{x}; you should output an assembler directive to force this much
7455 The argument @var{labelno} is a number to use in an internal label for
7456 the address of this pool entry. The definition of this macro is
7457 responsible for outputting the label definition at the proper place.
7458 Here is how to do this:
7461 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7464 When you output a pool entry specially, you should end with a
7465 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7466 entry from being output a second time in the usual manner.
7468 You need not define this macro if it would do nothing.
7471 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7472 A C statement to output assembler commands to at the end of the constant
7473 pool for a function. @var{funname} is a string giving the name of the
7474 function. Should the return type of the function be required, you can
7475 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7476 constant pool that GCC wrote immediately before this call.
7478 If no constant-pool epilogue is required, the usual case, you need not
7482 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7483 Define this macro as a C expression which is nonzero if @var{C} is
7484 used as a logical line separator by the assembler. @var{STR} points
7485 to the position in the string where @var{C} was found; this can be used if
7486 a line separator uses multiple characters.
7488 If you do not define this macro, the default is that only
7489 the character @samp{;} is treated as a logical line separator.
7492 @hook TARGET_ASM_OPEN_PAREN
7493 These target hooks are C string constants, describing the syntax in the
7494 assembler for grouping arithmetic expressions. If not overridden, they
7495 default to normal parentheses, which is correct for most assemblers.
7498 These macros are provided by @file{real.h} for writing the definitions
7499 of @code{ASM_OUTPUT_DOUBLE} and the like:
7501 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7502 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7503 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7504 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7505 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7506 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7507 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7508 target's floating point representation, and store its bit pattern in
7509 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7510 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7511 simple @code{long int}. For the others, it should be an array of
7512 @code{long int}. The number of elements in this array is determined
7513 by the size of the desired target floating point data type: 32 bits of
7514 it go in each @code{long int} array element. Each array element holds
7515 32 bits of the result, even if @code{long int} is wider than 32 bits
7516 on the host machine.
7518 The array element values are designed so that you can print them out
7519 using @code{fprintf} in the order they should appear in the target
7523 @node Uninitialized Data
7524 @subsection Output of Uninitialized Variables
7526 Each of the macros in this section is used to do the whole job of
7527 outputting a single uninitialized variable.
7529 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7530 A C statement (sans semicolon) to output to the stdio stream
7531 @var{stream} the assembler definition of a common-label named
7532 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7533 is the size rounded up to whatever alignment the caller wants. It is
7534 possible that @var{size} may be zero, for instance if a struct with no
7535 other member than a zero-length array is defined. In this case, the
7536 backend must output a symbol definition that allocates at least one
7537 byte, both so that the address of the resulting object does not compare
7538 equal to any other, and because some object formats cannot even express
7539 the concept of a zero-sized common symbol, as that is how they represent
7540 an ordinary undefined external.
7542 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7543 output the name itself; before and after that, output the additional
7544 assembler syntax for defining the name, and a newline.
7546 This macro controls how the assembler definitions of uninitialized
7547 common global variables are output.
7550 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7551 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7552 separate, explicit argument. If you define this macro, it is used in
7553 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7554 handling the required alignment of the variable. The alignment is specified
7555 as the number of bits.
7558 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7559 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7560 variable to be output, if there is one, or @code{NULL_TREE} if there
7561 is no corresponding variable. If you define this macro, GCC will use it
7562 in place of both @code{ASM_OUTPUT_COMMON} and
7563 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7564 the variable's decl in order to chose what to output.
7567 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
7568 A C statement (sans semicolon) to output to the stdio stream
7569 @var{stream} the assembler definition of uninitialized global @var{decl} named
7570 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7571 is the size rounded up to whatever alignment the caller wants.
7573 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
7574 defining this macro. If unable, use the expression
7575 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7576 before and after that, output the additional assembler syntax for defining
7577 the name, and a newline.
7579 There are two ways of handling global BSS@. One is to define either
7580 this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}.
7581 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7582 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7583 You do not need to do both.
7585 Some languages do not have @code{common} data, and require a
7586 non-common form of global BSS in order to handle uninitialized globals
7587 efficiently. C++ is one example of this. However, if the target does
7588 not support global BSS, the front end may choose to make globals
7589 common in order to save space in the object file.
7592 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7593 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
7594 separate, explicit argument. If you define this macro, it is used in
7595 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
7596 handling the required alignment of the variable. The alignment is specified
7597 as the number of bits.
7599 Try to use function @code{asm_output_aligned_bss} defined in file
7600 @file{varasm.c} when defining this macro.
7603 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7604 A C statement (sans semicolon) to output to the stdio stream
7605 @var{stream} the assembler definition of a local-common-label named
7606 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7607 is the size rounded up to whatever alignment the caller wants.
7609 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7610 output the name itself; before and after that, output the additional
7611 assembler syntax for defining the name, and a newline.
7613 This macro controls how the assembler definitions of uninitialized
7614 static variables are output.
7617 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7618 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7619 separate, explicit argument. If you define this macro, it is used in
7620 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7621 handling the required alignment of the variable. The alignment is specified
7622 as the number of bits.
7625 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7626 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7627 variable to be output, if there is one, or @code{NULL_TREE} if there
7628 is no corresponding variable. If you define this macro, GCC will use it
7629 in place of both @code{ASM_OUTPUT_DECL} and
7630 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7631 the variable's decl in order to chose what to output.
7635 @subsection Output and Generation of Labels
7637 @c prevent bad page break with this line
7638 This is about outputting labels.
7640 @findex assemble_name
7641 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7642 A C statement (sans semicolon) to output to the stdio stream
7643 @var{stream} the assembler definition of a label named @var{name}.
7644 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7645 output the name itself; before and after that, output the additional
7646 assembler syntax for defining the name, and a newline. A default
7647 definition of this macro is provided which is correct for most systems.
7650 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7651 A C statement (sans semicolon) to output to the stdio stream
7652 @var{stream} the assembler definition of a label named @var{name} of
7654 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7655 output the name itself; before and after that, output the additional
7656 assembler syntax for defining the name, and a newline. A default
7657 definition of this macro is provided which is correct for most systems.
7659 If this macro is not defined, then the function name is defined in the
7660 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7663 @findex assemble_name_raw
7664 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7665 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7666 to refer to a compiler-generated label. The default definition uses
7667 @code{assemble_name_raw}, which is like @code{assemble_name} except
7668 that it is more efficient.
7672 A C string containing the appropriate assembler directive to specify the
7673 size of a symbol, without any arguments. On systems that use ELF, the
7674 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7675 systems, the default is not to define this macro.
7677 Define this macro only if it is correct to use the default definitions
7678 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7679 for your system. If you need your own custom definitions of those
7680 macros, or if you do not need explicit symbol sizes at all, do not
7684 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7685 A C statement (sans semicolon) to output to the stdio stream
7686 @var{stream} a directive telling the assembler that the size of the
7687 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7688 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7692 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7693 A C statement (sans semicolon) to output to the stdio stream
7694 @var{stream} a directive telling the assembler to calculate the size of
7695 the symbol @var{name} by subtracting its address from the current
7698 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7699 provided. The default assumes that the assembler recognizes a special
7700 @samp{.} symbol as referring to the current address, and can calculate
7701 the difference between this and another symbol. If your assembler does
7702 not recognize @samp{.} or cannot do calculations with it, you will need
7703 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7707 A C string containing the appropriate assembler directive to specify the
7708 type of a symbol, without any arguments. On systems that use ELF, the
7709 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7710 systems, the default is not to define this macro.
7712 Define this macro only if it is correct to use the default definition of
7713 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7714 custom definition of this macro, or if you do not need explicit symbol
7715 types at all, do not define this macro.
7718 @defmac TYPE_OPERAND_FMT
7719 A C string which specifies (using @code{printf} syntax) the format of
7720 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7721 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7722 the default is not to define this macro.
7724 Define this macro only if it is correct to use the default definition of
7725 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7726 custom definition of this macro, or if you do not need explicit symbol
7727 types at all, do not define this macro.
7730 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7731 A C statement (sans semicolon) to output to the stdio stream
7732 @var{stream} a directive telling the assembler that the type of the
7733 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7734 that string is always either @samp{"function"} or @samp{"object"}, but
7735 you should not count on this.
7737 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7738 definition of this macro is provided.
7741 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7742 A C statement (sans semicolon) to output to the stdio stream
7743 @var{stream} any text necessary for declaring the name @var{name} of a
7744 function which is being defined. This macro is responsible for
7745 outputting the label definition (perhaps using
7746 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7747 @code{FUNCTION_DECL} tree node representing the function.
7749 If this macro is not defined, then the function name is defined in the
7750 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7752 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7756 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7757 A C statement (sans semicolon) to output to the stdio stream
7758 @var{stream} any text necessary for declaring the size of a function
7759 which is being defined. The argument @var{name} is the name of the
7760 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7761 representing the function.
7763 If this macro is not defined, then the function size is not defined.
7765 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7769 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7770 A C statement (sans semicolon) to output to the stdio stream
7771 @var{stream} any text necessary for declaring the name @var{name} of an
7772 initialized variable which is being defined. This macro must output the
7773 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7774 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7776 If this macro is not defined, then the variable name is defined in the
7777 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7779 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7780 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7783 @hook TARGET_ASM_DECLARE_CONSTANT_NAME
7784 A target hook to output to the stdio stream @var{file} any text necessary
7785 for declaring the name @var{name} of a constant which is being defined. This
7786 target hook is responsible for outputting the label definition (perhaps using
7787 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7788 and @var{size} is the size of the constant in bytes. The @var{name}
7789 will be an internal label.
7791 The default version of this target hook, define the @var{name} in the
7792 usual manner as a label (by means of @code{assemble_label}).
7794 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7797 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7798 A C statement (sans semicolon) to output to the stdio stream
7799 @var{stream} any text necessary for claiming a register @var{regno}
7800 for a global variable @var{decl} with name @var{name}.
7802 If you don't define this macro, that is equivalent to defining it to do
7806 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7807 A C statement (sans semicolon) to finish up declaring a variable name
7808 once the compiler has processed its initializer fully and thus has had a
7809 chance to determine the size of an array when controlled by an
7810 initializer. This is used on systems where it's necessary to declare
7811 something about the size of the object.
7813 If you don't define this macro, that is equivalent to defining it to do
7816 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7817 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7820 @hook TARGET_ASM_GLOBALIZE_LABEL
7821 This target hook is a function to output to the stdio stream
7822 @var{stream} some commands that will make the label @var{name} global;
7823 that is, available for reference from other files.
7825 The default implementation relies on a proper definition of
7826 @code{GLOBAL_ASM_OP}.
7829 @hook TARGET_ASM_GLOBALIZE_DECL_NAME
7830 This target hook is a function to output to the stdio stream
7831 @var{stream} some commands that will make the name associated with @var{decl}
7832 global; that is, available for reference from other files.
7834 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7837 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7838 A C statement (sans semicolon) to output to the stdio stream
7839 @var{stream} some commands that will make the label @var{name} weak;
7840 that is, available for reference from other files but only used if
7841 no other definition is available. Use the expression
7842 @code{assemble_name (@var{stream}, @var{name})} to output the name
7843 itself; before and after that, output the additional assembler syntax
7844 for making that name weak, and a newline.
7846 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7847 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7851 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7852 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7853 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7854 or variable decl. If @var{value} is not @code{NULL}, this C statement
7855 should output to the stdio stream @var{stream} assembler code which
7856 defines (equates) the weak symbol @var{name} to have the value
7857 @var{value}. If @var{value} is @code{NULL}, it should output commands
7858 to make @var{name} weak.
7861 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7862 Outputs a directive that enables @var{name} to be used to refer to
7863 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7864 declaration of @code{name}.
7867 @defmac SUPPORTS_WEAK
7868 A preprocessor constant expression which evaluates to true if the target
7869 supports weak symbols.
7871 If you don't define this macro, @file{defaults.h} provides a default
7872 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7873 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
7876 @defmac TARGET_SUPPORTS_WEAK
7877 A C expression which evaluates to true if the target supports weak symbols.
7879 If you don't define this macro, @file{defaults.h} provides a default
7880 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
7881 this macro if you want to control weak symbol support with a compiler
7882 flag such as @option{-melf}.
7885 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7886 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7887 public symbol such that extra copies in multiple translation units will
7888 be discarded by the linker. Define this macro if your object file
7889 format provides support for this concept, such as the @samp{COMDAT}
7890 section flags in the Microsoft Windows PE/COFF format, and this support
7891 requires changes to @var{decl}, such as putting it in a separate section.
7894 @defmac SUPPORTS_ONE_ONLY
7895 A C expression which evaluates to true if the target supports one-only
7898 If you don't define this macro, @file{varasm.c} provides a default
7899 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7900 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7901 you want to control one-only symbol support with a compiler flag, or if
7902 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7903 be emitted as one-only.
7906 @hook TARGET_ASM_ASSEMBLE_VISIBILITY
7907 This target hook is a function to output to @var{asm_out_file} some
7908 commands that will make the symbol(s) associated with @var{decl} have
7909 hidden, protected or internal visibility as specified by @var{visibility}.
7912 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7913 A C expression that evaluates to true if the target's linker expects
7914 that weak symbols do not appear in a static archive's table of contents.
7915 The default is @code{0}.
7917 Leaving weak symbols out of an archive's table of contents means that,
7918 if a symbol will only have a definition in one translation unit and
7919 will have undefined references from other translation units, that
7920 symbol should not be weak. Defining this macro to be nonzero will
7921 thus have the effect that certain symbols that would normally be weak
7922 (explicit template instantiations, and vtables for polymorphic classes
7923 with noninline key methods) will instead be nonweak.
7925 The C++ ABI requires this macro to be zero. Define this macro for
7926 targets where full C++ ABI compliance is impossible and where linker
7927 restrictions require weak symbols to be left out of a static archive's
7931 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7932 A C statement (sans semicolon) to output to the stdio stream
7933 @var{stream} any text necessary for declaring the name of an external
7934 symbol named @var{name} which is referenced in this compilation but
7935 not defined. The value of @var{decl} is the tree node for the
7938 This macro need not be defined if it does not need to output anything.
7939 The GNU assembler and most Unix assemblers don't require anything.
7942 @hook TARGET_ASM_EXTERNAL_LIBCALL
7943 This target hook is a function to output to @var{asm_out_file} an assembler
7944 pseudo-op to declare a library function name external. The name of the
7945 library function is given by @var{symref}, which is a @code{symbol_ref}.
7948 @hook TARGET_ASM_MARK_DECL_PRESERVED
7949 This target hook is a function to output to @var{asm_out_file} an assembler
7950 directive to annotate @var{symbol} as used. The Darwin target uses the
7951 .no_dead_code_strip directive.
7954 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
7955 A C statement (sans semicolon) to output to the stdio stream
7956 @var{stream} a reference in assembler syntax to a label named
7957 @var{name}. This should add @samp{_} to the front of the name, if that
7958 is customary on your operating system, as it is in most Berkeley Unix
7959 systems. This macro is used in @code{assemble_name}.
