1 @c Copyright (C) 1988-2014 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
6 @chapter Target Description Macros and Functions
7 @cindex machine description macros
8 @cindex target description macros
9 @cindex macros, target description
10 @cindex @file{tm.h} macros
12 In addition to the file @file{@var{machine}.md}, a machine description
13 includes a C header file conventionally given the name
14 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
15 The header file defines numerous macros that convey the information
16 about the target machine that does not fit into the scheme of the
17 @file{.md} file. The file @file{tm.h} should be a link to
18 @file{@var{machine}.h}. The header file @file{config.h} includes
19 @file{tm.h} and most compiler source files include @file{config.h}. The
20 source file defines a variable @code{targetm}, which is a structure
21 containing pointers to functions and data relating to the target
22 machine. @file{@var{machine}.c} should also contain their definitions,
23 if they are not defined elsewhere in GCC, and other functions called
24 through the macros defined in the @file{.h} file.
27 * Target Structure:: The @code{targetm} variable.
28 * Driver:: Controlling how the driver runs the compilation passes.
29 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
30 * Per-Function Data:: Defining data structures for per-function information.
31 * Storage Layout:: Defining sizes and alignments of data.
32 * Type Layout:: Defining sizes and properties of basic user data types.
33 * Registers:: Naming and describing the hardware registers.
34 * Register Classes:: Defining the classes of hardware registers.
35 * Old Constraints:: The old way to define machine-specific constraints.
36 * Stack and Calling:: Defining which way the stack grows and by how much.
37 * Varargs:: Defining the varargs macros.
38 * Trampolines:: Code set up at run time to enter a nested function.
39 * Library Calls:: Controlling how library routines are implicitly called.
40 * Addressing Modes:: Defining addressing modes valid for memory operands.
41 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
42 * Condition Code:: Defining how insns update the condition code.
43 * Costs:: Defining relative costs of different operations.
44 * Scheduling:: Adjusting the behavior of the instruction scheduler.
45 * Sections:: Dividing storage into text, data, and other sections.
46 * PIC:: Macros for position independent code.
47 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
48 * Debugging Info:: Defining the format of debugging output.
49 * Floating Point:: Handling floating point for cross-compilers.
50 * Mode Switching:: Insertion of mode-switching instructions.
51 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
52 * Emulated TLS:: Emulated TLS support.
53 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
54 * PCH Target:: Validity checking for precompiled headers.
55 * C++ ABI:: Controlling C++ ABI changes.
56 * Named Address Spaces:: Adding support for named address spaces
57 * Misc:: Everything else.
60 @node Target Structure
61 @section The Global @code{targetm} Variable
63 @cindex target functions
65 @deftypevar {struct gcc_target} targetm
66 The target @file{.c} file must define the global @code{targetm} variable
67 which contains pointers to functions and data relating to the target
68 machine. The variable is declared in @file{target.h};
69 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
70 used to initialize the variable, and macros for the default initializers
71 for elements of the structure. The @file{.c} file should override those
72 macros for which the default definition is inappropriate. For example:
75 #include "target-def.h"
77 /* @r{Initialize the GCC target structure.} */
79 #undef TARGET_COMP_TYPE_ATTRIBUTES
80 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
82 struct gcc_target targetm = TARGET_INITIALIZER;
86 Where a macro should be defined in the @file{.c} file in this manner to
87 form part of the @code{targetm} structure, it is documented below as a
88 ``Target Hook'' with a prototype. Many macros will change in future
89 from being defined in the @file{.h} file to being part of the
90 @code{targetm} structure.
92 Similarly, there is a @code{targetcm} variable for hooks that are
93 specific to front ends for C-family languages, documented as ``C
94 Target Hook''. This is declared in @file{c-family/c-target.h}, the
95 initializer @code{TARGETCM_INITIALIZER} in
96 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
97 themselves, they should set @code{target_has_targetcm=yes} in
98 @file{config.gcc}; otherwise a default definition is used.
100 Similarly, there is a @code{targetm_common} variable for hooks that
101 are shared between the compiler driver and the compilers proper,
102 documented as ``Common Target Hook''. This is declared in
103 @file{common/common-target.h}, the initializer
104 @code{TARGETM_COMMON_INITIALIZER} in
105 @file{common/common-target-def.h}. If targets initialize
106 @code{targetm_common} themselves, they should set
107 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
108 default definition is used.
111 @section Controlling the Compilation Driver, @file{gcc}
113 @cindex controlling the compilation driver
115 @c prevent bad page break with this line
116 You can control the compilation driver.
118 @defmac DRIVER_SELF_SPECS
119 A list of specs for the driver itself. It should be a suitable
120 initializer for an array of strings, with no surrounding braces.
122 The driver applies these specs to its own command line between loading
123 default @file{specs} files (but not command-line specified ones) and
124 choosing the multilib directory or running any subcommands. It
125 applies them in the order given, so each spec can depend on the
126 options added by earlier ones. It is also possible to remove options
127 using @samp{%<@var{option}} in the usual way.
129 This macro can be useful when a port has several interdependent target
130 options. It provides a way of standardizing the command line so
131 that the other specs are easier to write.
133 Do not define this macro if it does not need to do anything.
136 @defmac OPTION_DEFAULT_SPECS
137 A list of specs used to support configure-time default options (i.e.@:
138 @option{--with} options) in the driver. It should be a suitable initializer
139 for an array of structures, each containing two strings, without the
140 outermost pair of surrounding braces.
142 The first item in the pair is the name of the default. This must match
143 the code in @file{config.gcc} for the target. The second item is a spec
144 to apply if a default with this name was specified. The string
145 @samp{%(VALUE)} in the spec will be replaced by the value of the default
146 everywhere it occurs.
148 The driver will apply these specs to its own command line between loading
149 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
150 the same mechanism as @code{DRIVER_SELF_SPECS}.
152 Do not define this macro if it does not need to do anything.
156 A C string constant that tells the GCC driver program options to
157 pass to CPP@. It can also specify how to translate options you
158 give to GCC into options for GCC to pass to the CPP@.
160 Do not define this macro if it does not need to do anything.
163 @defmac CPLUSPLUS_CPP_SPEC
164 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
165 than C@. If you do not define this macro, then the value of
166 @code{CPP_SPEC} (if any) will be used instead.
170 A C string constant that tells the GCC driver program options to
171 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
173 It can also specify how to translate options you give to GCC into options
174 for GCC to pass to front ends.
176 Do not define this macro if it does not need to do anything.
180 A C string constant that tells the GCC driver program options to
181 pass to @code{cc1plus}. It can also specify how to translate options you
182 give to GCC into options for GCC to pass to the @code{cc1plus}.
184 Do not define this macro if it does not need to do anything.
185 Note that everything defined in CC1_SPEC is already passed to
186 @code{cc1plus} so there is no need to duplicate the contents of
187 CC1_SPEC in CC1PLUS_SPEC@.
191 A C string constant that tells the GCC driver program options to
192 pass to the assembler. It can also specify how to translate options
193 you give to GCC into options for GCC to pass to the assembler.
194 See the file @file{sun3.h} for an example of this.
196 Do not define this macro if it does not need to do anything.
199 @defmac ASM_FINAL_SPEC
200 A C string constant that tells the GCC driver program how to
201 run any programs which cleanup after the normal assembler.