7962 @hook TARGET_MANGLE_ASSEMBLER_NAME
7964 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
7965 A C statement (sans semicolon) to output a reference to
7966 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
7967 will be used to output the name of the symbol. This macro may be used
7968 to modify the way a symbol is referenced depending on information
7969 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
7972 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
7973 A C statement (sans semicolon) to output a reference to @var{buf}, the
7974 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
7975 @code{assemble_name} will be used to output the name of the symbol.
7976 This macro is not used by @code{output_asm_label}, or the @code{%l}
7977 specifier that calls it; the intention is that this macro should be set
7978 when it is necessary to output a label differently when its address is
7982 @hook TARGET_ASM_INTERNAL_LABEL
7983 A function to output to the stdio stream @var{stream} a label whose
7984 name is made from the string @var{prefix} and the number @var{labelno}.
7986 It is absolutely essential that these labels be distinct from the labels
7987 used for user-level functions and variables. Otherwise, certain programs
7988 will have name conflicts with internal labels.
7990 It is desirable to exclude internal labels from the symbol table of the
7991 object file. Most assemblers have a naming convention for labels that
7992 should be excluded; on many systems, the letter @samp{L} at the
7993 beginning of a label has this effect. You should find out what
7994 convention your system uses, and follow it.
7996 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
7999 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8000 A C statement to output to the stdio stream @var{stream} a debug info
8001 label whose name is made from the string @var{prefix} and the number
8002 @var{num}. This is useful for VLIW targets, where debug info labels
8003 may need to be treated differently than branch target labels. On some
8004 systems, branch target labels must be at the beginning of instruction
8005 bundles, but debug info labels can occur in the middle of instruction
8008 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8012 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8013 A C statement to store into the string @var{string} a label whose name
8014 is made from the string @var{prefix} and the number @var{num}.
8016 This string, when output subsequently by @code{assemble_name}, should
8017 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8018 with the same @var{prefix} and @var{num}.
8020 If the string begins with @samp{*}, then @code{assemble_name} will
8021 output the rest of the string unchanged. It is often convenient for
8022 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8023 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8024 to output the string, and may change it. (Of course,
8025 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8026 you should know what it does on your machine.)
8029 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8030 A C expression to assign to @var{outvar} (which is a variable of type
8031 @code{char *}) a newly allocated string made from the string
8032 @var{name} and the number @var{number}, with some suitable punctuation
8033 added. Use @code{alloca} to get space for the string.
8035 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8036 produce an assembler label for an internal static variable whose name is
8037 @var{name}. Therefore, the string must be such as to result in valid
8038 assembler code. The argument @var{number} is different each time this
8039 macro is executed; it prevents conflicts between similarly-named
8040 internal static variables in different scopes.
8042 Ideally this string should not be a valid C identifier, to prevent any
8043 conflict with the user's own symbols. Most assemblers allow periods
8044 or percent signs in assembler symbols; putting at least one of these
8045 between the name and the number will suffice.
8047 If this macro is not defined, a default definition will be provided
8048 which is correct for most systems.
8051 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8052 A C statement to output to the stdio stream @var{stream} assembler code
8053 which defines (equates) the symbol @var{name} to have the value @var{value}.
8056 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8057 correct for most systems.
8060 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8061 A C statement to output to the stdio stream @var{stream} assembler code
8062 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8063 to have the value of the tree node @var{decl_of_value}. This macro will
8064 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8065 the tree nodes are available.
8068 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8069 correct for most systems.
8072 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8073 A C statement that evaluates to true if the assembler code which defines
8074 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8075 of the tree node @var{decl_of_value} should be emitted near the end of the
8076 current compilation unit. The default is to not defer output of defines.
8077 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8078 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8081 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8082 A C statement to output to the stdio stream @var{stream} assembler code
8083 which defines (equates) the weak symbol @var{name} to have the value
8084 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8085 an undefined weak symbol.
8087 Define this macro if the target only supports weak aliases; define
8088 @code{ASM_OUTPUT_DEF} instead if possible.
8091 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8092 Define this macro to override the default assembler names used for
8093 Objective-C methods.
8095 The default name is a unique method number followed by the name of the
8096 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8097 the category is also included in the assembler name (e.g.@:
8100 These names are safe on most systems, but make debugging difficult since
8101 the method's selector is not present in the name. Therefore, particular
8102 systems define other ways of computing names.
8104 @var{buf} is an expression of type @code{char *} which gives you a
8105 buffer in which to store the name; its length is as long as
8106 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8107 50 characters extra.
8109 The argument @var{is_inst} specifies whether the method is an instance
8110 method or a class method; @var{class_name} is the name of the class;
8111 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8112 in a category); and @var{sel_name} is the name of the selector.
8114 On systems where the assembler can handle quoted names, you can use this
8115 macro to provide more human-readable names.
8118 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
8119 A C statement (sans semicolon) to output to the stdio stream
8120 @var{stream} commands to declare that the label @var{name} is an
8121 Objective-C class reference. This is only needed for targets whose
8122 linkers have special support for NeXT-style runtimes.
8125 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
8126 A C statement (sans semicolon) to output to the stdio stream
8127 @var{stream} commands to declare that the label @var{name} is an
8128 unresolved Objective-C class reference. This is only needed for targets
8129 whose linkers have special support for NeXT-style runtimes.
8132 @node Initialization
8133 @subsection How Initialization Functions Are Handled
8134 @cindex initialization routines
8135 @cindex termination routines
8136 @cindex constructors, output of
8137 @cindex destructors, output of
8139 The compiled code for certain languages includes @dfn{constructors}
8140 (also called @dfn{initialization routines})---functions to initialize
8141 data in the program when the program is started. These functions need
8142 to be called before the program is ``started''---that is to say, before
8143 @code{main} is called.
8145 Compiling some languages generates @dfn{destructors} (also called
8146 @dfn{termination routines}) that should be called when the program
8149 To make the initialization and termination functions work, the compiler
8150 must output something in the assembler code to cause those functions to
8151 be called at the appropriate time. When you port the compiler to a new
8152 system, you need to specify how to do this.
8154 There are two major ways that GCC currently supports the execution of
8155 initialization and termination functions. Each way has two variants.
8156 Much of the structure is common to all four variations.
8158 @findex __CTOR_LIST__
8159 @findex __DTOR_LIST__
8160 The linker must build two lists of these functions---a list of
8161 initialization functions, called @code{__CTOR_LIST__}, and a list of
8162 termination functions, called @code{__DTOR_LIST__}.
8164 Each list always begins with an ignored function pointer (which may hold
8165 0, @minus{}1, or a count of the function pointers after it, depending on
8166 the environment). This is followed by a series of zero or more function
8167 pointers to constructors (or destructors), followed by a function
8168 pointer containing zero.
8170 Depending on the operating system and its executable file format, either
8171 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8172 time and exit time. Constructors are called in reverse order of the
8173 list; destructors in forward order.
8175 The best way to handle static constructors works only for object file
8176 formats which provide arbitrarily-named sections. A section is set
8177 aside for a list of constructors, and another for a list of destructors.
8178 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8179 object file that defines an initialization function also puts a word in
8180 the constructor section to point to that function. The linker
8181 accumulates all these words into one contiguous @samp{.ctors} section.
8182 Termination functions are handled similarly.
8184 This method will be chosen as the default by @file{target-def.h} if
8185 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8186 support arbitrary sections, but does support special designated
8187 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8188 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8190 When arbitrary sections are available, there are two variants, depending
8191 upon how the code in @file{crtstuff.c} is called. On systems that
8192 support a @dfn{.init} section which is executed at program startup,
8193 parts of @file{crtstuff.c} are compiled into that section. The
8194 program is linked by the @command{gcc} driver like this:
8197 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8200 The prologue of a function (@code{__init}) appears in the @code{.init}
8201 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8202 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8203 files are provided by the operating system or by the GNU C library, but
8204 are provided by GCC for a few targets.
8206 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8207 compiled from @file{crtstuff.c}. They contain, among other things, code
8208 fragments within the @code{.init} and @code{.fini} sections that branch
8209 to routines in the @code{.text} section. The linker will pull all parts
8210 of a section together, which results in a complete @code{__init} function
8211 that invokes the routines we need at startup.
8213 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8216 If no init section is available, when GCC compiles any function called
8217 @code{main} (or more accurately, any function designated as a program
8218 entry point by the language front end calling @code{expand_main_function}),
8219 it inserts a procedure call to @code{__main} as the first executable code
8220 after the function prologue. The @code{__main} function is defined
8221 in @file{libgcc2.c} and runs the global constructors.
8223 In file formats that don't support arbitrary sections, there are again
8224 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8225 and an `a.out' format must be used. In this case,
8226 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8227 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8228 and with the address of the void function containing the initialization
8229 code as its value. The GNU linker recognizes this as a request to add
8230 the value to a @dfn{set}; the values are accumulated, and are eventually
8231 placed in the executable as a vector in the format described above, with
8232 a leading (ignored) count and a trailing zero element.
8233 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8234 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8235 the compilation of @code{main} to call @code{__main} as above, starting
8236 the initialization process.
8238 The last variant uses neither arbitrary sections nor the GNU linker.
8239 This is preferable when you want to do dynamic linking and when using
8240 file formats which the GNU linker does not support, such as `ECOFF'@. In
8241 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8242 termination functions are recognized simply by their names. This requires
8243 an extra program in the linkage step, called @command{collect2}. This program
8244 pretends to be the linker, for use with GCC; it does its job by running
8245 the ordinary linker, but also arranges to include the vectors of
8246 initialization and termination functions. These functions are called
8247 via @code{__main} as described above. In order to use this method,
8248 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8251 The following section describes the specific macros that control and
8252 customize the handling of initialization and termination functions.
8255 @node Macros for Initialization
8256 @subsection Macros Controlling Initialization Routines
8258 Here are the macros that control how the compiler handles initialization
8259 and termination functions:
8261 @defmac INIT_SECTION_ASM_OP
8262 If defined, a C string constant, including spacing, for the assembler
8263 operation to identify the following data as initialization code. If not
8264 defined, GCC will assume such a section does not exist. When you are
8265 using special sections for initialization and termination functions, this
8266 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8267 run the initialization functions.
8270 @defmac HAS_INIT_SECTION
8271 If defined, @code{main} will not call @code{__main} as described above.
8272 This macro should be defined for systems that control start-up code
8273 on a symbol-by-symbol basis, such as OSF/1, and should not
8274 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8277 @defmac LD_INIT_SWITCH
8278 If defined, a C string constant for a switch that tells the linker that
8279 the following symbol is an initialization routine.
8282 @defmac LD_FINI_SWITCH
8283 If defined, a C string constant for a switch that tells the linker that
8284 the following symbol is a finalization routine.
8287 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8288 If defined, a C statement that will write a function that can be
8289 automatically called when a shared library is loaded. The function
8290 should call @var{func}, which takes no arguments. If not defined, and
8291 the object format requires an explicit initialization function, then a
8292 function called @code{_GLOBAL__DI} will be generated.
8294 This function and the following one are used by collect2 when linking a
8295 shared library that needs constructors or destructors, or has DWARF2
8296 exception tables embedded in the code.
8299 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8300 If defined, a C statement that will write a function that can be
8301 automatically called when a shared library is unloaded. The function
8302 should call @var{func}, which takes no arguments. If not defined, and
8303 the object format requires an explicit finalization function, then a
8304 function called @code{_GLOBAL__DD} will be generated.
8307 @defmac INVOKE__main
8308 If defined, @code{main} will call @code{__main} despite the presence of
8309 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8310 where the init section is not actually run automatically, but is still
8311 useful for collecting the lists of constructors and destructors.
8314 @defmac SUPPORTS_INIT_PRIORITY
8315 If nonzero, the C++ @code{init_priority} attribute is supported and the
8316 compiler should emit instructions to control the order of initialization
8317 of objects. If zero, the compiler will issue an error message upon
8318 encountering an @code{init_priority} attribute.
8321 @hook TARGET_HAVE_CTORS_DTORS
8322 This value is true if the target supports some ``native'' method of
8323 collecting constructors and destructors to be run at startup and exit.
8324 It is false if we must use @command{collect2}.
8327 @hook TARGET_ASM_CONSTRUCTOR
8328 If defined, a function that outputs assembler code to arrange to call
8329 the function referenced by @var{symbol} at initialization time.
8331 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8332 no arguments and with no return value. If the target supports initialization
8333 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8334 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8336 If this macro is not defined by the target, a suitable default will
8337 be chosen if (1) the target supports arbitrary section names, (2) the
8338 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8342 @hook TARGET_ASM_DESTRUCTOR
8343 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8344 functions rather than initialization functions.
8347 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8348 generated for the generated object file will have static linkage.
8350 If your system uses @command{collect2} as the means of processing
8351 constructors, then that program normally uses @command{nm} to scan
8352 an object file for constructor functions to be called.
8354 On certain kinds of systems, you can define this macro to make
8355 @command{collect2} work faster (and, in some cases, make it work at all):
8357 @defmac OBJECT_FORMAT_COFF
8358 Define this macro if the system uses COFF (Common Object File Format)
8359 object files, so that @command{collect2} can assume this format and scan
8360 object files directly for dynamic constructor/destructor functions.
8362 This macro is effective only in a native compiler; @command{collect2} as
8363 part of a cross compiler always uses @command{nm} for the target machine.
8366 @defmac REAL_NM_FILE_NAME
8367 Define this macro as a C string constant containing the file name to use
8368 to execute @command{nm}. The default is to search the path normally for
8373 @command{collect2} calls @command{nm} to scan object files for static
8374 constructors and destructors and LTO info. By default, @option{-n} is
8375 passed. Define @code{NM_FLAGS} to a C string constant if other options
8376 are needed to get the same output format as GNU @command{nm -n}
8380 If your system supports shared libraries and has a program to list the
8381 dynamic dependencies of a given library or executable, you can define
8382 these macros to enable support for running initialization and
8383 termination functions in shared libraries:
8386 Define this macro to a C string constant containing the name of the program
8387 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8390 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8391 Define this macro to be C code that extracts filenames from the output
8392 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8393 of type @code{char *} that points to the beginning of a line of output
8394 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8395 code must advance @var{ptr} to the beginning of the filename on that
8396 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8399 @defmac SHLIB_SUFFIX
8400 Define this macro to a C string constant containing the default shared
8401 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8402 strips version information after this suffix when generating global
8403 constructor and destructor names. This define is only needed on targets
8404 that use @command{collect2} to process constructors and destructors.
8407 @node Instruction Output
8408 @subsection Output of Assembler Instructions
8410 @c prevent bad page break with this line
8411 This describes assembler instruction output.
8413 @defmac REGISTER_NAMES
8414 A C initializer containing the assembler's names for the machine
8415 registers, each one as a C string constant. This is what translates
8416 register numbers in the compiler into assembler language.
8419 @defmac ADDITIONAL_REGISTER_NAMES
8420 If defined, a C initializer for an array of structures containing a name
8421 and a register number. This macro defines additional names for hard
8422 registers, thus allowing the @code{asm} option in declarations to refer
8423 to registers using alternate names.
8426 @defmac OVERLAPPING_REGISTER_NAMES
8427 If defined, a C initializer for an array of structures containing a
8428 name, a register number and a count of the number of consecutive
8429 machine registers the name overlaps. This macro defines additional
8430 names for hard registers, thus allowing the @code{asm} option in
8431 declarations to refer to registers using alternate names. Unlike
8432 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8433 register name implies multiple underlying registers.
8435 This macro should be used when it is important that a clobber in an
8436 @code{asm} statement clobbers all the underlying values implied by the
8437 register name. For example, on ARM, clobbering the double-precision
8438 VFP register ``d0'' implies clobbering both single-precision registers
8442 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8443 Define this macro if you are using an unusual assembler that
8444 requires different names for the machine instructions.
8446 The definition is a C statement or statements which output an
8447 assembler instruction opcode to the stdio stream @var{stream}. The
8448 macro-operand @var{ptr} is a variable of type @code{char *} which
8449 points to the opcode name in its ``internal'' form---the form that is
8450 written in the machine description. The definition should output the
8451 opcode name to @var{stream}, performing any translation you desire, and
8452 increment the variable @var{ptr} to point at the end of the opcode
8453 so that it will not be output twice.
8455 In fact, your macro definition may process less than the entire opcode
8456 name, or more than the opcode name; but if you want to process text
8457 that includes @samp{%}-sequences to substitute operands, you must take
8458 care of the substitution yourself. Just be sure to increment
8459 @var{ptr} over whatever text should not be output normally.
8461 @findex recog_data.operand
8462 If you need to look at the operand values, they can be found as the
8463 elements of @code{recog_data.operand}.
8465 If the macro definition does nothing, the instruction is output
8469 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8470 If defined, a C statement to be executed just prior to the output of
8471 assembler code for @var{insn}, to modify the extracted operands so
8472 they will be output differently.
8474 Here the argument @var{opvec} is the vector containing the operands
8475 extracted from @var{insn}, and @var{noperands} is the number of
8476 elements of the vector which contain meaningful data for this insn.
8477 The contents of this vector are what will be used to convert the insn
8478 template into assembler code, so you can change the assembler output
8479 by changing the contents of the vector.
8481 This macro is useful when various assembler syntaxes share a single
8482 file of instruction patterns; by defining this macro differently, you
8483 can cause a large class of instructions to be output differently (such
8484 as with rearranged operands). Naturally, variations in assembler
8485 syntax affecting individual insn patterns ought to be handled by
8486 writing conditional output routines in those patterns.
8488 If this macro is not defined, it is equivalent to a null statement.
8491 @hook TARGET_ASM_FINAL_POSTSCAN_INSN
8492 If defined, this target hook is a function which is executed just after the
8493 output of assembler code for @var{insn}, to change the mode of the assembler
8496 Here the argument @var{opvec} is the vector containing the operands
8497 extracted from @var{insn}, and @var{noperands} is the number of
8498 elements of the vector which contain meaningful data for this insn.
8499 The contents of this vector are what was used to convert the insn
8500 template into assembler code, so you can change the assembler mode
8501 by checking the contents of the vector.
8504 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8505 A C compound statement to output to stdio stream @var{stream} the
8506 assembler syntax for an instruction operand @var{x}. @var{x} is an
8509 @var{code} is a value that can be used to specify one of several ways
8510 of printing the operand. It is used when identical operands must be
8511 printed differently depending on the context. @var{code} comes from
8512 the @samp{%} specification that was used to request printing of the
8513 operand. If the specification was just @samp{%@var{digit}} then
8514 @var{code} is 0; if the specification was @samp{%@var{ltr}
8515 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8518 If @var{x} is a register, this macro should print the register's name.
8519 The names can be found in an array @code{reg_names} whose type is
8520 @code{char *[]}. @code{reg_names} is initialized from
8521 @code{REGISTER_NAMES}.
8523 When the machine description has a specification @samp{%@var{punct}}
8524 (a @samp{%} followed by a punctuation character), this macro is called
8525 with a null pointer for @var{x} and the punctuation character for
8529 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8530 A C expression which evaluates to true if @var{code} is a valid
8531 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8532 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8533 punctuation characters (except for the standard one, @samp{%}) are used
8537 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8538 A C compound statement to output to stdio stream @var{stream} the
8539 assembler syntax for an instruction operand that is a memory reference
8540 whose address is @var{x}. @var{x} is an RTL expression.
8542 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8543 On some machines, the syntax for a symbolic address depends on the
8544 section that the address refers to. On these machines, define the hook
8545 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8546 @code{symbol_ref}, and then check for it here. @xref{Assembler
8550 @findex dbr_sequence_length
8551 @defmac DBR_OUTPUT_SEQEND (@var{file})
8552 A C statement, to be executed after all slot-filler instructions have
8553 been output. If necessary, call @code{dbr_sequence_length} to
8554 determine the number of slots filled in a sequence (zero if not
8555 currently outputting a sequence), to decide how many no-ops to output,
8558 Don't define this macro if it has nothing to do, but it is helpful in
8559 reading assembly output if the extent of the delay sequence is made
8560 explicit (e.g.@: with white space).
8563 @findex final_sequence
8564 Note that output routines for instructions with delay slots must be
8565 prepared to deal with not being output as part of a sequence
8566 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8567 found.) The variable @code{final_sequence} is null when not
8568 processing a sequence, otherwise it contains the @code{sequence} rtx
8572 @defmac REGISTER_PREFIX
8573 @defmacx LOCAL_LABEL_PREFIX
8574 @defmacx USER_LABEL_PREFIX
8575 @defmacx IMMEDIATE_PREFIX
8576 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8577 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8578 @file{final.c}). These are useful when a single @file{md} file must
8579 support multiple assembler formats. In that case, the various @file{tm.h}
8580 files can define these macros differently.
8583 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8584 If defined this macro should expand to a series of @code{case}
8585 statements which will be parsed inside the @code{switch} statement of
8586 the @code{asm_fprintf} function. This allows targets to define extra
8587 printf formats which may useful when generating their assembler
8588 statements. Note that uppercase letters are reserved for future
8589 generic extensions to asm_fprintf, and so are not available to target
8590 specific code. The output file is given by the parameter @var{file}.
8591 The varargs input pointer is @var{argptr} and the rest of the format
8592 string, starting the character after the one that is being switched
8593 upon, is pointed to by @var{format}.
8596 @defmac ASSEMBLER_DIALECT
8597 If your target supports multiple dialects of assembler language (such as
8598 different opcodes), define this macro as a C expression that gives the
8599 numeric index of the assembler language dialect to use, with zero as the
8602 If this macro is defined, you may use constructs of the form
8604 @samp{@{option0|option1|option2@dots{}@}}
8607 in the output templates of patterns (@pxref{Output Template}) or in the
8608 first argument of @code{asm_fprintf}. This construct outputs
8609 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8610 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8611 within these strings retain their usual meaning. If there are fewer
8612 alternatives within the braces than the value of
8613 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8615 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8616 @samp{@}} do not have any special meaning when used in templates or
8617 operands to @code{asm_fprintf}.
8619 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8620 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8621 the variations in assembler language syntax with that mechanism. Define
8622 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8623 if the syntax variant are larger and involve such things as different
8624 opcodes or operand order.
8627 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8628 A C expression to output to @var{stream} some assembler code
8629 which will push hard register number @var{regno} onto the stack.
8630 The code need not be optimal, since this macro is used only when
8634 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8635 A C expression to output to @var{stream} some assembler code
8636 which will pop hard register number @var{regno} off of the stack.
8637 The code need not be optimal, since this macro is used only when
8641 @node Dispatch Tables
8642 @subsection Output of Dispatch Tables
8644 @c prevent bad page break with this line
8645 This concerns dispatch tables.
8647 @cindex dispatch table
8648 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8649 A C statement to output to the stdio stream @var{stream} an assembler
8650 pseudo-instruction to generate a difference between two labels.
8651 @var{value} and @var{rel} are the numbers of two internal labels. The
8652 definitions of these labels are output using
8653 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8654 way here. For example,
8657 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8658 @var{value}, @var{rel})
8661 You must provide this macro on machines where the addresses in a
8662 dispatch table are relative to the table's own address. If defined, GCC
8663 will also use this macro on all machines when producing PIC@.
8664 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8665 mode and flags can be read.
8668 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8669 This macro should be provided on machines where the addresses
8670 in a dispatch table are absolute.
8672 The definition should be a C statement to output to the stdio stream
8673 @var{stream} an assembler pseudo-instruction to generate a reference to
8674 a label. @var{value} is the number of an internal label whose
8675 definition is output using @code{(*targetm.asm_out.internal_label)}.
8679 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8683 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8684 Define this if the label before a jump-table needs to be output
8685 specially. The first three arguments are the same as for
8686 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8687 jump-table which follows (a @code{jump_insn} containing an
8688 @code{addr_vec} or @code{addr_diff_vec}).
8690 This feature is used on system V to output a @code{swbeg} statement
8693 If this macro is not defined, these labels are output with
8694 @code{(*targetm.asm_out.internal_label)}.
8697 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8698 Define this if something special must be output at the end of a
8699 jump-table. The definition should be a C statement to be executed
8700 after the assembler code for the table is written. It should write
8701 the appropriate code to stdio stream @var{stream}. The argument
8702 @var{table} is the jump-table insn, and @var{num} is the label-number
8703 of the preceding label.
8705 If this macro is not defined, nothing special is output at the end of
8709 @hook TARGET_ASM_EMIT_UNWIND_LABEL
8710 This target hook emits a label at the beginning of each FDE@. It
8711 should be defined on targets where FDEs need special labels, and it
8712 should write the appropriate label, for the FDE associated with the
8713 function declaration @var{decl}, to the stdio stream @var{stream}.
8714 The third argument, @var{for_eh}, is a boolean: true if this is for an
8715 exception table. The fourth argument, @var{empty}, is a boolean:
8716 true if this is a placeholder label for an omitted FDE@.
8718 The default is that FDEs are not given nonlocal labels.
8721 @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
8722 This target hook emits a label at the beginning of the exception table.
8723 It should be defined on targets where it is desirable for the table
8724 to be broken up according to function.
8726 The default is that no label is emitted.
8729 @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
8731 @hook TARGET_ASM_UNWIND_EMIT
8732 This target hook emits assembly directives required to unwind the
8733 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8734 returns @code{UI_TARGET}.
8737 @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8739 @node Exception Region Output
8740 @subsection Assembler Commands for Exception Regions
8742 @c prevent bad page break with this line
8744 This describes commands marking the start and the end of an exception
8747 @defmac EH_FRAME_SECTION_NAME
8748 If defined, a C string constant for the name of the section containing
8749 exception handling frame unwind information. If not defined, GCC will
8750 provide a default definition if the target supports named sections.
8751 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8753 You should define this symbol if your target supports DWARF 2 frame
8754 unwind information and the default definition does not work.
8757 @defmac EH_FRAME_IN_DATA_SECTION
8758 If defined, DWARF 2 frame unwind information will be placed in the
8759 data section even though the target supports named sections. This
8760 might be necessary, for instance, if the system linker does garbage
8761 collection and sections cannot be marked as not to be collected.
8763 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8767 @defmac EH_TABLES_CAN_BE_READ_ONLY
8768 Define this macro to 1 if your target is such that no frame unwind
8769 information encoding used with non-PIC code will ever require a
8770 runtime relocation, but the linker may not support merging read-only
8771 and read-write sections into a single read-write section.
8774 @defmac MASK_RETURN_ADDR
8775 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8776 that it does not contain any extraneous set bits in it.
8779 @defmac DWARF2_UNWIND_INFO
8780 Define this macro to 0 if your target supports DWARF 2 frame unwind
8781 information, but it does not yet work with exception handling.
8782 Otherwise, if your target supports this information (if it defines
8783 @code{INCOMING_RETURN_ADDR_RTX} and either @code{UNALIGNED_INT_ASM_OP}
8784 or @code{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8787 @hook TARGET_EXCEPT_UNWIND_INFO
8788 This hook defines the mechanism that will be used for exception handling
8789 by the target. If the target has ABI specified unwind tables, the hook
8790 should return @code{UI_TARGET}. If the target is to use the
8791 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8792 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8793 information, the hook should return @code{UI_DWARF2}.
8795 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8796 This may end up simplifying other parts of target-specific code. The
8797 default implementation of this hook never returns @code{UI_NONE}.
8799 Note that the value returned by this hook should be constant. It should
8800 not depend on anything except the command-line switches described by
8801 @var{opts}. In particular, the
8802 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8803 macros and builtin functions related to exception handling are set up
8804 depending on this setting.
8806 The default implementation of the hook first honors the
8807 @option{--enable-sjlj-exceptions} configure option, then
8808 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
8809 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
8810 must define this hook so that @var{opts} is used correctly.
8813 @hook TARGET_UNWIND_TABLES_DEFAULT
8814 This variable should be set to @code{true} if the target ABI requires unwinding
8815 tables even when exceptions are not used. It must not be modified by
8816 command-line option processing.
8819 @defmac DONT_USE_BUILTIN_SETJMP
8820 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8821 should use the @code{setjmp}/@code{longjmp} functions from the C library
8822 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8825 @defmac DWARF_CIE_DATA_ALIGNMENT
8826 This macro need only be defined if the target might save registers in the
8827 function prologue at an offset to the stack pointer that is not aligned to
8828 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8829 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8830 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8831 the target supports DWARF 2 frame unwind information.
8834 @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
8835 Contains the value true if the target should add a zero word onto the
8836 end of a Dwarf-2 frame info section when used for exception handling.
8837 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8841 @hook TARGET_DWARF_REGISTER_SPAN
8842 Given a register, this hook should return a parallel of registers to
8843 represent where to find the register pieces. Define this hook if the
8844 register and its mode are represented in Dwarf in non-contiguous
8845 locations, or if the register should be represented in more than one
8846 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8847 If not defined, the default is to return @code{NULL_RTX}.
8850 @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
8851 If some registers are represented in Dwarf-2 unwind information in
8852 multiple pieces, define this hook to fill in information about the
8853 sizes of those pieces in the table used by the unwinder at runtime.
8854 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8855 filling in a single size corresponding to each hard register;
8856 @var{address} is the address of the table.
8859 @hook TARGET_ASM_TTYPE
8860 This hook is used to output a reference from a frame unwinding table to
8861 the type_info object identified by @var{sym}. It should return @code{true}
8862 if the reference was output. Returning @code{false} will cause the
8863 reference to be output using the normal Dwarf2 routines.
8866 @hook TARGET_ARM_EABI_UNWINDER
8867 This flag should be set to @code{true} on targets that use an ARM EABI
8868 based unwinding library, and @code{false} on other targets. This effects
8869 the format of unwinding tables, and how the unwinder in entered after
8870 running a cleanup. The default is @code{false}.