202 Normally, this is not needed. See the file @file{mips.h} for
205 Do not define this macro if it does not need to do anything.
208 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
209 Define this macro, with no value, if the driver should give the assembler
210 an argument consisting of a single dash, @option{-}, to instruct it to
211 read from its standard input (which will be a pipe connected to the
212 output of the compiler proper). This argument is given after any
213 @option{-o} option specifying the name of the output file.
215 If you do not define this macro, the assembler is assumed to read its
216 standard input if given no non-option arguments. If your assembler
217 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
218 see @file{mips.h} for instance.
222 A C string constant that tells the GCC driver program options to
223 pass to the linker. It can also specify how to translate options you
224 give to GCC into options for GCC to pass to the linker.
226 Do not define this macro if it does not need to do anything.
230 Another C string constant used much like @code{LINK_SPEC}. The difference
231 between the two is that @code{LIB_SPEC} is used at the end of the
232 command given to the linker.
234 If this macro is not defined, a default is provided that
235 loads the standard C library from the usual place. See @file{gcc.c}.
239 Another C string constant that tells the GCC driver program
240 how and when to place a reference to @file{libgcc.a} into the
241 linker command line. This constant is placed both before and after
242 the value of @code{LIB_SPEC}.
244 If this macro is not defined, the GCC driver provides a default that
245 passes the string @option{-lgcc} to the linker.
248 @defmac REAL_LIBGCC_SPEC
249 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
250 @code{LIBGCC_SPEC} is not directly used by the driver program but is
251 instead modified to refer to different versions of @file{libgcc.a}
252 depending on the values of the command line flags @option{-static},
253 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
254 targets where these modifications are inappropriate, define
255 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
256 driver how to place a reference to @file{libgcc} on the link command
257 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
260 @defmac USE_LD_AS_NEEDED
261 A macro that controls the modifications to @code{LIBGCC_SPEC}
262 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
263 generated that uses @option{--as-needed} or equivalent options and the
264 shared @file{libgcc} in place of the
265 static exception handler library, when linking without any of
266 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
270 If defined, this C string constant is added to @code{LINK_SPEC}.
271 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
272 the modifications to @code{LIBGCC_SPEC} mentioned in
273 @code{REAL_LIBGCC_SPEC}.
276 @defmac STARTFILE_SPEC
277 Another C string constant used much like @code{LINK_SPEC}. The
278 difference between the two is that @code{STARTFILE_SPEC} is used at
279 the very beginning of the command given to the linker.
281 If this macro is not defined, a default is provided that loads the
282 standard C startup file from the usual place. See @file{gcc.c}.
286 Another C string constant used much like @code{LINK_SPEC}. The
287 difference between the two is that @code{ENDFILE_SPEC} is used at
288 the very end of the command given to the linker.
290 Do not define this macro if it does not need to do anything.
293 @defmac THREAD_MODEL_SPEC
294 GCC @code{-v} will print the thread model GCC was configured to use.
295 However, this doesn't work on platforms that are multilibbed on thread
296 models, such as AIX 4.3. On such platforms, define
297 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
298 blanks that names one of the recognized thread models. @code{%*}, the
299 default value of this macro, will expand to the value of
300 @code{thread_file} set in @file{config.gcc}.
303 @defmac SYSROOT_SUFFIX_SPEC
304 Define this macro to add a suffix to the target sysroot when GCC is
305 configured with a sysroot. This will cause GCC to search for usr/lib,
306 et al, within sysroot+suffix.
309 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
310 Define this macro to add a headers_suffix to the target sysroot when
311 GCC is configured with a sysroot. This will cause GCC to pass the
312 updated sysroot+headers_suffix to CPP, causing it to search for
313 usr/include, et al, within sysroot+headers_suffix.
317 Define this macro to provide additional specifications to put in the
318 @file{specs} file that can be used in various specifications like
321 The definition should be an initializer for an array of structures,
322 containing a string constant, that defines the specification name, and a
323 string constant that provides the specification.
325 Do not define this macro if it does not need to do anything.
327 @code{EXTRA_SPECS} is useful when an architecture contains several
328 related targets, which have various @code{@dots{}_SPECS} which are similar
329 to each other, and the maintainer would like one central place to keep
332 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
333 define either @code{_CALL_SYSV} when the System V calling sequence is
334 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
337 The @file{config/rs6000/rs6000.h} target file defines:
340 #define EXTRA_SPECS \
341 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
343 #define CPP_SYS_DEFAULT ""
346 The @file{config/rs6000/sysv.h} target file defines:
350 "%@{posix: -D_POSIX_SOURCE @} \
351 %@{mcall-sysv: -D_CALL_SYSV @} \
352 %@{!mcall-sysv: %(cpp_sysv_default) @} \
353 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
355 #undef CPP_SYSV_DEFAULT
356 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
359 while the @file{config/rs6000/eabiaix.h} target file defines
360 @code{CPP_SYSV_DEFAULT} as:
363 #undef CPP_SYSV_DEFAULT
364 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
368 @defmac LINK_LIBGCC_SPECIAL_1
369 Define this macro if the driver program should find the library
370 @file{libgcc.a}. If you do not define this macro, the driver program will pass
371 the argument @option{-lgcc} to tell the linker to do the search.
374 @defmac LINK_GCC_C_SEQUENCE_SPEC
375 The sequence in which libgcc and libc are specified to the linker.
376 By default this is @code{%G %L %G}.
379 @defmac LINK_COMMAND_SPEC
380 A C string constant giving the complete command line need to execute the
381 linker. When you do this, you will need to update your port each time a
382 change is made to the link command line within @file{gcc.c}. Therefore,
383 define this macro only if you need to completely redefine the command
384 line for invoking the linker and there is no other way to accomplish
385 the effect you need. Overriding this macro may be avoidable by overriding
386 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
389 @deftypevr {Common Target Hook} bool TARGET_ALWAYS_STRIP_DOTDOT
390 True if @file{..} components should always be removed from directory names computed relative to GCC's internal directories, false (default) if such components should be preserved and directory names containing them passed to other tools such as the linker.
393 @defmac MULTILIB_DEFAULTS
394 Define this macro as a C expression for the initializer of an array of
395 string to tell the driver program which options are defaults for this
396 target and thus do not need to be handled specially when using
397 @code{MULTILIB_OPTIONS}.
399 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
400 the target makefile fragment or if none of the options listed in
401 @code{MULTILIB_OPTIONS} are set by default.
402 @xref{Target Fragment}.
405 @defmac RELATIVE_PREFIX_NOT_LINKDIR
406 Define this macro to tell @command{gcc} that it should only translate
407 a @option{-B} prefix into a @option{-L} linker option if the prefix
408 indicates an absolute file name.
411 @defmac MD_EXEC_PREFIX
412 If defined, this macro is an additional prefix to try after
413 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
414 when the compiler is built as a cross
415 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
416 to the list of directories used to find the assembler in @file{configure.in}.
419 @defmac STANDARD_STARTFILE_PREFIX
420 Define this macro as a C string constant if you wish to override the
421 standard choice of @code{libdir} as the default prefix to
422 try when searching for startup files such as @file{crt0.o}.
423 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
424 is built as a cross compiler.
427 @defmac STANDARD_STARTFILE_PREFIX_1
428 Define this macro as a C string constant if you wish to override the
429 standard choice of @code{/lib} as a prefix to try after the default prefix
430 when searching for startup files such as @file{crt0.o}.