8873 @node Alignment Output
8874 @subsection Assembler Commands for Alignment
8876 @c prevent bad page break with this line
8877 This describes commands for alignment.
8879 @defmac JUMP_ALIGN (@var{label})
8880 The alignment (log base 2) to put in front of @var{label}, which is
8881 a common destination of jumps and has no fallthru incoming edge.
8883 This macro need not be defined if you don't want any special alignment
8884 to be done at such a time. Most machine descriptions do not currently
8887 Unless it's necessary to inspect the @var{label} parameter, it is better
8888 to set the variable @var{align_jumps} in the target's
8889 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8890 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8893 @hook TARGET_ASM_JUMP_ALIGN_MAX_SKIP
8894 The maximum number of bytes to skip before @var{label} when applying
8895 @code{JUMP_ALIGN}. This works only if
8896 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8899 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8900 The alignment (log base 2) to put in front of @var{label}, which follows
8903 This macro need not be defined if you don't want any special alignment
8904 to be done at such a time. Most machine descriptions do not currently
8908 @hook TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8909 The maximum number of bytes to skip before @var{label} when applying
8910 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8911 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8914 @defmac LOOP_ALIGN (@var{label})
8915 The alignment (log base 2) to put in front of @var{label}, which follows
8916 a @code{NOTE_INSN_LOOP_BEG} note.
8918 This macro need not be defined if you don't want any special alignment
8919 to be done at such a time. Most machine descriptions do not currently
8922 Unless it's necessary to inspect the @var{label} parameter, it is better
8923 to set the variable @code{align_loops} in the target's
8924 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8925 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8928 @hook TARGET_ASM_LOOP_ALIGN_MAX_SKIP
8929 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
8930 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
8934 @defmac LABEL_ALIGN (@var{label})
8935 The alignment (log base 2) to put in front of @var{label}.
8936 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8937 the maximum of the specified values is used.
8939 Unless it's necessary to inspect the @var{label} parameter, it is better
8940 to set the variable @code{align_labels} in the target's
8941 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8942 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8945 @hook TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8946 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
8947 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
8951 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8952 A C statement to output to the stdio stream @var{stream} an assembler
8953 instruction to advance the location counter by @var{nbytes} bytes.
8954 Those bytes should be zero when loaded. @var{nbytes} will be a C
8955 expression of type @code{unsigned HOST_WIDE_INT}.
8958 @defmac ASM_NO_SKIP_IN_TEXT
8959 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
8960 text section because it fails to put zeros in the bytes that are skipped.
8961 This is true on many Unix systems, where the pseudo--op to skip bytes
8962 produces no-op instructions rather than zeros when used in the text
8966 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
8967 A C statement to output to the stdio stream @var{stream} an assembler
8968 command to advance the location counter to a multiple of 2 to the
8969 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
8972 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
8973 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
8974 for padding, if necessary.
8977 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
8978 A C statement to output to the stdio stream @var{stream} an assembler
8979 command to advance the location counter to a multiple of 2 to the
8980 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
8981 satisfy the alignment request. @var{power} and @var{max_skip} will be
8982 a C expression of type @code{int}.
8986 @node Debugging Info
8987 @section Controlling Debugging Information Format
8989 @c prevent bad page break with this line
8990 This describes how to specify debugging information.
8993 * All Debuggers:: Macros that affect all debugging formats uniformly.
8994 * DBX Options:: Macros enabling specific options in DBX format.
8995 * DBX Hooks:: Hook macros for varying DBX format.
8996 * File Names and DBX:: Macros controlling output of file names in DBX format.
8997 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
8998 * VMS Debug:: Macros for VMS debug format.
9002 @subsection Macros Affecting All Debugging Formats
9004 @c prevent bad page break with this line
9005 These macros affect all debugging formats.
9007 @defmac DBX_REGISTER_NUMBER (@var{regno})
9008 A C expression that returns the DBX register number for the compiler
9009 register number @var{regno}. In the default macro provided, the value
9010 of this expression will be @var{regno} itself. But sometimes there are
9011 some registers that the compiler knows about and DBX does not, or vice
9012 versa. In such cases, some register may need to have one number in the
9013 compiler and another for DBX@.
9015 If two registers have consecutive numbers inside GCC, and they can be
9016 used as a pair to hold a multiword value, then they @emph{must} have
9017 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9018 Otherwise, debuggers will be unable to access such a pair, because they
9019 expect register pairs to be consecutive in their own numbering scheme.
9021 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9022 does not preserve register pairs, then what you must do instead is
9023 redefine the actual register numbering scheme.
9026 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9027 A C expression that returns the integer offset value for an automatic
9028 variable having address @var{x} (an RTL expression). The default
9029 computation assumes that @var{x} is based on the frame-pointer and
9030 gives the offset from the frame-pointer. This is required for targets
9031 that produce debugging output for DBX or COFF-style debugging output
9032 for SDB and allow the frame-pointer to be eliminated when the
9033 @option{-g} options is used.
9036 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9037 A C expression that returns the integer offset value for an argument
9038 having address @var{x} (an RTL expression). The nominal offset is
9042 @defmac PREFERRED_DEBUGGING_TYPE
9043 A C expression that returns the type of debugging output GCC should
9044 produce when the user specifies just @option{-g}. Define
9045 this if you have arranged for GCC to support more than one format of
9046 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9047 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9048 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9050 When the user specifies @option{-ggdb}, GCC normally also uses the
9051 value of this macro to select the debugging output format, but with two
9052 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9053 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9054 defined, GCC uses @code{DBX_DEBUG}.
9056 The value of this macro only affects the default debugging output; the
9057 user can always get a specific type of output by using @option{-gstabs},
9058 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9062 @subsection Specific Options for DBX Output
9064 @c prevent bad page break with this line
9065 These are specific options for DBX output.
9067 @defmac DBX_DEBUGGING_INFO
9068 Define this macro if GCC should produce debugging output for DBX
9069 in response to the @option{-g} option.
9072 @defmac XCOFF_DEBUGGING_INFO
9073 Define this macro if GCC should produce XCOFF format debugging output
9074 in response to the @option{-g} option. This is a variant of DBX format.
9077 @defmac DEFAULT_GDB_EXTENSIONS
9078 Define this macro to control whether GCC should by default generate
9079 GDB's extended version of DBX debugging information (assuming DBX-format
9080 debugging information is enabled at all). If you don't define the
9081 macro, the default is 1: always generate the extended information
9082 if there is any occasion to.
9085 @defmac DEBUG_SYMS_TEXT
9086 Define this macro if all @code{.stabs} commands should be output while
9087 in the text section.
9090 @defmac ASM_STABS_OP
9091 A C string constant, including spacing, naming the assembler pseudo op to
9092 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9093 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9094 applies only to DBX debugging information format.
9097 @defmac ASM_STABD_OP
9098 A C string constant, including spacing, naming the assembler pseudo op to
9099 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9100 value is the current location. If you don't define this macro,
9101 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9105 @defmac ASM_STABN_OP
9106 A C string constant, including spacing, naming the assembler pseudo op to
9107 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9108 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9109 macro applies only to DBX debugging information format.
9112 @defmac DBX_NO_XREFS
9113 Define this macro if DBX on your system does not support the construct
9114 @samp{xs@var{tagname}}. On some systems, this construct is used to
9115 describe a forward reference to a structure named @var{tagname}.
9116 On other systems, this construct is not supported at all.
9119 @defmac DBX_CONTIN_LENGTH
9120 A symbol name in DBX-format debugging information is normally
9121 continued (split into two separate @code{.stabs} directives) when it
9122 exceeds a certain length (by default, 80 characters). On some
9123 operating systems, DBX requires this splitting; on others, splitting
9124 must not be done. You can inhibit splitting by defining this macro
9125 with the value zero. You can override the default splitting-length by
9126 defining this macro as an expression for the length you desire.
9129 @defmac DBX_CONTIN_CHAR
9130 Normally continuation is indicated by adding a @samp{\} character to
9131 the end of a @code{.stabs} string when a continuation follows. To use
9132 a different character instead, define this macro as a character
9133 constant for the character you want to use. Do not define this macro
9134 if backslash is correct for your system.
9137 @defmac DBX_STATIC_STAB_DATA_SECTION
9138 Define this macro if it is necessary to go to the data section before
9139 outputting the @samp{.stabs} pseudo-op for a non-global static
9143 @defmac DBX_TYPE_DECL_STABS_CODE
9144 The value to use in the ``code'' field of the @code{.stabs} directive
9145 for a typedef. The default is @code{N_LSYM}.
9148 @defmac DBX_STATIC_CONST_VAR_CODE
9149 The value to use in the ``code'' field of the @code{.stabs} directive
9150 for a static variable located in the text section. DBX format does not
9151 provide any ``right'' way to do this. The default is @code{N_FUN}.
9154 @defmac DBX_REGPARM_STABS_CODE
9155 The value to use in the ``code'' field of the @code{.stabs} directive
9156 for a parameter passed in registers. DBX format does not provide any
9157 ``right'' way to do this. The default is @code{N_RSYM}.
9160 @defmac DBX_REGPARM_STABS_LETTER
9161 The letter to use in DBX symbol data to identify a symbol as a parameter
9162 passed in registers. DBX format does not customarily provide any way to
9163 do this. The default is @code{'P'}.
9166 @defmac DBX_FUNCTION_FIRST
9167 Define this macro if the DBX information for a function and its
9168 arguments should precede the assembler code for the function. Normally,
9169 in DBX format, the debugging information entirely follows the assembler
9173 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9174 Define this macro, with value 1, if the value of a symbol describing
9175 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9176 relative to the start of the enclosing function. Normally, GCC uses
9177 an absolute address.
9180 @defmac DBX_LINES_FUNCTION_RELATIVE
9181 Define this macro, with value 1, if the value of a symbol indicating
9182 the current line number (@code{N_SLINE}) should be relative to the
9183 start of the enclosing function. Normally, GCC uses an absolute address.
9186 @defmac DBX_USE_BINCL
9187 Define this macro if GCC should generate @code{N_BINCL} and
9188 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9189 macro also directs GCC to output a type number as a pair of a file
9190 number and a type number within the file. Normally, GCC does not
9191 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9192 number for a type number.
9196 @subsection Open-Ended Hooks for DBX Format
9198 @c prevent bad page break with this line
9199 These are hooks for DBX format.
9201 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9202 Define this macro to say how to output to @var{stream} the debugging
9203 information for the start of a scope level for variable names. The
9204 argument @var{name} is the name of an assembler symbol (for use with
9205 @code{assemble_name}) whose value is the address where the scope begins.
9208 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9209 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9212 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9213 Define this macro if the target machine requires special handling to
9214 output an @code{N_FUN} entry for the function @var{decl}.
9217 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9218 A C statement to output DBX debugging information before code for line
9219 number @var{line} of the current source file to the stdio stream
9220 @var{stream}. @var{counter} is the number of time the macro was
9221 invoked, including the current invocation; it is intended to generate
9222 unique labels in the assembly output.
9224 This macro should not be defined if the default output is correct, or
9225 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9228 @defmac NO_DBX_FUNCTION_END
9229 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9230 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9231 On those machines, define this macro to turn this feature off without
9232 disturbing the rest of the gdb extensions.
9235 @defmac NO_DBX_BNSYM_ENSYM
9236 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9237 extension construct. On those machines, define this macro to turn this
9238 feature off without disturbing the rest of the gdb extensions.
9241 @node File Names and DBX
9242 @subsection File Names in DBX Format
9244 @c prevent bad page break with this line
9245 This describes file names in DBX format.
9247 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9248 A C statement to output DBX debugging information to the stdio stream
9249 @var{stream}, which indicates that file @var{name} is the main source
9250 file---the file specified as the input file for compilation.
9251 This macro is called only once, at the beginning of compilation.
9253 This macro need not be defined if the standard form of output
9254 for DBX debugging information is appropriate.
9256 It may be necessary to refer to a label equal to the beginning of the
9257 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9258 to do so. If you do this, you must also set the variable
9259 @var{used_ltext_label_name} to @code{true}.
9262 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9263 Define this macro, with value 1, if GCC should not emit an indication
9264 of the current directory for compilation and current source language at
9265 the beginning of the file.
9268 @defmac NO_DBX_GCC_MARKER
9269 Define this macro, with value 1, if GCC should not emit an indication
9270 that this object file was compiled by GCC@. The default is to emit
9271 an @code{N_OPT} stab at the beginning of every source file, with
9272 @samp{gcc2_compiled.} for the string and value 0.
9275 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9276 A C statement to output DBX debugging information at the end of
9277 compilation of the main source file @var{name}. Output should be
9278 written to the stdio stream @var{stream}.
9280 If you don't define this macro, nothing special is output at the end
9281 of compilation, which is correct for most machines.
9284 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9285 Define this macro @emph{instead of} defining
9286 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9287 the end of compilation is an @code{N_SO} stab with an empty string,
9288 whose value is the highest absolute text address in the file.
9293 @subsection Macros for SDB and DWARF Output
9295 @c prevent bad page break with this line
9296 Here are macros for SDB and DWARF output.
9298 @defmac SDB_DEBUGGING_INFO
9299 Define this macro if GCC should produce COFF-style debugging output
9300 for SDB in response to the @option{-g} option.
9303 @defmac DWARF2_DEBUGGING_INFO
9304 Define this macro if GCC should produce dwarf version 2 format
9305 debugging output in response to the @option{-g} option.
9307 @hook TARGET_DWARF_CALLING_CONVENTION
9308 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9309 be emitted for each function. Instead of an integer return the enum
9310 value for the @code{DW_CC_} tag.
9313 To support optional call frame debugging information, you must also
9314 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9315 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9316 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9317 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9320 @defmac DWARF2_FRAME_INFO
9321 Define this macro to a nonzero value if GCC should always output
9322 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9323 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9324 exceptions are enabled, GCC will output this information not matter
9325 how you define @code{DWARF2_FRAME_INFO}.
9328 @hook TARGET_DEBUG_UNWIND_INFO
9329 This hook defines the mechanism that will be used for describing frame
9330 unwind information to the debugger. Normally the hook will return
9331 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9332 return @code{UI_NONE} otherwise.
9334 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9335 is disabled in order to always output DWARF 2 frame information.
9337 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9338 This will suppress generation of the normal debug frame unwind information.
9341 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9342 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9343 line debug info sections. This will result in much more compact line number
9344 tables, and hence is desirable if it works.
9347 @hook TARGET_WANT_DEBUG_PUB_SECTIONS
9349 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9350 A C statement to issue assembly directives that create a difference
9351 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9354 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9355 A C statement to issue assembly directives that create a difference
9356 between the two given labels in system defined units, e.g. instruction
9357 slots on IA64 VMS, using an integer of the given size.
9360 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9361 A C statement to issue assembly directives that create a
9362 section-relative reference to the given @var{label}, using an integer of the
9363 given @var{size}. The label is known to be defined in the given @var{section}.
9366 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9367 A C statement to issue assembly directives that create a self-relative
9368 reference to the given @var{label}, using an integer of the given @var{size}.