431 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
432 is built as a cross compiler.
435 @defmac STANDARD_STARTFILE_PREFIX_2
436 Define this macro as a C string constant if you wish to override the
437 standard choice of @code{/lib} as yet another prefix to try after the
438 default prefix when searching for startup files such as @file{crt0.o}.
439 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
440 is built as a cross compiler.
443 @defmac MD_STARTFILE_PREFIX
444 If defined, this macro supplies an additional prefix to try after the
445 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
446 compiler is built as a cross compiler.
449 @defmac MD_STARTFILE_PREFIX_1
450 If defined, this macro supplies yet another prefix to try after the
451 standard prefixes. It is not searched when the compiler is built as a
455 @defmac INIT_ENVIRONMENT
456 Define this macro as a C string constant if you wish to set environment
457 variables for programs called by the driver, such as the assembler and
458 loader. The driver passes the value of this macro to @code{putenv} to
459 initialize the necessary environment variables.
462 @defmac LOCAL_INCLUDE_DIR
463 Define this macro as a C string constant if you wish to override the
464 standard choice of @file{/usr/local/include} as the default prefix to
465 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
466 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
467 @file{config.gcc}, normally @file{/usr/include}) in the search order.
469 Cross compilers do not search either @file{/usr/local/include} or its
473 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
474 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
475 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
476 If you do not define this macro, no component is used.
479 @defmac INCLUDE_DEFAULTS
480 Define this macro if you wish to override the entire default search path
481 for include files. For a native compiler, the default search path
482 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
483 @code{GPLUSPLUS_INCLUDE_DIR}, and
484 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
485 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
486 and specify private search areas for GCC@. The directory
487 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
489 The definition should be an initializer for an array of structures.
490 Each array element should have four elements: the directory name (a
491 string constant), the component name (also a string constant), a flag
492 for C++-only directories,
493 and a flag showing that the includes in the directory don't need to be
494 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
495 the array with a null element.
497 The component name denotes what GNU package the include file is part of,
498 if any, in all uppercase letters. For example, it might be @samp{GCC}
499 or @samp{BINUTILS}. If the package is part of a vendor-supplied
500 operating system, code the component name as @samp{0}.
502 For example, here is the definition used for VAX/VMS:
505 #define INCLUDE_DEFAULTS \
507 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
508 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
509 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
516 Here is the order of prefixes tried for exec files:
520 Any prefixes specified by the user with @option{-B}.
523 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
524 is not set and the compiler has not been installed in the configure-time
525 @var{prefix}, the location in which the compiler has actually been installed.
528 The directories specified by the environment variable @code{COMPILER_PATH}.
531 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
532 in the configured-time @var{prefix}.
535 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
538 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
541 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
545 Here is the order of prefixes tried for startfiles:
549 Any prefixes specified by the user with @option{-B}.
552 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
553 value based on the installed toolchain location.
556 The directories specified by the environment variable @code{LIBRARY_PATH}
557 (or port-specific name; native only, cross compilers do not use this).
560 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
561 in the configured @var{prefix} or this is a native compiler.
564 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
567 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
571 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
572 native compiler, or we have a target system root.
575 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
576 native compiler, or we have a target system root.
579 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
580 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
581 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
584 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
585 compiler, or we have a target system root. The default for this macro is
589 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
590 compiler, or we have a target system root. The default for this macro is
594 @node Run-time Target
595 @section Run-time Target Specification
596 @cindex run-time target specification
597 @cindex predefined macros
598 @cindex target specifications
600 @c prevent bad page break with this line
601 Here are run-time target specifications.
603 @defmac TARGET_CPU_CPP_BUILTINS ()
604 This function-like macro expands to a block of code that defines
605 built-in preprocessor macros and assertions for the target CPU, using
606 the functions @code{builtin_define}, @code{builtin_define_std} and
607 @code{builtin_assert}. When the front end
608 calls this macro it provides a trailing semicolon, and since it has
609 finished command line option processing your code can use those
612 @code{builtin_assert} takes a string in the form you pass to the
613 command-line option @option{-A}, such as @code{cpu=mips}, and creates
614 the assertion. @code{builtin_define} takes a string in the form
615 accepted by option @option{-D} and unconditionally defines the macro.
617 @code{builtin_define_std} takes a string representing the name of an
618 object-like macro. If it doesn't lie in the user's namespace,
619 @code{builtin_define_std} defines it unconditionally. Otherwise, it
620 defines a version with two leading underscores, and another version
621 with two leading and trailing underscores, and defines the original
622 only if an ISO standard was not requested on the command line. For
623 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
624 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
625 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
626 defines only @code{_ABI64}.
628 You can also test for the C dialect being compiled. The variable
629 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
630 or @code{clk_objective_c}. Note that if we are preprocessing
631 assembler, this variable will be @code{clk_c} but the function-like
632 macro @code{preprocessing_asm_p()} will return true, so you might want
633 to check for that first. If you need to check for strict ANSI, the
634 variable @code{flag_iso} can be used. The function-like macro
635 @code{preprocessing_trad_p()} can be used to check for traditional
639 @defmac TARGET_OS_CPP_BUILTINS ()
640 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
641 and is used for the target operating system instead.
644 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
645 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
646 and is used for the target object format. @file{elfos.h} uses this
647 macro to define @code{__ELF__}, so you probably do not need to define
651 @deftypevar {extern int} target_flags
652 This variable is declared in @file{options.h}, which is included before
653 any target-specific headers.
656 @deftypevr {Common Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
657 This variable specifies the initial value of @code{target_flags}.
658 Its default setting is 0.
661 @cindex optional hardware or system features
662 @cindex features, optional, in system conventions
664 @deftypefn {Common Target Hook} bool TARGET_HANDLE_OPTION (struct gcc_options *@var{opts}, struct gcc_options *@var{opts_set}, const struct cl_decoded_option *@var{decoded}, location_t @var{loc})
665 This hook is called whenever the user specifies one of the
666 target-specific options described by the @file{.opt} definition files
667 (@pxref{Options}). It has the opportunity to do some option-specific
668 processing and should return true if the option is valid. The default
669 definition does nothing but return true.
671 @var{decoded} specifies the option and its arguments. @var{opts} and
672 @var{opts_set} are the @code{gcc_options} structures to be used for
673 storing option state, and @var{loc} is the location at which the
674 option was passed (@code{UNKNOWN_LOCATION} except for options passed
678 @deftypefn {C Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
679 This target hook is called whenever the user specifies one of the
680 target-specific C language family options described by the @file{.opt}
681 definition files(@pxref{Options}). It has the opportunity to do some
682 option-specific processing and should return true if the option is
683 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
684 default definition does nothing but return false.
686 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
687 options. However, if processing an option requires routines that are
688 only available in the C (and related language) front ends, then you
689 should use @code{TARGET_HANDLE_C_OPTION} instead.
692 @deftypefn {C Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING_OBJECT (tree @var{string})
693 Targets may provide a string object type that can be used within and between C, C++ and their respective Objective-C dialects. A string object might, for example, embed encoding and length information. These objects are considered opaque to the compiler and handled as references. An ideal implementation makes the composition of the string object match that of the Objective-C @code{NSString} (@code{NXString} for GNUStep), allowing efficient interworking between C-only and Objective-C code. If a target implements string objects then this hook should return a reference to such an object constructed from the normal `C' string representation provided in @var{string}. At present, the hook is used by Objective-C only, to obtain a common-format string object when the target provides one.
696 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE (const char *@var{classname})
697 Declare that Objective C class @var{classname} is referenced by the current TU.
700 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_CLASS_DEFINITION (const char *@var{classname})
701 Declare that Objective C class @var{classname} is defined by the current TU.
704 @deftypefn {C Target Hook} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref})
705 If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object.
708 @deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list})
709 If a target implements string objects then this hook should should provide a facility to check the function arguments in @var{args_list} against the format specifiers in @var{format_arg} where the type of @var{format_arg} is one recognized as a valid string reference type.
712 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
713 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
714 but is called when the optimize level is changed via an attribute or
715 pragma or when it is reset at the end of the code affected by the
716 attribute or pragma. It is not called at the beginning of compilation
717 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
718 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
719 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
722 @defmac C_COMMON_OVERRIDE_OPTIONS
723 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
724 but is only used in the C
725 language frontends (C, Objective-C, C++, Objective-C++) and so can be
726 used to alter option flag variables which only exist in those
730 @deftypevr {Common Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
731 Some machines may desire to change what optimizations are performed for
732 various optimization levels. This variable, if defined, describes
733 options to enable at particular sets of optimization levels. These
734 options are processed once
735 just after the optimization level is determined and before the remainder
736 of the command options have been parsed, so may be overridden by other
737 options passed explicitly.
739 This processing is run once at program startup and when the optimization
740 options are changed via @code{#pragma GCC optimize} or by using the
741 @code{optimize} attribute.
744 @deftypefn {Common Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
745 Set target-dependent initial values of fields in @var{opts}.
748 @deftypefn {Common Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
749 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
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 @deftypefn {Target Hook} bool TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P (void)
772 Returns true if the target supports IEEE 754 floating-point exceptions and rounding modes, false otherwise. This is intended to relate to the @code{float} and @code{double} types, but not necessarily @code{long double}. By default, returns true if the @code{adddf3} instruction pattern is available and false otherwise, on the assumption that hardware floating point supports exceptions and rounding modes but software floating point does not.
775 @node Per-Function Data
776 @section Defining data structures for per-function information.
777 @cindex per-function data
778 @cindex data structures
780 If the target needs to store information on a per-function basis, GCC
781 provides a macro and a couple of variables to allow this. Note, just
782 using statics to store the information is a bad idea, since GCC supports
783 nested functions, so you can be halfway through encoding one function
784 when another one comes along.
786 GCC defines a data structure called @code{struct function} which
787 contains all of the data specific to an individual function. This
788 structure contains a field called @code{machine} whose type is
789 @code{struct machine_function *}, which can be used by targets to point
790 to their own specific data.
792 If a target needs per-function specific data it should define the type
793 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
794 This macro should be used to initialize the function pointer
795 @code{init_machine_status}. This pointer is explained below.
797 One typical use of per-function, target specific data is to create an
798 RTX to hold the register containing the function's return address. This
799 RTX can then be used to implement the @code{__builtin_return_address}
800 function, for level 0.
802 Note---earlier implementations of GCC used a single data area to hold
803 all of the per-function information. Thus when processing of a nested
804 function began the old per-function data had to be pushed onto a
805 stack, and when the processing was finished, it had to be popped off the
806 stack. GCC used to provide function pointers called
807 @code{save_machine_status} and @code{restore_machine_status} to handle
808 the saving and restoring of the target specific information. Since the
809 single data area approach is no longer used, these pointers are no
812 @defmac INIT_EXPANDERS
813 Macro called to initialize any target specific information. This macro
814 is called once per function, before generation of any RTL has begun.
815 The intention of this macro is to allow the initialization of the
816 function pointer @code{init_machine_status}.
819 @deftypevar {void (*)(struct function *)} init_machine_status
820 If this function pointer is non-@code{NULL} it will be called once per
821 function, before function compilation starts, in order to allow the
822 target to perform any target specific initialization of the
823 @code{struct function} structure. It is intended that this would be
824 used to initialize the @code{machine} of that structure.
826 @code{struct machine_function} structures are expected to be freed by GC@.
827 Generally, any memory that they reference must be allocated by using
828 GC allocation, including the structure itself.
832 @section Storage Layout
833 @cindex storage layout
835 Note that the definitions of the macros in this table which are sizes or
836 alignments measured in bits do not need to be constant. They can be C
837 expressions that refer to static variables, such as the @code{target_flags}.
838 @xref{Run-time Target}.
840 @defmac BITS_BIG_ENDIAN
841 Define this macro to have the value 1 if the most significant bit in a
842 byte has the lowest number; otherwise define it to have the value zero.
843 This means that bit-field instructions count from the most significant
844 bit. If the machine has no bit-field instructions, then this must still
845 be defined, but it doesn't matter which value it is defined to. This
846 macro need not be a constant.
848 This macro does not affect the way structure fields are packed into
849 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
852 @defmac BYTES_BIG_ENDIAN
853 Define this macro to have the value 1 if the most significant byte in a
854 word has the lowest number. This macro need not be a constant.
857 @defmac WORDS_BIG_ENDIAN
858 Define this macro to have the value 1 if, in a multiword object, the
859 most significant word has the lowest number. This applies to both
860 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
861 order of words in memory is not the same as the order in registers. This
862 macro need not be a constant.
865 @defmac REG_WORDS_BIG_ENDIAN
866 On some machines, the order of words in a multiword object differs between
867 registers in memory. In such a situation, define this macro to describe
868 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
869 the order of words in memory.
872 @defmac FLOAT_WORDS_BIG_ENDIAN
873 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
874 @code{TFmode} floating point numbers are stored in memory with the word
875 containing the sign bit at the lowest address; otherwise define it to
876 have the value 0. This macro need not be a constant.
878 You need not define this macro if the ordering is the same as for
882 @defmac BITS_PER_WORD
883 Number of bits in a word. If you do not define this macro, the default
884 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
887 @defmac MAX_BITS_PER_WORD
888 Maximum number of bits in a word. If this is undefined, the default is
889 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
890 largest value that @code{BITS_PER_WORD} can have at run-time.
893 @defmac UNITS_PER_WORD
894 Number of storage units in a word; normally the size of a general-purpose
895 register, a power of two from 1 or 8.
898 @defmac MIN_UNITS_PER_WORD
899 Minimum number of units in a word. If this is undefined, the default is
900 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
901 smallest value that @code{UNITS_PER_WORD} can have at run-time.
905 Width of a pointer, in bits. You must specify a value no wider than the
906 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
907 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
908 a value the default is @code{BITS_PER_WORD}.
911 @defmac POINTERS_EXTEND_UNSIGNED
912 A C expression that determines how pointers should be extended from
913 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
914 greater than zero if pointers should be zero-extended, zero if they
915 should be sign-extended, and negative if some other sort of conversion
916 is needed. In the last case, the extension is done by the target's
917 @code{ptr_extend} instruction.
919 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
920 and @code{word_mode} are all the same width.