9371 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9372 A C statement to issue assembly directives that create a reference to
9373 the DWARF table identifier @var{label} from the current section. This
9374 is used on some systems to avoid garbage collecting a DWARF table which
9375 is referenced by a function.
9378 @hook TARGET_ASM_OUTPUT_DWARF_DTPREL
9379 If defined, this target hook is a function which outputs a DTP-relative
9380 reference to the given TLS symbol of the specified size.
9383 @defmac PUT_SDB_@dots{}
9384 Define these macros to override the assembler syntax for the special
9385 SDB assembler directives. See @file{sdbout.c} for a list of these
9386 macros and their arguments. If the standard syntax is used, you need
9387 not define them yourself.
9391 Some assemblers do not support a semicolon as a delimiter, even between
9392 SDB assembler directives. In that case, define this macro to be the
9393 delimiter to use (usually @samp{\n}). It is not necessary to define
9394 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9398 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9399 Define this macro to allow references to unknown structure,
9400 union, or enumeration tags to be emitted. Standard COFF does not
9401 allow handling of unknown references, MIPS ECOFF has support for
9405 @defmac SDB_ALLOW_FORWARD_REFERENCES
9406 Define this macro to allow references to structure, union, or
9407 enumeration tags that have not yet been seen to be handled. Some
9408 assemblers choke if forward tags are used, while some require it.
9411 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9412 A C statement to output SDB debugging information before code for line
9413 number @var{line} of the current source file to the stdio stream
9414 @var{stream}. The default is to emit an @code{.ln} directive.
9419 @subsection Macros for VMS Debug Format
9421 @c prevent bad page break with this line
9422 Here are macros for VMS debug format.
9424 @defmac VMS_DEBUGGING_INFO
9425 Define this macro if GCC should produce debugging output for VMS
9426 in response to the @option{-g} option. The default behavior for VMS
9427 is to generate minimal debug info for a traceback in the absence of
9428 @option{-g} unless explicitly overridden with @option{-g0}. This
9429 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9430 @code{TARGET_OPTION_OVERRIDE}.
9433 @node Floating Point
9434 @section Cross Compilation and Floating Point
9435 @cindex cross compilation and floating point
9436 @cindex floating point and cross compilation
9438 While all modern machines use twos-complement representation for integers,
9439 there are a variety of representations for floating point numbers. This
9440 means that in a cross-compiler the representation of floating point numbers
9441 in the compiled program may be different from that used in the machine
9442 doing the compilation.
9444 Because different representation systems may offer different amounts of
9445 range and precision, all floating point constants must be represented in
9446 the target machine's format. Therefore, the cross compiler cannot
9447 safely use the host machine's floating point arithmetic; it must emulate
9448 the target's arithmetic. To ensure consistency, GCC always uses
9449 emulation to work with floating point values, even when the host and
9450 target floating point formats are identical.
9452 The following macros are provided by @file{real.h} for the compiler to
9453 use. All parts of the compiler which generate or optimize
9454 floating-point calculations must use these macros. They may evaluate
9455 their operands more than once, so operands must not have side effects.
9457 @defmac REAL_VALUE_TYPE
9458 The C data type to be used to hold a floating point value in the target
9459 machine's format. Typically this is a @code{struct} containing an
9460 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9464 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9465 Compares for equality the two values, @var{x} and @var{y}. If the target
9466 floating point format supports negative zeroes and/or NaNs,
9467 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9468 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9471 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9472 Tests whether @var{x} is less than @var{y}.
9475 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9476 Truncates @var{x} to a signed integer, rounding toward zero.
9479 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9480 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9481 @var{x} is negative, returns zero.
9484 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9485 Converts @var{string} into a floating point number in the target machine's
9486 representation for mode @var{mode}. This routine can handle both
9487 decimal and hexadecimal floating point constants, using the syntax
9488 defined by the C language for both.
9491 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9492 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9495 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9496 Determines whether @var{x} represents infinity (positive or negative).
9499 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9500 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9503 @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})
9504 Calculates an arithmetic operation on the two floating point values
9505 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9508 The operation to be performed is specified by @var{code}. Only the
9509 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9510 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9512 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9513 target's floating point format cannot represent infinity, it will call
9514 @code{abort}. Callers should check for this situation first, using
9515 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9518 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9519 Returns the negative of the floating point value @var{x}.
9522 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9523 Returns the absolute value of @var{x}.
9526 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9527 Truncates the floating point value @var{x} to fit in @var{mode}. The
9528 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9529 appropriate bit pattern to be output as a floating constant whose
9530 precision accords with mode @var{mode}.
9533 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9534 Converts a floating point value @var{x} into a double-precision integer
9535 which is then stored into @var{low} and @var{high}. If the value is not
9536 integral, it is truncated.
9539 @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})
9540 Converts a double-precision integer found in @var{low} and @var{high},
9541 into a floating point value which is then stored into @var{x}. The
9542 value is truncated to fit in mode @var{mode}.
9545 @node Mode Switching
9546 @section Mode Switching Instructions
9547 @cindex mode switching
9548 The following macros control mode switching optimizations:
9550 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9551 Define this macro if the port needs extra instructions inserted for mode
9552 switching in an optimizing compilation.
9554 For an example, the SH4 can perform both single and double precision
9555 floating point operations, but to perform a single precision operation,
9556 the FPSCR PR bit has to be cleared, while for a double precision
9557 operation, this bit has to be set. Changing the PR bit requires a general
9558 purpose register as a scratch register, hence these FPSCR sets have to
9559 be inserted before reload, i.e.@: you can't put this into instruction emitting
9560 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9562 You can have multiple entities that are mode-switched, and select at run time
9563 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9564 return nonzero for any @var{entity} that needs mode-switching.
9565 If you define this macro, you also have to define
9566 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9567 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9568 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9572 @defmac NUM_MODES_FOR_MODE_SWITCHING
9573 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9574 initializer for an array of integers. Each initializer element
9575 N refers to an entity that needs mode switching, and specifies the number
9576 of different modes that might need to be set for this entity.
9577 The position of the initializer in the initializer---starting counting at
9578 zero---determines the integer that is used to refer to the mode-switched
9580 In macros that take mode arguments / yield a mode result, modes are
9581 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9582 switch is needed / supplied.
9585 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9586 @var{entity} is an integer specifying a mode-switched entity. If
9587 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9588 return an integer value not larger than the corresponding element in
9589 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9590 be switched into prior to the execution of @var{insn}.
9593 @defmac MODE_AFTER (@var{mode}, @var{insn})
9594 If this macro is defined, it is evaluated for every @var{insn} during
9595 mode switching. It determines the mode that an insn results in (if
9596 different from the incoming mode).
9599 @defmac MODE_ENTRY (@var{entity})
9600 If this macro is defined, it is evaluated for every @var{entity} that needs
9601 mode switching. It should evaluate to an integer, which is a mode that
9602 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9603 is defined then @code{MODE_EXIT} must be defined.
9606 @defmac MODE_EXIT (@var{entity})
9607 If this macro is defined, it is evaluated for every @var{entity} that needs
9608 mode switching. It should evaluate to an integer, which is a mode that
9609 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9610 is defined then @code{MODE_ENTRY} must be defined.
9613 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9614 This macro specifies the order in which modes for @var{entity} are processed.
9615 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9616 lowest. The value of the macro should be an integer designating a mode
9617 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9618 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9619 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9622 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9623 Generate one or more insns to set @var{entity} to @var{mode}.
9624 @var{hard_reg_live} is the set of hard registers live at the point where
9625 the insn(s) are to be inserted.
9628 @node Target Attributes
9629 @section Defining target-specific uses of @code{__attribute__}
9630 @cindex target attributes
9631 @cindex machine attributes
9632 @cindex attributes, target-specific
9634 Target-specific attributes may be defined for functions, data and types.
9635 These are described using the following target hooks; they also need to
9636 be documented in @file{extend.texi}.
9638 @hook TARGET_ATTRIBUTE_TABLE
9639 If defined, this target hook points to an array of @samp{struct
9640 attribute_spec} (defined in @file{tree.h}) specifying the machine
9641 specific attributes for this target and some of the restrictions on the
9642 entities to which these attributes are applied and the arguments they
9646 @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
9647 If defined, this target hook is a function which returns true if the
9648 machine-specific attribute named @var{name} expects an identifier
9649 given as its first argument to be passed on as a plain identifier, not
9650 subjected to name lookup. If this is not defined, the default is
9651 false for all machine-specific attributes.
9654 @hook TARGET_COMP_TYPE_ATTRIBUTES
9655 If defined, this target hook is a function which returns zero if the attributes on
9656 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9657 and two if they are nearly compatible (which causes a warning to be
9658 generated). If this is not defined, machine-specific attributes are
9659 supposed always to be compatible.
9662 @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
9663 If defined, this target hook is a function which assigns default attributes to
9664 the newly defined @var{type}.
9667 @hook TARGET_MERGE_TYPE_ATTRIBUTES
9668 Define this target hook if the merging of type attributes needs special
9669 handling. If defined, the result is a list of the combined
9670 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9671 that @code{comptypes} has already been called and returned 1. This
9672 function may call @code{merge_attributes} to handle machine-independent
9676 @hook TARGET_MERGE_DECL_ATTRIBUTES
9677 Define this target hook if the merging of decl attributes needs special
9678 handling. If defined, the result is a list of the combined
9679 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9680 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9681 when this is needed are when one attribute overrides another, or when an
9682 attribute is nullified by a subsequent definition. This function may
9683 call @code{merge_attributes} to handle machine-independent merging.
9685 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9686 If the only target-specific handling you require is @samp{dllimport}
9687 for Microsoft Windows targets, you should define the macro
9688 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9689 will then define a function called
9690 @code{merge_dllimport_decl_attributes} which can then be defined as
9691 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9692 add @code{handle_dll_attribute} in the attribute table for your port
9693 to perform initial processing of the @samp{dllimport} and
9694 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9695 @file{i386/i386.c}, for example.
9698 @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
9700 @defmac TARGET_DECLSPEC
9701 Define this macro to a nonzero value if you want to treat
9702 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9703 default, this behavior is enabled only for targets that define
9704 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9705 of @code{__declspec} is via a built-in macro, but you should not rely
9706 on this implementation detail.
9709 @hook TARGET_INSERT_ATTRIBUTES
9710 Define this target hook if you want to be able to add attributes to a decl
9711 when it is being created. This is normally useful for back ends which
9712 wish to implement a pragma by using the attributes which correspond to
9713 the pragma's effect. The @var{node} argument is the decl which is being
9714 created. The @var{attr_ptr} argument is a pointer to the attribute list
9715 for this decl. The list itself should not be modified, since it may be
9716 shared with other decls, but attributes may be chained on the head of
9717 the list and @code{*@var{attr_ptr}} modified to point to the new
9718 attributes, or a copy of the list may be made if further changes are
9722 @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
9724 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9725 into the current function, despite its having target-specific
9726 attributes, @code{false} otherwise. By default, if a function has a
9727 target specific attribute attached to it, it will not be inlined.
9730 @hook TARGET_OPTION_VALID_ATTRIBUTE_P
9731 This hook is called to parse the @code{attribute(option("..."))}, and
9732 it allows the function to set different target machine compile time
9733 options for the current function that might be different than the
9734 options specified on the command line. The hook should return
9735 @code{true} if the options are valid.
9737 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9738 the function declaration to hold a pointer to a target specific
9739 @var{struct cl_target_option} structure.
9742 @hook TARGET_OPTION_SAVE
9743 This hook is called to save any additional target specific information
9744 in the @var{struct cl_target_option} structure for function specific
9746 @xref{Option file format}.
9749 @hook TARGET_OPTION_RESTORE
9750 This hook is called to restore any additional target specific
9751 information in the @var{struct cl_target_option} structure for
9752 function specific options.
9755 @hook TARGET_OPTION_PRINT
9756 This hook is called to print any additional target specific
9757 information in the @var{struct cl_target_option} structure for
9758 function specific options.
9761 @hook TARGET_OPTION_PRAGMA_PARSE
9762 This target hook parses the options for @code{#pragma GCC option} to
9763 set the machine specific options for functions that occur later in the
9764 input stream. The options should be the same as handled by the
9765 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9768 @hook TARGET_OPTION_OVERRIDE
9769 Sometimes certain combinations of command options do not make sense on
9770 a particular target machine. You can override the hook
9771 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9772 once just after all the command options have been parsed.
9774 Don't use this hook to turn on various extra optimizations for
9775 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9777 If you need to do something whenever the optimization level is
9778 changed via the optimize attribute or pragma, see
9779 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9782 @hook TARGET_CAN_INLINE_P
9783 This target hook returns @code{false} if the @var{caller} function
9784 cannot inline @var{callee}, based on target specific information. By
9785 default, inlining is not allowed if the callee function has function
9786 specific target options and the caller does not use the same options.
9790 @section Emulating TLS
9791 @cindex Emulated TLS
9793 For targets whose psABI does not provide Thread Local Storage via
9794 specific relocations and instruction sequences, an emulation layer is
9795 used. A set of target hooks allows this emulation layer to be
9796 configured for the requirements of a particular target. For instance
9797 the psABI may in fact specify TLS support in terms of an emulation
9800 The emulation layer works by creating a control object for every TLS
9801 object. To access the TLS object, a lookup function is provided
9802 which, when given the address of the control object, will return the
9803 address of the current thread's instance of the TLS object.
9805 @hook TARGET_EMUTLS_GET_ADDRESS
9806 Contains the name of the helper function that uses a TLS control
9807 object to locate a TLS instance. The default causes libgcc's
9808 emulated TLS helper function to be used.
9811 @hook TARGET_EMUTLS_REGISTER_COMMON
9812 Contains the name of the helper function that should be used at
9813 program startup to register TLS objects that are implicitly
9814 initialized to zero. If this is @code{NULL}, all TLS objects will
9815 have explicit initializers. The default causes libgcc's emulated TLS
9816 registration function to be used.
9819 @hook TARGET_EMUTLS_VAR_SECTION
9820 Contains the name of the section in which TLS control variables should
9821 be placed. The default of @code{NULL} allows these to be placed in
9825 @hook TARGET_EMUTLS_TMPL_SECTION
9826 Contains the name of the section in which TLS initializers should be
9827 placed. The default of @code{NULL} allows these to be placed in any
9831 @hook TARGET_EMUTLS_VAR_PREFIX
9832 Contains the prefix to be prepended to TLS control variable names.
9833 The default of @code{NULL} uses a target-specific prefix.
9836 @hook TARGET_EMUTLS_TMPL_PREFIX
9837 Contains the prefix to be prepended to TLS initializer objects. The
9838 default of @code{NULL} uses a target-specific prefix.
9841 @hook TARGET_EMUTLS_VAR_FIELDS
9842 Specifies a function that generates the FIELD_DECLs for a TLS control
9843 object type. @var{type} is the RECORD_TYPE the fields are for and
9844 @var{name} should be filled with the structure tag, if the default of
9845 @code{__emutls_object} is unsuitable. The default creates a type suitable
9846 for libgcc's emulated TLS function.