923 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
924 A macro to update @var{m} and @var{unsignedp} when an object whose type
925 is @var{type} and which has the specified mode and signedness is to be
926 stored in a register. This macro is only called when @var{type} is a
929 On most RISC machines, which only have operations that operate on a full
930 register, define this macro to set @var{m} to @code{word_mode} if
931 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
932 cases, only integer modes should be widened because wider-precision
933 floating-point operations are usually more expensive than their narrower
936 For most machines, the macro definition does not change @var{unsignedp}.
937 However, some machines, have instructions that preferentially handle
938 either signed or unsigned quantities of certain modes. For example, on
939 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
940 sign-extend the result to 64 bits. On such machines, set
941 @var{unsignedp} according to which kind of extension is more efficient.
943 Do not define this macro if it would never modify @var{m}.
946 @deftypefn {Target Hook} {enum machine_mode} TARGET_PROMOTE_FUNCTION_MODE (const_tree @var{type}, enum machine_mode @var{mode}, int *@var{punsignedp}, const_tree @var{funtype}, int @var{for_return})
947 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
948 function return values. The target hook should return the new mode
949 and possibly change @code{*@var{punsignedp}} if the promotion should
950 change signedness. This function is called only for scalar @emph{or
953 @var{for_return} allows to distinguish the promotion of arguments and
954 return values. If it is @code{1}, a return value is being promoted and
955 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
956 If it is @code{2}, the returned mode should be that of the register in
957 which an incoming parameter is copied, or the outgoing result is computed;
958 then the hook should return the same mode as @code{promote_mode}, though
959 the signedness may be different.
961 @var{type} can be NULL when promoting function arguments of libcalls.
963 The default is to not promote arguments and return values. You can
964 also define the hook to @code{default_promote_function_mode_always_promote}
965 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
968 @defmac PARM_BOUNDARY
969 Normal alignment required for function parameters on the stack, in
970 bits. All stack parameters receive at least this much alignment
971 regardless of data type. On most machines, this is the same as the
975 @defmac STACK_BOUNDARY
976 Define this macro to the minimum alignment enforced by hardware for the
977 stack pointer on this machine. The definition is a C expression for the
978 desired alignment (measured in bits). This value is used as a default
979 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
980 this should be the same as @code{PARM_BOUNDARY}.
983 @defmac PREFERRED_STACK_BOUNDARY
984 Define this macro if you wish to preserve a certain alignment for the
985 stack pointer, greater than what the hardware enforces. The definition
986 is a C expression for the desired alignment (measured in bits). This
987 macro must evaluate to a value equal to or larger than
988 @code{STACK_BOUNDARY}.
991 @defmac INCOMING_STACK_BOUNDARY
992 Define this macro if the incoming stack boundary may be different
993 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
994 to a value equal to or larger than @code{STACK_BOUNDARY}.
997 @defmac FUNCTION_BOUNDARY
998 Alignment required for a function entry point, in bits.
1001 @defmac BIGGEST_ALIGNMENT
1002 Biggest alignment that any data type can require on this machine, in
1003 bits. Note that this is not the biggest alignment that is supported,
1004 just the biggest alignment that, when violated, may cause a fault.
1007 @defmac MALLOC_ABI_ALIGNMENT
1008 Alignment, in bits, a C conformant malloc implementation has to
1009 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1012 @defmac ATTRIBUTE_ALIGNED_VALUE
1013 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1014 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1017 @defmac MINIMUM_ATOMIC_ALIGNMENT
1018 If defined, the smallest alignment, in bits, that can be given to an
1019 object that can be referenced in one operation, without disturbing any
1020 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1021 on machines that don't have byte or half-word store operations.
1024 @defmac BIGGEST_FIELD_ALIGNMENT
1025 Biggest alignment that any structure or union field can require on this
1026 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1027 structure and union fields only, unless the field alignment has been set
1028 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1031 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1032 An expression for the alignment of a structure field @var{field} if the
1033 alignment computed in the usual way (including applying of
1034 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1035 alignment) is @var{computed}. It overrides alignment only if the
1036 field alignment has not been set by the
1037 @code{__attribute__ ((aligned (@var{n})))} construct.
1040 @defmac MAX_STACK_ALIGNMENT
1041 Biggest stack alignment guaranteed by the backend. Use this macro
1042 to specify the maximum alignment of a variable on stack.
1044 If not defined, the default value is @code{STACK_BOUNDARY}.
1046 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1047 @c But the fix for PR 32893 indicates that we can only guarantee
1048 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1049 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1052 @defmac MAX_OFILE_ALIGNMENT
1053 Biggest alignment supported by the object file format of this machine.
1054 Use this macro to limit the alignment which can be specified using the
1055 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1056 the default value is @code{BIGGEST_ALIGNMENT}.
1058 On systems that use ELF, the default (in @file{config/elfos.h}) is
1059 the largest supported 32-bit ELF section alignment representable on
1060 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1061 On 32-bit ELF the largest supported section alignment in bits is
1062 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1065 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1066 If defined, a C expression to compute the alignment for a variable in
1067 the static store. @var{type} is the data type, and @var{basic-align} is
1068 the alignment that the object would ordinarily have. The value of this
1069 macro is used instead of that alignment to align the object.
1071 If this macro is not defined, then @var{basic-align} is used.
1074 One use of this macro is to increase alignment of medium-size data to
1075 make it all fit in fewer cache lines. Another is to cause character
1076 arrays to be word-aligned so that @code{strcpy} calls that copy
1077 constants to character arrays can be done inline.
1080 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1081 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1082 some alignment increase, instead of optimization only purposes. E.g.@
1083 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1084 must be aligned to 16 byte boundaries.
1086 If this macro is not defined, then @var{basic-align} is used.
1089 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1090 If defined, a C expression to compute the alignment given to a constant
1091 that is being placed in memory. @var{constant} is the constant and
1092 @var{basic-align} is the alignment that the object would ordinarily
1093 have. The value of this macro is used instead of that alignment to
1096 If this macro is not defined, then @var{basic-align} is used.
1098 The typical use of this macro is to increase alignment for string
1099 constants to be word aligned so that @code{strcpy} calls that copy
1100 constants can be done inline.
1103 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1104 If defined, a C expression to compute the alignment for a variable in
1105 the local store. @var{type} is the data type, and @var{basic-align} is
1106 the alignment that the object would ordinarily have. The value of this
1107 macro is used instead of that alignment to align the object.
1109 If this macro is not defined, then @var{basic-align} is used.
1111 One use of this macro is to increase alignment of medium-size data to
1112 make it all fit in fewer cache lines.
1114 If the value of this macro has a type, it should be an unsigned type.
1117 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_VECTOR_ALIGNMENT (const_tree @var{type})
1118 This hook can be used to define the alignment for a vector of type
1119 @var{type}, in order to comply with a platform ABI. The default is to
1120 require natural alignment for vector types. The alignment returned by
1121 this hook must be a power-of-two multiple of the default alignment of
1122 the vector element type.
1125 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1126 If defined, a C expression to compute the alignment for stack slot.
1127 @var{type} is the data type, @var{mode} is the widest mode available,
1128 and @var{basic-align} is the alignment that the slot would ordinarily
1129 have. The value of this macro is used instead of that alignment to
1132 If this macro is not defined, then @var{basic-align} is used when
1133 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1136 This macro is to set alignment of stack slot to the maximum alignment
1137 of all possible modes which the slot may have.