9849 @hook TARGET_EMUTLS_VAR_INIT
9850 Specifies a function that generates the CONSTRUCTOR to initialize a
9851 TLS control object. @var{var} is the TLS control object, @var{decl}
9852 is the TLS object and @var{tmpl_addr} is the address of the
9853 initializer. The default initializes libgcc's emulated TLS control object.
9856 @hook TARGET_EMUTLS_VAR_ALIGN_FIXED
9857 Specifies whether the alignment of TLS control variable objects is
9858 fixed and should not be increased as some backends may do to optimize
9859 single objects. The default is false.
9862 @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9863 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9864 may be used to describe emulated TLS control objects.
9867 @node MIPS Coprocessors
9868 @section Defining coprocessor specifics for MIPS targets.
9869 @cindex MIPS coprocessor-definition macros
9871 The MIPS specification allows MIPS implementations to have as many as 4
9872 coprocessors, each with as many as 32 private registers. GCC supports
9873 accessing these registers and transferring values between the registers
9874 and memory using asm-ized variables. For example:
9877 register unsigned int cp0count asm ("c0r1");
9883 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9884 names may be added as described below, or the default names may be
9885 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9887 Coprocessor registers are assumed to be epilogue-used; sets to them will
9888 be preserved even if it does not appear that the register is used again
9889 later in the function.
9891 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9892 the FPU@. One accesses COP1 registers through standard mips
9893 floating-point support; they are not included in this mechanism.
9895 There is one macro used in defining the MIPS coprocessor interface which
9896 you may want to override in subtargets; it is described below.
9898 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9899 A comma-separated list (with leading comma) of pairs describing the
9900 alternate names of coprocessor registers. The format of each entry should be
9902 @{ @var{alternatename}, @var{register_number}@}
9908 @section Parameters for Precompiled Header Validity Checking
9909 @cindex parameters, precompiled headers
9911 @hook TARGET_GET_PCH_VALIDITY
9912 This hook returns a pointer to the data needed by
9913 @code{TARGET_PCH_VALID_P} and sets
9914 @samp{*@var{sz}} to the size of the data in bytes.
9917 @hook TARGET_PCH_VALID_P
9918 This hook checks whether the options used to create a PCH file are
9919 compatible with the current settings. It returns @code{NULL}
9920 if so and a suitable error message if not. Error messages will
9921 be presented to the user and must be localized using @samp{_(@var{msg})}.
9923 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9924 when the PCH file was created and @var{sz} is the size of that data in bytes.
9925 It's safe to assume that the data was created by the same version of the
9926 compiler, so no format checking is needed.
9928 The default definition of @code{default_pch_valid_p} should be
9929 suitable for most targets.
9932 @hook TARGET_CHECK_PCH_TARGET_FLAGS
9933 If this hook is nonnull, the default implementation of
9934 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9935 of @code{target_flags}. @var{pch_flags} specifies the value that
9936 @code{target_flags} had when the PCH file was created. The return
9937 value is the same as for @code{TARGET_PCH_VALID_P}.
9941 @section C++ ABI parameters
9942 @cindex parameters, c++ abi
9944 @hook TARGET_CXX_GUARD_TYPE
9945 Define this hook to override the integer type used for guard variables.
9946 These are used to implement one-time construction of static objects. The
9947 default is long_long_integer_type_node.
9950 @hook TARGET_CXX_GUARD_MASK_BIT
9951 This hook determines how guard variables are used. It should return
9952 @code{false} (the default) if the first byte should be used. A return value of
9953 @code{true} indicates that only the least significant bit should be used.
9956 @hook TARGET_CXX_GET_COOKIE_SIZE
9957 This hook returns the size of the cookie to use when allocating an array
9958 whose elements have the indicated @var{type}. Assumes that it is already
9959 known that a cookie is needed. The default is
9960 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
9961 IA64/Generic C++ ABI@.
9964 @hook TARGET_CXX_COOKIE_HAS_SIZE
9965 This hook should return @code{true} if the element size should be stored in
9966 array cookies. The default is to return @code{false}.
9969 @hook TARGET_CXX_IMPORT_EXPORT_CLASS
9970 If defined by a backend this hook allows the decision made to export
9971 class @var{type} to be overruled. Upon entry @var{import_export}
9972 will contain 1 if the class is going to be exported, @minus{}1 if it is going
9973 to be imported and 0 otherwise. This function should return the
9974 modified value and perform any other actions necessary to support the
9975 backend's targeted operating system.
9978 @hook TARGET_CXX_CDTOR_RETURNS_THIS
9979 This hook should return @code{true} if constructors and destructors return
9980 the address of the object created/destroyed. The default is to return
9984 @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
9985 This hook returns true if the key method for a class (i.e., the method
9986 which, if defined in the current translation unit, causes the virtual
9987 table to be emitted) may be an inline function. Under the standard
9988 Itanium C++ ABI the key method may be an inline function so long as
9989 the function is not declared inline in the class definition. Under
9990 some variants of the ABI, an inline function can never be the key
9991 method. The default is to return @code{true}.
9994 @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
9996 @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
9997 This hook returns true (the default) if virtual tables and other
9998 similar implicit class data objects are always COMDAT if they have
9999 external linkage. If this hook returns false, then class data for
10000 classes whose virtual table will be emitted in only one translation
10001 unit will not be COMDAT.
10004 @hook TARGET_CXX_LIBRARY_RTTI_COMDAT
10005 This hook returns true (the default) if the RTTI information for
10006 the basic types which is defined in the C++ runtime should always
10007 be COMDAT, false if it should not be COMDAT.
10010 @hook TARGET_CXX_USE_AEABI_ATEXIT
10011 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10012 should be used to register static destructors when @option{-fuse-cxa-atexit}
10013 is in effect. The default is to return false to use @code{__cxa_atexit}.
10016 @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
10017 This hook returns true if the target @code{atexit} function can be used
10018 in the same manner as @code{__cxa_atexit} to register C++ static
10019 destructors. This requires that @code{atexit}-registered functions in
10020 shared libraries are run in the correct order when the libraries are
10021 unloaded. The default is to return false.
10024 @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
10026 @node Named Address Spaces
10027 @section Adding support for named address spaces
10028 @cindex named address spaces
10030 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10031 standards committee, @cite{Programming Languages - C - Extensions to
10032 support embedded processors}, specifies a syntax for embedded
10033 processors to specify alternate address spaces. You can configure a
10034 GCC port to support section 5.1 of the draft report to add support for
10035 address spaces other than the default address space. These address
10036 spaces are new keywords that are similar to the @code{volatile} and
10037 @code{const} type attributes.
10039 Pointers to named address spaces can have a different size than
10040 pointers to the generic address space.
10042 For example, the SPU port uses the @code{__ea} address space to refer
10043 to memory in the host processor, rather than memory local to the SPU
10044 processor. Access to memory in the @code{__ea} address space involves
10045 issuing DMA operations to move data between the host processor and the
10046 local processor memory address space. Pointers in the @code{__ea}
10047 address space are either 32 bits or 64 bits based on the
10048 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10051 Internally, address spaces are represented as a small integer in the
10052 range 0 to 15 with address space 0 being reserved for the generic
10055 To register a named address space qualifier keyword with the C front end,
10056 the target may call the @code{c_register_addr_space} routine. For example,
10057 the SPU port uses the following to declare @code{__ea} as the keyword for
10058 named address space #1:
10060 #define ADDR_SPACE_EA 1
10061 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10064 @hook TARGET_ADDR_SPACE_POINTER_MODE
10065 Define this to return the machine mode to use for pointers to
10066 @var{address_space} if the target supports named address spaces.
10067 The default version of this hook returns @code{ptr_mode} for the
10068 generic address space only.
10071 @hook TARGET_ADDR_SPACE_ADDRESS_MODE
10072 Define this to return the machine mode to use for addresses in
10073 @var{address_space} if the target supports named address spaces.
10074 The default version of this hook returns @code{Pmode} for the
10075 generic address space only.
10078 @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
10079 Define this to return nonzero if the port can handle pointers
10080 with machine mode @var{mode} to address space @var{as}. This target
10081 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10082 except that it includes explicit named address space support. The default
10083 version of this hook returns true for the modes returned by either the
10084 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10085 target hooks for the given address space.
10088 @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
10089 Define this to return true if @var{exp} is a valid address for mode
10090 @var{mode} in the named address space @var{as}. The @var{strict}
10091 parameter says whether strict addressing is in effect after reload has
10092 finished. This target hook is the same as the
10093 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10094 explicit named address space support.
10097 @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
10098 Define this to modify an invalid address @var{x} to be a valid address
10099 with mode @var{mode} in the named address space @var{as}. This target
10100 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10101 except that it includes explicit named address space support.
10104 @hook TARGET_ADDR_SPACE_SUBSET_P
10105 Define this to return whether the @var{subset} named address space is
10106 contained within the @var{superset} named address space. Pointers to
10107 a named address space that is a subset of another named address space
10108 will be converted automatically without a cast if used together in
10109 arithmetic operations. Pointers to a superset address space can be
10110 converted to pointers to a subset address space via explicit casts.
10113 @hook TARGET_ADDR_SPACE_CONVERT
10114 Define this to convert the pointer expression represented by the RTL
10115 @var{op} with type @var{from_type} that points to a named address
10116 space to a new pointer expression with type @var{to_type} that points
10117 to a different named address space. When this hook it called, it is
10118 guaranteed that one of the two address spaces is a subset of the other,
10119 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10123 @section Miscellaneous Parameters
10124 @cindex parameters, miscellaneous
10126 @c prevent bad page break with this line
10127 Here are several miscellaneous parameters.
10129 @defmac HAS_LONG_COND_BRANCH
10130 Define this boolean macro to indicate whether or not your architecture
10131 has conditional branches that can span all of memory. It is used in
10132 conjunction with an optimization that partitions hot and cold basic
10133 blocks into separate sections of the executable. If this macro is
10134 set to false, gcc will convert any conditional branches that attempt
10135 to cross between sections into unconditional branches or indirect jumps.
10138 @defmac HAS_LONG_UNCOND_BRANCH
10139 Define this boolean macro to indicate whether or not your architecture
10140 has unconditional branches that can span all of memory. It is used in
10141 conjunction with an optimization that partitions hot and cold basic
10142 blocks into separate sections of the executable. If this macro is
10143 set to false, gcc will convert any unconditional branches that attempt
10144 to cross between sections into indirect jumps.
10147 @defmac CASE_VECTOR_MODE
10148 An alias for a machine mode name. This is the machine mode that
10149 elements of a jump-table should have.
10152 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10153 Optional: return the preferred mode for an @code{addr_diff_vec}
10154 when the minimum and maximum offset are known. If you define this,
10155 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10156 To make this work, you also have to define @code{INSN_ALIGN} and
10157 make the alignment for @code{addr_diff_vec} explicit.
10158 The @var{body} argument is provided so that the offset_unsigned and scale
10159 flags can be updated.
10162 @defmac CASE_VECTOR_PC_RELATIVE
10163 Define this macro to be a C expression to indicate when jump-tables
10164 should contain relative addresses. You need not define this macro if
10165 jump-tables never contain relative addresses, or jump-tables should
10166 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10170 @hook TARGET_CASE_VALUES_THRESHOLD
10171 This function return the smallest number of different values for which it
10172 is best to use a jump-table instead of a tree of conditional branches.
10173 The default is four for machines with a @code{casesi} instruction and
10174 five otherwise. This is best for most machines.
10177 @defmac CASE_USE_BIT_TESTS
10178 Define this macro to be a C expression to indicate whether C switch
10179 statements may be implemented by a sequence of bit tests. This is
10180 advantageous on processors that can efficiently implement left shift
10181 of 1 by the number of bits held in a register, but inappropriate on
10182 targets that would require a loop. By default, this macro returns
10183 @code{true} if the target defines an @code{ashlsi3} pattern, and
10184 @code{false} otherwise.
10187 @defmac WORD_REGISTER_OPERATIONS
10188 Define this macro if operations between registers with integral mode
10189 smaller than a word are always performed on the entire register.
10190 Most RISC machines have this property and most CISC machines do not.
10193 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10194 Define this macro to be a C expression indicating when insns that read
10195 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10196 bits outside of @var{mem_mode} to be either the sign-extension or the
10197 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10198 of @var{mem_mode} for which the
10199 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10200 @code{UNKNOWN} for other modes.
10202 This macro is not called with @var{mem_mode} non-integral or with a width
10203 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10204 value in this case. Do not define this macro if it would always return
10205 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10206 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10208 You may return a non-@code{UNKNOWN} value even if for some hard registers
10209 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10210 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10211 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10212 integral mode larger than this but not larger than @code{word_mode}.
10214 You must return @code{UNKNOWN} if for some hard registers that allow this
10215 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10216 @code{word_mode}, but that they can change to another integral mode that
10217 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10220 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10221 Define this macro if loading short immediate values into registers sign
10225 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
10226 Define this macro if the same instructions that convert a floating
10227 point number to a signed fixed point number also convert validly to an
10231 @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
10232 When @option{-ffast-math} is in effect, GCC tries to optimize
10233 divisions by the same divisor, by turning them into multiplications by
10234 the reciprocal. This target hook specifies the minimum number of divisions
10235 that should be there for GCC to perform the optimization for a variable
10236 of mode @var{mode}. The default implementation returns 3 if the machine
10237 has an instruction for the division, and 2 if it does not.
10241 The maximum number of bytes that a single instruction can move quickly
10242 between memory and registers or between two memory locations.
10245 @defmac MAX_MOVE_MAX
10246 The maximum number of bytes that a single instruction can move quickly
10247 between memory and registers or between two memory locations. If this
10248 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10249 constant value that is the largest value that @code{MOVE_MAX} can have
10253 @defmac SHIFT_COUNT_TRUNCATED
10254 A C expression that is nonzero if on this machine the number of bits
10255 actually used for the count of a shift operation is equal to the number
10256 of bits needed to represent the size of the object being shifted. When
10257 this macro is nonzero, the compiler will assume that it is safe to omit
10258 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10259 truncates the count of a shift operation. On machines that have
10260 instructions that act on bit-fields at variable positions, which may
10261 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10262 also enables deletion of truncations of the values that serve as
10263 arguments to bit-field instructions.
10265 If both types of instructions truncate the count (for shifts) and
10266 position (for bit-field operations), or if no variable-position bit-field
10267 instructions exist, you should define this macro.
10269 However, on some machines, such as the 80386 and the 680x0, truncation
10270 only applies to shift operations and not the (real or pretended)
10271 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10272 such machines. Instead, add patterns to the @file{md} file that include
10273 the implied truncation of the shift instructions.
10275 You need not define this macro if it would always have the value of zero.
10278 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10279 @hook TARGET_SHIFT_TRUNCATION_MASK
10280 This function describes how the standard shift patterns for @var{mode}
10281 deal with shifts by negative amounts or by more than the width of the mode.
10282 @xref{shift patterns}.
10284 On many machines, the shift patterns will apply a mask @var{m} to the
10285 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10286 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10287 this is true for mode @var{mode}, the function should return @var{m},
10288 otherwise it should return 0. A return value of 0 indicates that no
10289 particular behavior is guaranteed.
10291 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10292 @emph{not} apply to general shift rtxes; it applies only to instructions
10293 that are generated by the named shift patterns.
10295 The default implementation of this function returns
10296 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10297 and 0 otherwise. This definition is always safe, but if
10298 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10299 nevertheless truncate the shift count, you may get better code
10303 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10304 A C expression which is nonzero if on this machine it is safe to
10305 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10306 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10307 operating on it as if it had only @var{outprec} bits.
10309 On many machines, this expression can be 1.
10311 @c rearranged this, removed the phrase "it is reported that". this was
10312 @c to fix an overfull hbox. --mew 10feb93
10313 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10314 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10315 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10316 such cases may improve things.
10319 @hook TARGET_MODE_REP_EXTENDED
10320 The representation of an integral mode can be such that the values
10321 are always extended to a wider integral mode. Return
10322 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10323 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10324 otherwise. (Currently, none of the targets use zero-extended
10325 representation this way so unlike @code{LOAD_EXTEND_OP},
10326 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10327 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10328 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10329 widest integral mode and currently we take advantage of this fact.)
10331 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10332 value even if the extension is not performed on certain hard registers
10333 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10334 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10336 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10337 describe two related properties. If you define
10338 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10339 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10342 In order to enforce the representation of @code{mode},
10343 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10347 @defmac STORE_FLAG_VALUE
10348 A C expression describing the value returned by a comparison operator
10349 with an integral mode and stored by a store-flag instruction
10350 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10351 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10352 comparison operators whose results have a @code{MODE_INT} mode.
10354 A value of 1 or @minus{}1 means that the instruction implementing the
10355 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10356 and 0 when the comparison is false. Otherwise, the value indicates
10357 which bits of the result are guaranteed to be 1 when the comparison is
10358 true. This value is interpreted in the mode of the comparison
10359 operation, which is given by the mode of the first operand in the
10360 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10361 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10364 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10365 generate code that depends only on the specified bits. It can also
10366 replace comparison operators with equivalent operations if they cause
10367 the required bits to be set, even if the remaining bits are undefined.
10368 For example, on a machine whose comparison operators return an
10369 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10370 @samp{0x80000000}, saying that just the sign bit is relevant, the
10374 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10378 can be converted to
10381 (ashift:SI @var{x} (const_int @var{n}))
10385 where @var{n} is the appropriate shift count to move the bit being
10386 tested into the sign bit.
10388 There is no way to describe a machine that always sets the low-order bit
10389 for a true value, but does not guarantee the value of any other bits,
10390 but we do not know of any machine that has such an instruction. If you
10391 are trying to port GCC to such a machine, include an instruction to
10392 perform a logical-and of the result with 1 in the pattern for the
10393 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10395 Often, a machine will have multiple instructions that obtain a value
10396 from a comparison (or the condition codes). Here are rules to guide the
10397 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10402 Use the shortest sequence that yields a valid definition for
10403 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10404 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10405 comparison operators to do so because there may be opportunities to
10406 combine the normalization with other operations.
10409 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10410 slightly preferred on machines with expensive jumps and 1 preferred on
10414 As a second choice, choose a value of @samp{0x80000001} if instructions
10415 exist that set both the sign and low-order bits but do not define the
10419 Otherwise, use a value of @samp{0x80000000}.
10422 Many machines can produce both the value chosen for
10423 @code{STORE_FLAG_VALUE} and its negation in the same number of
10424 instructions. On those machines, you should also define a pattern for
10425 those cases, e.g., one matching
10428 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10431 Some machines can also perform @code{and} or @code{plus} operations on
10432 condition code values with less instructions than the corresponding
10433 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10434 machines, define the appropriate patterns. Use the names @code{incscc}
10435 and @code{decscc}, respectively, for the patterns which perform
10436 @code{plus} or @code{minus} operations on condition code values. See
10437 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10438 find such instruction sequences on other machines.
10440 If this macro is not defined, the default value, 1, is used. You need
10441 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10442 instructions, or if the value generated by these instructions is 1.
10445 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10446 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10447 returned when comparison operators with floating-point results are true.
10448 Define this macro on machines that have comparison operations that return
10449 floating-point values. If there are no such operations, do not define
10453 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10454 A C expression that gives a rtx representing the nonzero true element
10455 for vector comparisons. The returned rtx should be valid for the inner
10456 mode of @var{mode} which is guaranteed to be a vector mode. Define
10457 this macro on machines that have vector comparison operations that
10458 return a vector result. If there are no such operations, do not define
10459 this macro. Typically, this macro is defined as @code{const1_rtx} or
10460 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10461 the compiler optimizing such vector comparison operations for the
10465 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10466 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10467 A C expression that indicates whether the architecture defines a value
10468 for @code{clz} or @code{ctz} with a zero operand.
10469 A result of @code{0} indicates the value is undefined.
10470 If the value is defined for only the RTL expression, the macro should
10471 evaluate to @code{1}; if the value applies also to the corresponding optab
10472 entry (which is normally the case if it expands directly into
10473 the corresponding RTL), then the macro should evaluate to @code{2}.
10474 In the cases where the value is defined, @var{value} should be set to
10477 If this macro is not defined, the value of @code{clz} or
10478 @code{ctz} at zero is assumed to be undefined.
10480 This macro must be defined if the target's expansion for @code{ffs}
10481 relies on a particular value to get correct results. Otherwise it
10482 is not necessary, though it may be used to optimize some corner cases, and
10483 to provide a default expansion for the @code{ffs} optab.
10485 Note that regardless of this macro the ``definedness'' of @code{clz}
10486 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10487 visible to the user. Thus one may be free to adjust the value at will
10488 to match the target expansion of these operations without fear of
10493 An alias for the machine mode for pointers. On most machines, define
10494 this to be the integer mode corresponding to the width of a hardware
10495 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10496 On some machines you must define this to be one of the partial integer
10497 modes, such as @code{PSImode}.
10499 The width of @code{Pmode} must be at least as large as the value of
10500 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10501 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10505 @defmac FUNCTION_MODE
10506 An alias for the machine mode used for memory references to functions
10507 being called, in @code{call} RTL expressions. On most CISC machines,
10508 where an instruction can begin at any byte address, this should be
10509 @code{QImode}. On most RISC machines, where all instructions have fixed
10510 size and alignment, this should be a mode with the same size and alignment
10511 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10514 @defmac STDC_0_IN_SYSTEM_HEADERS
10515 In normal operation, the preprocessor expands @code{__STDC__} to the
10516 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10517 hosts, like Solaris, the system compiler uses a different convention,
10518 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10519 strict conformance to the C Standard.
10521 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10522 convention when processing system header files, but when processing user
10523 files @code{__STDC__} will always expand to 1.
10526 @defmac NO_IMPLICIT_EXTERN_C
10527 Define this macro if the system header files support C++ as well as C@.
10528 This macro inhibits the usual method of using system header files in
10529 C++, which is to pretend that the file's contents are enclosed in
10530 @samp{extern "C" @{@dots{}@}}.
10535 @defmac REGISTER_TARGET_PRAGMAS ()
10536 Define this macro if you want to implement any target-specific pragmas.
10537 If defined, it is a C expression which makes a series of calls to
10538 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10539 for each pragma. The macro may also do any
10540 setup required for the pragmas.
10542 The primary reason to define this macro is to provide compatibility with
10543 other compilers for the same target. In general, we discourage
10544 definition of target-specific pragmas for GCC@.
10546 If the pragma can be implemented by attributes then you should consider
10547 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10549 Preprocessor macros that appear on pragma lines are not expanded. All
10550 @samp{#pragma} directives that do not match any registered pragma are
10551 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10554 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10555 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10557 Each call to @code{c_register_pragma} or
10558 @code{c_register_pragma_with_expansion} establishes one pragma. The
10559 @var{callback} routine will be called when the preprocessor encounters a
10563 #pragma [@var{space}] @var{name} @dots{}
10566 @var{space} is the case-sensitive namespace of the pragma, or
10567 @code{NULL} to put the pragma in the global namespace. The callback
10568 routine receives @var{pfile} as its first argument, which can be passed
10569 on to cpplib's functions if necessary. You can lex tokens after the
10570 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10571 callback will be silently ignored. The end of the line is indicated by
10572 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10573 arguments of pragmas registered with
10574 @code{c_register_pragma_with_expansion} but not on the arguments of
10575 pragmas registered with @code{c_register_pragma}.
10577 Note that the use of @code{pragma_lex} is specific to the C and C++
10578 compilers. It will not work in the Java or Fortran compilers, or any
10579 other language compilers for that matter. Thus if @code{pragma_lex} is going
10580 to be called from target-specific code, it must only be done so when
10581 building the C and C++ compilers. This can be done by defining the
10582 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10583 target entry in the @file{config.gcc} file. These variables should name
10584 the target-specific, language-specific object file which contains the
10585 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10586 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10587 how to build this object file.
10590 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10591 Define this macro if macros should be expanded in the
10592 arguments of @samp{#pragma pack}.
10595 @hook TARGET_HANDLE_PRAGMA_EXTERN_PREFIX
10597 @defmac TARGET_DEFAULT_PACK_STRUCT
10598 If your target requires a structure packing default other than 0 (meaning
10599 the machine default), define this macro to the necessary value (in bytes).
10600 This must be a value that would also be valid to use with
10601 @samp{#pragma pack()} (that is, a small power of two).
10604 @defmac DOLLARS_IN_IDENTIFIERS
10605 Define this macro to control use of the character @samp{$} in
10606 identifier names for the C family of languages. 0 means @samp{$} is
10607 not allowed by default; 1 means it is allowed. 1 is the default;
10608 there is no need to define this macro in that case.
10611 @defmac NO_DOLLAR_IN_LABEL
10612 Define this macro if the assembler does not accept the character
10613 @samp{$} in label names. By default constructors and destructors in
10614 G++ have @samp{$} in the identifiers. If this macro is defined,
10615 @samp{.} is used instead.
10618 @defmac NO_DOT_IN_LABEL
10619 Define this macro if the assembler does not accept the character
10620 @samp{.} in label names. By default constructors and destructors in G++
10621 have names that use @samp{.}. If this macro is defined, these names
10622 are rewritten to avoid @samp{.}.
10625 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10626 Define this macro as a C expression that is nonzero if it is safe for the
10627 delay slot scheduler to place instructions in the delay slot of @var{insn},
10628 even if they appear to use a resource set or clobbered in @var{insn}.
10629 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10630 every @code{call_insn} has this behavior. On machines where some @code{insn}
10631 or @code{jump_insn} is really a function call and hence has this behavior,
10632 you should define this macro.
10634 You need not define this macro if it would always return zero.
10637 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10638 Define this macro as a C expression that is nonzero if it is safe for the
10639 delay slot scheduler to place instructions in the delay slot of @var{insn},
10640 even if they appear to set or clobber a resource referenced in @var{insn}.
10641 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10642 some @code{insn} or @code{jump_insn} is really a function call and its operands
10643 are registers whose use is actually in the subroutine it calls, you should
10644 define this macro. Doing so allows the delay slot scheduler to move
10645 instructions which copy arguments into the argument registers into the delay
10646 slot of @var{insn}.
10648 You need not define this macro if it would always return zero.
10651 @defmac MULTIPLE_SYMBOL_SPACES
10652 Define this macro as a C expression that is nonzero if, in some cases,
10653 global symbols from one translation unit may not be bound to undefined
10654 symbols in another translation unit without user intervention. For
10655 instance, under Microsoft Windows symbols must be explicitly imported
10656 from shared libraries (DLLs).
10658 You need not define this macro if it would always evaluate to zero.
10661 @hook TARGET_MD_ASM_CLOBBERS
10662 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10663 any hard regs the port wishes to automatically clobber for an asm.
10664 It should return the result of the last @code{tree_cons} used to add a
10665 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10666 corresponding parameters to the asm and may be inspected to avoid
10667 clobbering a register that is an input or output of the asm. You can use
10668 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10669 for overlap with regards to asm-declared registers.
10672 @defmac MATH_LIBRARY
10673 Define this macro as a C string constant for the linker argument to link
10674 in the system math library, minus the initial @samp{"-l"}, or
10675 @samp{""} if the target does not have a
10676 separate math library.
10678 You need only define this macro if the default of @samp{"m"} is wrong.
10681 @defmac LIBRARY_PATH_ENV
10682 Define this macro as a C string constant for the environment variable that
10683 specifies where the linker should look for libraries.
10685 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10689 @defmac TARGET_POSIX_IO
10690 Define this macro if the target supports the following POSIX@ file
10691 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10692 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10693 to use file locking when exiting a program, which avoids race conditions
10694 if the program has forked. It will also create directories at run-time
10695 for cross-profiling.
10698 @defmac MAX_CONDITIONAL_EXECUTE
10700 A C expression for the maximum number of instructions to execute via
10701 conditional execution instructions instead of a branch. A value of
10702 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10703 1 if it does use cc0.
10706 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10707 Used if the target needs to perform machine-dependent modifications on the
10708 conditionals used for turning basic blocks into conditionally executed code.
10709 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10710 contains information about the currently processed blocks. @var{true_expr}
10711 and @var{false_expr} are the tests that are used for converting the
10712 then-block and the else-block, respectively. Set either @var{true_expr} or
10713 @var{false_expr} to a null pointer if the tests cannot be converted.
10716 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10717 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10718 if-statements into conditions combined by @code{and} and @code{or} operations.
10719 @var{bb} contains the basic block that contains the test that is currently
10720 being processed and about to be turned into a condition.
10723 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10724 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10725 be converted to conditional execution format. @var{ce_info} points to
10726 a data structure, @code{struct ce_if_block}, which contains information
10727 about the currently processed blocks.
10730 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10731 A C expression to perform any final machine dependent modifications in
10732 converting code to conditional execution. The involved basic blocks
10733 can be found in the @code{struct ce_if_block} structure that is pointed
10734 to by @var{ce_info}.
10737 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10738 A C expression to cancel any machine dependent modifications in
10739 converting code to conditional execution. The involved basic blocks
10740 can be found in the @code{struct ce_if_block} structure that is pointed
10741 to by @var{ce_info}.
10744 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10745 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10746 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10749 @defmac IFCVT_EXTRA_FIELDS
10750 If defined, it should expand to a set of field declarations that will be
10751 added to the @code{struct ce_if_block} structure. These should be initialized
10752 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10755 @hook TARGET_MACHINE_DEPENDENT_REORG
10756 If non-null, this hook performs a target-specific pass over the
10757 instruction stream. The compiler will run it at all optimization levels,
10758 just before the point at which it normally does delayed-branch scheduling.
10760 The exact purpose of the hook varies from target to target. Some use
10761 it to do transformations that are necessary for correctness, such as
10762 laying out in-function constant pools or avoiding hardware hazards.