1139 If the value of this macro has a type, it should be an unsigned type.
1142 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1143 If defined, a C expression to compute the alignment for a local
1144 variable @var{decl}.
1146 If this macro is not defined, then
1147 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1150 One use of this macro is to increase alignment of medium-size data to
1151 make it all fit in fewer cache lines.
1153 If the value of this macro has a type, it should be an unsigned type.
1156 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1157 If defined, a C expression to compute the minimum required alignment
1158 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1159 @var{mode}, assuming normal alignment @var{align}.
1161 If this macro is not defined, then @var{align} will be used.
1164 @defmac EMPTY_FIELD_BOUNDARY
1165 Alignment in bits to be given to a structure bit-field that follows an
1166 empty field such as @code{int : 0;}.
1168 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1171 @defmac STRUCTURE_SIZE_BOUNDARY
1172 Number of bits which any structure or union's size must be a multiple of.
1173 Each structure or union's size is rounded up to a multiple of this.
1175 If you do not define this macro, the default is the same as
1176 @code{BITS_PER_UNIT}.
1179 @defmac STRICT_ALIGNMENT
1180 Define this macro to be the value 1 if instructions will fail to work
1181 if given data not on the nominal alignment. If instructions will merely
1182 go slower in that case, define this macro as 0.
1185 @defmac PCC_BITFIELD_TYPE_MATTERS
1186 Define this if you wish to imitate the way many other C compilers handle
1187 alignment of bit-fields and the structures that contain them.
1189 The behavior is that the type written for a named bit-field (@code{int},
1190 @code{short}, or other integer type) imposes an alignment for the entire
1191 structure, as if the structure really did contain an ordinary field of
1192 that type. In addition, the bit-field is placed within the structure so
1193 that it would fit within such a field, not crossing a boundary for it.
1195 Thus, on most machines, a named bit-field whose type is written as
1196 @code{int} would not cross a four-byte boundary, and would force
1197 four-byte alignment for the whole structure. (The alignment used may
1198 not be four bytes; it is controlled by the other alignment parameters.)
1200 An unnamed bit-field will not affect the alignment of the containing
1203 If the macro is defined, its definition should be a C expression;
1204 a nonzero value for the expression enables this behavior.
1206 Note that if this macro is not defined, or its value is zero, some
1207 bit-fields may cross more than one alignment boundary. The compiler can
1208 support such references if there are @samp{insv}, @samp{extv}, and
1209 @samp{extzv} insns that can directly reference memory.
1211 The other known way of making bit-fields work is to define
1212 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1213 Then every structure can be accessed with fullwords.
1215 Unless the machine has bit-field instructions or you define
1216 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1217 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1219 If your aim is to make GCC use the same conventions for laying out
1220 bit-fields as are used by another compiler, here is how to investigate
1221 what the other compiler does. Compile and run this program:
1240 printf ("Size of foo1 is %d\n",
1241 sizeof (struct foo1));
1242 printf ("Size of foo2 is %d\n",
1243 sizeof (struct foo2));
1248 If this prints 2 and 5, then the compiler's behavior is what you would
1249 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1252 @defmac BITFIELD_NBYTES_LIMITED
1253 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1254 to aligning a bit-field within the structure.
1257 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1258 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1259 whether unnamed bitfields affect the alignment of the containing
1260 structure. The hook should return true if the structure should inherit
1261 the alignment requirements of an unnamed bitfield's type.
1264 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1265 This target hook should return @code{true} if accesses to volatile bitfields
1266 should use the narrowest mode possible. It should return @code{false} if
1267 these accesses should use the bitfield container type.
1269 The default is @code{false}.
1272 @deftypefn {Target Hook} bool TARGET_MEMBER_TYPE_FORCES_BLK (const_tree @var{field}, enum machine_mode @var{mode})
1273 Return true if a structure, union or array containing @var{field} should
1274 be accessed using @code{BLKMODE}.
1276 If @var{field} is the only field in the structure, @var{mode} is its
1277 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1278 case where structures of one field would require the structure's mode to
1279 retain the field's mode.
1281 Normally, this is not needed.
1284 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1285 Define this macro as an expression for the alignment of a type (given
1286 by @var{type} as a tree node) if the alignment computed in the usual
1287 way is @var{computed} and the alignment explicitly specified was
1290 The default is to use @var{specified} if it is larger; otherwise, use
1291 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1294 @defmac MAX_FIXED_MODE_SIZE
1295 An integer expression for the size in bits of the largest integer
1296 machine mode that should actually be used. All integer machine modes of
1297 this size or smaller can be used for structures and unions with the
1298 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1299 (DImode)} is assumed.
1302 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1303 If defined, an expression of type @code{enum machine_mode} that
1304 specifies the mode of the save area operand of a
1305 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1306 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1307 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1308 having its mode specified.
1310 You need not define this macro if it always returns @code{Pmode}. You
1311 would most commonly define this macro if the
1312 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1316 @defmac STACK_SIZE_MODE
1317 If defined, an expression of type @code{enum machine_mode} that
1318 specifies the mode of the size increment operand of an
1319 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1321 You need not define this macro if it always returns @code{word_mode}.
1322 You would most commonly define this macro if the @code{allocate_stack}
1323 pattern needs to support both a 32- and a 64-bit mode.
1326 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE (void)
1327 This target hook should return the mode to be used for the return value
1328 of compare instructions expanded to libgcc calls. If not defined
1329 @code{word_mode} is returned which is the right choice for a majority of
1333 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1334 This target hook should return the mode to be used for the shift count operand
1335 of shift instructions expanded to libgcc calls. If not defined
1336 @code{word_mode} is returned which is the right choice for a majority of
1340 @deftypefn {Target Hook} {enum machine_mode} TARGET_UNWIND_WORD_MODE (void)
1341 Return machine mode to be used for @code{_Unwind_Word} type.
1342 The default is to use @code{word_mode}.
1345 @defmac ROUND_TOWARDS_ZERO
1346 If defined, this macro should be true if the prevailing rounding
1347 mode is towards zero.
1349 Defining this macro only affects the way @file{libgcc.a} emulates
1350 floating-point arithmetic.
1352 Not defining this macro is equivalent to returning zero.
1355 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1356 This macro should return true if floats with @var{size}
1357 bits do not have a NaN or infinity representation, but use the largest
1358 exponent for normal numbers instead.
1360 Defining this macro only affects the way @file{libgcc.a} emulates
1361 floating-point arithmetic.
1363 The default definition of this macro returns false for all sizes.
1366 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1367 This target hook returns @code{true} if bit-fields in the given
1368 @var{record_type} are to be laid out following the rules of Microsoft
1369 Visual C/C++, namely: (i) a bit-field won't share the same storage
1370 unit with the previous bit-field if their underlying types have
1371 different sizes, and the bit-field will be aligned to the highest
1372 alignment of the underlying types of itself and of the previous
1373 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1374 the whole enclosing structure, even if it is unnamed; except that
1375 (iii) a zero-sized bit-field will be disregarded unless it follows
1376 another bit-field of nonzero size. If this hook returns @code{true},
1377 other macros that control bit-field layout are ignored.
1379 When a bit-field is inserted into a packed record, the whole size
1380 of the underlying type is used by one or more same-size adjacent
1381 bit-fields (that is, if its long:3, 32 bits is used in the record,
1382 and any additional adjacent long bit-fields are packed into the same
1383 chunk of 32 bits. However, if the size changes, a new field of that
1384 size is allocated). In an unpacked record, this is the same as using
1385 alignment, but not equivalent when packing.