10763 Others use it as an opportunity to do some machine-dependent optimizations.
10765 You need not implement the hook if it has nothing to do. The default
10766 definition is null.
10769 @hook TARGET_INIT_BUILTINS
10770 Define this hook if you have any machine-specific built-in functions
10771 that need to be defined. It should be a function that performs the
10774 Machine specific built-in functions can be useful to expand special machine
10775 instructions that would otherwise not normally be generated because
10776 they have no equivalent in the source language (for example, SIMD vector
10777 instructions or prefetch instructions).
10779 To create a built-in function, call the function
10780 @code{lang_hooks.builtin_function}
10781 which is defined by the language front end. You can use any type nodes set
10782 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
10783 only language front ends that use those two functions will call
10784 @samp{TARGET_INIT_BUILTINS}.
10787 @hook TARGET_BUILTIN_DECL
10788 Define this hook if you have any machine-specific built-in functions
10789 that need to be defined. It should be a function that returns the
10790 builtin function declaration for the builtin function code @var{code}.
10791 If there is no such builtin and it cannot be initialized at this time
10792 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10793 If @var{code} is out of range the function should return
10794 @code{error_mark_node}.
10797 @hook TARGET_EXPAND_BUILTIN
10799 Expand a call to a machine specific built-in function that was set up by
10800 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10801 function call; the result should go to @var{target} if that is
10802 convenient, and have mode @var{mode} if that is convenient.
10803 @var{subtarget} may be used as the target for computing one of
10804 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10805 ignored. This function should return the result of the call to the
10809 @hook TARGET_RESOLVE_OVERLOADED_BUILTIN
10810 Select a replacement for a machine specific built-in function that
10811 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10812 @emph{before} regular type checking, and so allows the target to
10813 implement a crude form of function overloading. @var{fndecl} is the
10814 declaration of the built-in function. @var{arglist} is the list of
10815 arguments passed to the built-in function. The result is a
10816 complete expression that implements the operation, usually
10817 another @code{CALL_EXPR}.
10818 @var{arglist} really has type @samp{VEC(tree,gc)*}
10821 @hook TARGET_FOLD_BUILTIN
10822 Fold a call to a machine specific built-in function that was set up by
10823 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10824 built-in function. @var{n_args} is the number of arguments passed to
10825 the function; the arguments themselves are pointed to by @var{argp}.
10826 The result is another tree containing a simplified expression for the
10827 call's result. If @var{ignore} is true the value will be ignored.
10830 @hook TARGET_INVALID_WITHIN_DOLOOP
10832 Take an instruction in @var{insn} and return NULL if it is valid within a
10833 low-overhead loop, otherwise return a string explaining why doloop
10834 could not be applied.
10836 Many targets use special registers for low-overhead looping. For any
10837 instruction that clobbers these this function should return a string indicating
10838 the reason why the doloop could not be applied.
10839 By default, the RTL loop optimizer does not use a present doloop pattern for
10840 loops containing function calls or branch on table instructions.
10843 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10845 Take a branch insn in @var{branch1} and another in @var{branch2}.
10846 Return true if redirecting @var{branch1} to the destination of
10847 @var{branch2} is possible.
10849 On some targets, branches may have a limited range. Optimizing the
10850 filling of delay slots can result in branches being redirected, and this
10851 may in turn cause a branch offset to overflow.
10854 @hook TARGET_COMMUTATIVE_P
10855 This target hook returns @code{true} if @var{x} is considered to be commutative.
10856 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10857 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10858 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10861 @hook TARGET_ALLOCATE_INITIAL_VALUE
10863 When the initial value of a hard register has been copied in a pseudo
10864 register, it is often not necessary to actually allocate another register
10865 to this pseudo register, because the original hard register or a stack slot
10866 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10867 is called at the start of register allocation once for each hard register
10868 that had its initial value copied by using
10869 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10870 Possible values are @code{NULL_RTX}, if you don't want
10871 to do any special allocation, a @code{REG} rtx---that would typically be
10872 the hard register itself, if it is known not to be clobbered---or a
10874 If you are returning a @code{MEM}, this is only a hint for the allocator;
10875 it might decide to use another register anyways.
10876 You may use @code{current_function_leaf_function} in the hook, functions
10877 that use @code{REG_N_SETS}, to determine if the hard
10878 register in question will not be clobbered.
10879 The default value of this hook is @code{NULL}, which disables any special
10883 @hook TARGET_UNSPEC_MAY_TRAP_P
10884 This target hook returns nonzero if @var{x}, an @code{unspec} or
10885 @code{unspec_volatile} operation, might cause a trap. Targets can use
10886 this hook to enhance precision of analysis for @code{unspec} and
10887 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10888 to analyze inner elements of @var{x} in which case @var{flags} should be
10892 @hook TARGET_SET_CURRENT_FUNCTION
10893 The compiler invokes this hook whenever it changes its current function
10894 context (@code{cfun}). You can define this function if
10895 the back end needs to perform any initialization or reset actions on a
10896 per-function basis. For example, it may be used to implement function
10897 attributes that affect register usage or code generation patterns.
10898 The argument @var{decl} is the declaration for the new function context,
10899 and may be null to indicate that the compiler has left a function context
10900 and is returning to processing at the top level.
10901 The default hook function does nothing.
10903 GCC sets @code{cfun} to a dummy function context during initialization of
10904 some parts of the back end. The hook function is not invoked in this
10905 situation; you need not worry about the hook being invoked recursively,
10906 or when the back end is in a partially-initialized state.
10907 @code{cfun} might be @code{NULL} to indicate processing at top level,
10908 outside of any function scope.
10911 @defmac TARGET_OBJECT_SUFFIX
10912 Define this macro to be a C string representing the suffix for object
10913 files on your target machine. If you do not define this macro, GCC will
10914 use @samp{.o} as the suffix for object files.
10917 @defmac TARGET_EXECUTABLE_SUFFIX
10918 Define this macro to be a C string representing the suffix to be
10919 automatically added to executable files on your target machine. If you
10920 do not define this macro, GCC will use the null string as the suffix for
10924 @defmac COLLECT_EXPORT_LIST
10925 If defined, @code{collect2} will scan the individual object files
10926 specified on its command line and create an export list for the linker.
10927 Define this macro for systems like AIX, where the linker discards
10928 object files that are not referenced from @code{main} and uses export
10932 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
10933 Define this macro to a C expression representing a variant of the
10934 method call @var{mdecl}, if Java Native Interface (JNI) methods
10935 must be invoked differently from other methods on your target.
10936 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
10937 the @code{stdcall} calling convention and this macro is then
10938 defined as this expression:
10941 build_type_attribute_variant (@var{mdecl},
10943 (get_identifier ("stdcall"),
10948 @hook TARGET_CANNOT_MODIFY_JUMPS_P
10949 This target hook returns @code{true} past the point in which new jump
10950 instructions could be created. On machines that require a register for
10951 every jump such as the SHmedia ISA of SH5, this point would typically be
10952 reload, so this target hook should be defined to a function such as:
10956 cannot_modify_jumps_past_reload_p ()
10958 return (reload_completed || reload_in_progress);
10963 @hook TARGET_BRANCH_TARGET_REGISTER_CLASS
10964 This target hook returns a register class for which branch target register
10965 optimizations should be applied. All registers in this class should be
10966 usable interchangeably. After reload, registers in this class will be
10967 re-allocated and loads will be hoisted out of loops and be subjected
10968 to inter-block scheduling.
10971 @hook TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED
10972 Branch target register optimization will by default exclude callee-saved
10974 that are not already live during the current function; if this target hook
10975 returns true, they will be included. The target code must than make sure
10976 that all target registers in the class returned by
10977 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
10978 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
10979 epilogues have already been generated. Note, even if you only return
10980 true when @var{after_prologue_epilogue_gen} is false, you still are likely
10981 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
10982 to reserve space for caller-saved target registers.
10985 @hook TARGET_HAVE_CONDITIONAL_EXECUTION
10986 This target hook returns true if the target supports conditional execution.
10987 This target hook is required only when the target has several different
10988 modes and they have different conditional execution capability, such as ARM.
10991 @hook TARGET_LOOP_UNROLL_ADJUST
10992 This target hook returns a new value for the number of times @var{loop}
10993 should be unrolled. The parameter @var{nunroll} is the number of times
10994 the loop is to be unrolled. The parameter @var{loop} is a pointer to
10995 the loop, which is going to be checked for unrolling. This target hook
10996 is required only when the target has special constraints like maximum
10997 number of memory accesses.
11000 @defmac POWI_MAX_MULTS
11001 If defined, this macro is interpreted as a signed integer C expression
11002 that specifies the maximum number of floating point multiplications
11003 that should be emitted when expanding exponentiation by an integer
11004 constant inline. When this value is defined, exponentiation requiring
11005 more than this number of multiplications is implemented by calling the
11006 system library's @code{pow}, @code{powf} or @code{powl} routines.
11007 The default value places no upper bound on the multiplication count.
11010 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11011 This target hook should register any extra include files for the
11012 target. The parameter @var{stdinc} indicates if normal include files
11013 are present. The parameter @var{sysroot} is the system root directory.
11014 The parameter @var{iprefix} is the prefix for the gcc directory.
11017 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11018 This target hook should register any extra include files for the
11019 target before any standard headers. The parameter @var{stdinc}
11020 indicates if normal include files are present. The parameter
11021 @var{sysroot} is the system root directory. The parameter
11022 @var{iprefix} is the prefix for the gcc directory.
11025 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11026 This target hook should register special include paths for the target.
11027 The parameter @var{path} is the include to register. On Darwin
11028 systems, this is used for Framework includes, which have semantics
11029 that are different from @option{-I}.
11032 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11033 This target macro returns @code{true} if it is safe to use a local alias
11034 for a virtual function @var{fndecl} when constructing thunks,
11035 @code{false} otherwise. By default, the macro returns @code{true} for all
11036 functions, if a target supports aliases (i.e.@: defines
11037 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11040 @defmac TARGET_FORMAT_TYPES
11041 If defined, this macro is the name of a global variable containing
11042 target-specific format checking information for the @option{-Wformat}
11043 option. The default is to have no target-specific format checks.
11046 @defmac TARGET_N_FORMAT_TYPES
11047 If defined, this macro is the number of entries in
11048 @code{TARGET_FORMAT_TYPES}.
11051 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11052 If defined, this macro is the name of a global variable containing
11053 target-specific format overrides for the @option{-Wformat} option. The
11054 default is to have no target-specific format overrides. If defined,
11055 @code{TARGET_FORMAT_TYPES} must be defined, too.
11058 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11059 If defined, this macro specifies the number of entries in
11060 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11063 @defmac TARGET_OVERRIDES_FORMAT_INIT
11064 If defined, this macro specifies the optional initialization
11065 routine for target specific customizations of the system printf
11066 and scanf formatter settings.
11069 @hook TARGET_RELAXED_ORDERING
11070 If set to @code{true}, means that the target's memory model does not
11071 guarantee that loads which do not depend on one another will access
11072 main memory in the order of the instruction stream; if ordering is
11073 important, an explicit memory barrier must be used. This is true of
11074 many recent processors which implement a policy of ``relaxed,''
11075 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11076 and ia64. The default is @code{false}.
11079 @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
11080 If defined, this macro returns the diagnostic message when it is
11081 illegal to pass argument @var{val} to function @var{funcdecl}
11082 with prototype @var{typelist}.
11085 @hook TARGET_INVALID_CONVERSION
11086 If defined, this macro returns the diagnostic message when it is
11087 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11088 if validity should be determined by the front end.
11091 @hook TARGET_INVALID_UNARY_OP
11092 If defined, this macro returns the diagnostic message when it is
11093 invalid to apply operation @var{op} (where unary plus is denoted by
11094 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11095 if validity should be determined by the front end.
11098 @hook TARGET_INVALID_BINARY_OP
11099 If defined, this macro returns the diagnostic message when it is
11100 invalid to apply operation @var{op} to operands of types @var{type1}
11101 and @var{type2}, or @code{NULL} if validity should be determined by
11105 @hook TARGET_INVALID_PARAMETER_TYPE
11106 If defined, this macro returns the diagnostic message when it is
11107 invalid for functions to include parameters of type @var{type},
11108 or @code{NULL} if validity should be determined by
11109 the front end. This is currently used only by the C and C++ front ends.
11112 @hook TARGET_INVALID_RETURN_TYPE
11113 If defined, this macro returns the diagnostic message when it is
11114 invalid for functions to have return type @var{type},
11115 or @code{NULL} if validity should be determined by
11116 the front end. This is currently used only by the C and C++ front ends.
11119 @hook TARGET_PROMOTED_TYPE
11120 If defined, this target hook returns the type to which values of
11121 @var{type} should be promoted when they appear in expressions,
11122 analogous to the integer promotions, or @code{NULL_TREE} to use the
11123 front end's normal promotion rules. This hook is useful when there are
11124 target-specific types with special promotion rules.
11125 This is currently used only by the C and C++ front ends.
11128 @hook TARGET_CONVERT_TO_TYPE
11129 If defined, this hook returns the result of converting @var{expr} to
11130 @var{type}. It should return the converted expression,
11131 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11132 This hook is useful when there are target-specific types with special
11134 This is currently used only by the C and C++ front ends.
11137 @defmac TARGET_USE_JCR_SECTION
11138 This macro determines whether to use the JCR section to register Java
11139 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11140 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11144 This macro determines the size of the objective C jump buffer for the
11145 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11148 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11149 Define this macro if any target-specific attributes need to be attached
11150 to the functions in @file{libgcc} that provide low-level support for
11151 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11152 and the associated definitions of those functions.
11155 @hook TARGET_UPDATE_STACK_BOUNDARY
11156 Define this macro to update the current function stack boundary if
11160 @hook TARGET_GET_DRAP_RTX
11161 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11162 different argument pointer register is needed to access the function's
11163 argument list due to stack realignment. Return @code{NULL} if no DRAP
11167 @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
11168 When optimization is disabled, this hook indicates whether or not
11169 arguments should be allocated to stack slots. Normally, GCC allocates
11170 stacks slots for arguments when not optimizing in order to make
11171 debugging easier. However, when a function is declared with
11172 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11173 cannot safely move arguments from the registers in which they are passed
11174 to the stack. Therefore, this hook should return true in general, but
11175 false for naked functions. The default implementation always returns true.
11178 @hook TARGET_CONST_ANCHOR
11179 On some architectures it can take multiple instructions to synthesize
11180 a constant. If there is another constant already in a register that
11181 is close enough in value then it is preferable that the new constant
11182 is computed from this register using immediate addition or
11183 subtraction. We accomplish this through CSE. Besides the value of
11184 the constant we also add a lower and an upper constant anchor to the
11185 available expressions. These are then queried when encountering new
11186 constants. The anchors are computed by rounding the constant up and
11187 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11188 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11189 accepted by immediate-add plus one. We currently assume that the
11190 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11191 MIPS, where add-immediate takes a 16-bit signed value,
11192 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11193 is zero, which disables this optimization. @end deftypevr