1387 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1388 the latter will take precedence. If @samp{__attribute__((packed))} is
1389 used on a single field when MS bit-fields are in use, it will take
1390 precedence for that field, but the alignment of the rest of the structure
1391 may affect its placement.
1394 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1395 Returns true if the target supports decimal floating point.
1398 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1399 Returns true if the target supports fixed-point arithmetic.
1402 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1403 This hook is called just before expansion into rtl, allowing the target
1404 to perform additional initializations or analysis before the expansion.
1405 For example, the rs6000 port uses it to allocate a scratch stack slot
1406 for use in copying SDmode values between memory and floating point
1407 registers whenever the function being expanded has any SDmode
1411 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1412 This hook allows the backend to perform additional instantiations on rtl
1413 that are not actually in any insns yet, but will be later.
1416 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1417 If your target defines any fundamental types, or any types your target
1418 uses should be mangled differently from the default, define this hook
1419 to return the appropriate encoding for these types as part of a C++
1420 mangled name. The @var{type} argument is the tree structure representing
1421 the type to be mangled. The hook may be applied to trees which are
1422 not target-specific fundamental types; it should return @code{NULL}
1423 for all such types, as well as arguments it does not recognize. If the
1424 return value is not @code{NULL}, it must point to a statically-allocated
1427 Target-specific fundamental types might be new fundamental types or
1428 qualified versions of ordinary fundamental types. Encode new
1429 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1430 is the name used for the type in source code, and @var{n} is the
1431 length of @var{name} in decimal. Encode qualified versions of
1432 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1433 @var{name} is the name used for the type qualifier in source code,
1434 @var{n} is the length of @var{name} as above, and @var{code} is the
1435 code used to represent the unqualified version of this type. (See
1436 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1437 codes.) In both cases the spaces are for clarity; do not include any
1438 spaces in your string.
1440 This hook is applied to types prior to typedef resolution. If the mangled
1441 name for a particular type depends only on that type's main variant, you
1442 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1445 The default version of this hook always returns @code{NULL}, which is
1446 appropriate for a target that does not define any new fundamental
1451 @section Layout of Source Language Data Types
1453 These macros define the sizes and other characteristics of the standard
1454 basic data types used in programs being compiled. Unlike the macros in
1455 the previous section, these apply to specific features of C and related
1456 languages, rather than to fundamental aspects of storage layout.
1458 @defmac INT_TYPE_SIZE
1459 A C expression for the size in bits of the type @code{int} on the
1460 target machine. If you don't define this, the default is one word.
1463 @defmac SHORT_TYPE_SIZE
1464 A C expression for the size in bits of the type @code{short} on the
1465 target machine. If you don't define this, the default is half a word.
1466 (If this would be less than one storage unit, it is rounded up to one
1470 @defmac LONG_TYPE_SIZE
1471 A C expression for the size in bits of the type @code{long} on the
1472 target machine. If you don't define this, the default is one word.
1475 @defmac ADA_LONG_TYPE_SIZE
1476 On some machines, the size used for the Ada equivalent of the type
1477 @code{long} by a native Ada compiler differs from that used by C@. In
1478 that situation, define this macro to be a C expression to be used for
1479 the size of that type. If you don't define this, the default is the
1480 value of @code{LONG_TYPE_SIZE}.
1483 @defmac LONG_LONG_TYPE_SIZE
1484 A C expression for the size in bits of the type @code{long long} on the
1485 target machine. If you don't define this, the default is two
1486 words. If you want to support GNU Ada on your machine, the value of this
1487 macro must be at least 64.
1490 @defmac CHAR_TYPE_SIZE
1491 A C expression for the size in bits of the type @code{char} on the
1492 target machine. If you don't define this, the default is
1493 @code{BITS_PER_UNIT}.
1496 @defmac BOOL_TYPE_SIZE
1497 A C expression for the size in bits of the C++ type @code{bool} and
1498 C99 type @code{_Bool} on the target machine. If you don't define
1499 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1502 @defmac FLOAT_TYPE_SIZE
1503 A C expression for the size in bits of the type @code{float} on the
1504 target machine. If you don't define this, the default is one word.
1507 @defmac DOUBLE_TYPE_SIZE
1508 A C expression for the size in bits of the type @code{double} on the
1509 target machine. If you don't define this, the default is two
1513 @defmac LONG_DOUBLE_TYPE_SIZE
1514 A C expression for the size in bits of the type @code{long double} on
1515 the target machine. If you don't define this, the default is two
1519 @defmac SHORT_FRACT_TYPE_SIZE
1520 A C expression for the size in bits of the type @code{short _Fract} on
1521 the target machine. If you don't define this, the default is
1522 @code{BITS_PER_UNIT}.
1525 @defmac FRACT_TYPE_SIZE
1526 A C expression for the size in bits of the type @code{_Fract} on
1527 the target machine. If you don't define this, the default is
1528 @code{BITS_PER_UNIT * 2}.
1531 @defmac LONG_FRACT_TYPE_SIZE
1532 A C expression for the size in bits of the type @code{long _Fract} on
1533 the target machine. If you don't define this, the default is
1534 @code{BITS_PER_UNIT * 4}.
1537 @defmac LONG_LONG_FRACT_TYPE_SIZE
1538 A C expression for the size in bits of the type @code{long long _Fract} on
1539 the target machine. If you don't define this, the default is
1540 @code{BITS_PER_UNIT * 8}.
1543 @defmac SHORT_ACCUM_TYPE_SIZE
1544 A C expression for the size in bits of the type @code{short _Accum} on
1545 the target machine. If you don't define this, the default is
1546 @code{BITS_PER_UNIT * 2}.
1549 @defmac ACCUM_TYPE_SIZE
1550 A C expression for the size in bits of the type @code{_Accum} on
1551 the target machine. If you don't define this, the default is
1552 @code{BITS_PER_UNIT * 4}.
1555 @defmac LONG_ACCUM_TYPE_SIZE
1556 A C expression for the size in bits of the type @code{long _Accum} on
1557 the target machine. If you don't define this, the default is
1558 @code{BITS_PER_UNIT * 8}.
1561 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1562 A C expression for the size in bits of the type @code{long long _Accum} on
1563 the target machine. If you don't define this, the default is
1564 @code{BITS_PER_UNIT * 16}.
1567 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1568 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1569 if you want routines in @file{libgcc2.a} for a size other than
1570 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1571 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1574 @defmac LIBGCC2_HAS_DF_MODE
1575 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1576 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1577 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1578 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1579 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1583 @defmac LIBGCC2_HAS_XF_MODE
1584 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1585 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1586 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1587 is 80 then the default is 1, otherwise it is 0.
1590 @defmac LIBGCC2_HAS_TF_MODE
1591 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1592 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1593 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1594 is 128 then the default is 1, otherwise it is 0.
1597 @defmac LIBGCC2_GNU_PREFIX
1598 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1599 hook and should be defined if that hook is overriden to be true. It
1600 causes function names in libgcc to be changed to use a @code{__gnu_}
1601 prefix for their name rather than the default @code{__}. A port which
1602 uses this macro should also arrange to use @file{t-gnu-prefix} in
1603 the libgcc @file{config.host}.
1610 Define these macros to be the size in bits of the mantissa of
1611 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1612 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1613 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1614 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1615 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1616 @code{DOUBLE_TYPE_SIZE} or
1617 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1620 @defmac TARGET_FLT_EVAL_METHOD
1621 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1622 assuming, if applicable, that the floating-point control word is in its
1623 default state. If you do not define this macro the value of
1624 @code{FLT_EVAL_METHOD} will be zero.
1627 @defmac WIDEST_HARDWARE_FP_SIZE
1628 A C expression for the size in bits of the widest floating-point format
1629 supported by the hardware. If you define this macro, you must specify a
1630 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1631 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1635 @defmac DEFAULT_SIGNED_CHAR
1636 An expression whose value is 1 or 0, according to whether the type
1637 @code{char} should be signed or unsigned by default. The user can
1638 always override this default with the options @option{-fsigned-char}
1639 and @option{-funsigned-char}.
1642 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1643 This target hook should return true if the compiler should give an
1644 @code{enum} type only as many bytes as it takes to represent the range
1645 of possible values of that type. It should return false if all
1646 @code{enum} types should be allocated like @code{int}.
1648 The default is to return false.
1652 A C expression for a string describing the name of the data type to use
1653 for size values. The typedef name @code{size_t} is defined using the
1654 contents of the string.
1656 The string can contain more than one keyword. If so, separate them with
1657 spaces, and write first any length keyword, then @code{unsigned} if
1658 appropriate, and finally @code{int}. The string must exactly match one
1659 of the data type names defined in the function
1660 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1661 You may not omit @code{int} or change the order---that would cause the
1662 compiler to crash on startup.
1664 If you don't define this macro, the default is @code{"long unsigned
1669 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1670 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1671 dealing with size. This macro is a C expression for a string describing
1672 the name of the data type from which the precision of @code{sizetype}
1675 The string has the same restrictions as @code{SIZE_TYPE} string.
1677 If you don't define this macro, the default is @code{SIZE_TYPE}.
1680 @defmac PTRDIFF_TYPE
1681 A C expression for a string describing the name of the data type to use
1682 for the result of subtracting two pointers. The typedef name
1683 @code{ptrdiff_t} is defined using the contents of the string. See
1684 @code{SIZE_TYPE} above for more information.
1686 If you don't define this macro, the default is @code{"long int"}.
1690 A C expression for a string describing the name of the data type to use
1691 for wide characters. The typedef name @code{wchar_t} is defined using
1692 the contents of the string. See @code{SIZE_TYPE} above for more
1695 If you don't define this macro, the default is @code{"int"}.
1698 @defmac WCHAR_TYPE_SIZE
1699 A C expression for the size in bits of the data type for wide
1700 characters. This is used in @code{cpp}, which cannot make use of
1705 A C expression for a string describing the name of the data type to
1706 use for wide characters passed to @code{printf} and returned from
1707 @code{getwc}. The typedef name @code{wint_t} is defined using the
1708 contents of the string. See @code{SIZE_TYPE} above for more
1711 If you don't define this macro, the default is @code{"unsigned int"}.
1715 A C expression for a string describing the name of the data type that
1716 can represent any value of any standard or extended signed integer type.
1717 The typedef name @code{intmax_t} is defined using the contents of the
1718 string. See @code{SIZE_TYPE} above for more information.
1720 If you don't define this macro, the default is the first of
1721 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1722 much precision as @code{long long int}.
1725 @defmac UINTMAX_TYPE
1726 A C expression for a string describing the name of the data type that
1727 can represent any value of any standard or extended unsigned integer
1728 type. The typedef name @code{uintmax_t} is defined using the contents
1729 of the string. See @code{SIZE_TYPE} above for more information.
1731 If you don't define this macro, the default is the first of
1732 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1733 unsigned int"} that has as much precision as @code{long long unsigned
1737 @defmac SIG_ATOMIC_TYPE
1743 @defmacx UINT16_TYPE
1744 @defmacx UINT32_TYPE
1745 @defmacx UINT64_TYPE
1746 @defmacx INT_LEAST8_TYPE
1747 @defmacx INT_LEAST16_TYPE
1748 @defmacx INT_LEAST32_TYPE
1749 @defmacx INT_LEAST64_TYPE
1750 @defmacx UINT_LEAST8_TYPE
1751 @defmacx UINT_LEAST16_TYPE
1752 @defmacx UINT_LEAST32_TYPE
1753 @defmacx UINT_LEAST64_TYPE
1754 @defmacx INT_FAST8_TYPE
1755 @defmacx INT_FAST16_TYPE
1756 @defmacx INT_FAST32_TYPE
1757 @defmacx INT_FAST64_TYPE
1758 @defmacx UINT_FAST8_TYPE
1759 @defmacx UINT_FAST16_TYPE
1760 @defmacx UINT_FAST32_TYPE
1761 @defmacx UINT_FAST64_TYPE
1762 @defmacx INTPTR_TYPE
1763 @defmacx UINTPTR_TYPE
1764 C expressions for the standard types @code{sig_atomic_t},
1765 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1766 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1767 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1768 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1769 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1770 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1771 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1772 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1773 @code{SIZE_TYPE} above for more information.
1775 If any of these macros evaluates to a null pointer, the corresponding
1776 type is not supported; if GCC is configured to provide
1777 @code{<stdint.h>} in such a case, the header provided may not conform
1778 to C99, depending on the type in question. The defaults for all of
1779 these macros are null pointers.
1782 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1783 The C++ compiler represents a pointer-to-member-function with a struct
1790 ptrdiff_t vtable_index;
1797 The C++ compiler must use one bit to indicate whether the function that
1798 will be called through a pointer-to-member-function is virtual.
1799 Normally, we assume that the low-order bit of a function pointer must
1800 always be zero. Then, by ensuring that the vtable_index is odd, we can
1801 distinguish which variant of the union is in use. But, on some
1802 platforms function pointers can be odd, and so this doesn't work. In
1803 that case, we use the low-order bit of the @code{delta} field, and shift
1804 the remainder of the @code{delta} field to the left.
1806 GCC will automatically make the right selection about where to store
1807 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1808 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1809 set such that functions always start at even addresses, but the lowest
1810 bit of pointers to functions indicate whether the function at that
1811 address is in ARM or Thumb mode. If this is the case of your
1812 architecture, you should define this macro to
1813 @code{ptrmemfunc_vbit_in_delta}.
1815 In general, you should not have to define this macro. On architectures
1816 in which function addresses are always even, according to
1817 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1818 @code{ptrmemfunc_vbit_in_pfn}.
1821 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1822 Normally, the C++ compiler uses function pointers in vtables. This
1823 macro allows the target to change to use ``function descriptors''
1824 instead. Function descriptors are found on targets for whom a
1825 function pointer is actually a small data structure. Normally the
1826 data structure consists of the actual code address plus a data
1827 pointer to which the function's data is relative.
1829 If vtables are used, the value of this macro should be the number
1830 of words that the function descriptor occupies.
1833 @defmac TARGET_VTABLE_ENTRY_ALIGN
1834 By default, the vtable entries are void pointers, the so the alignment
1835 is the same as pointer alignment. The value of this macro specifies
1836 the alignment of the vtable entry in bits. It should be defined only
1837 when special alignment is necessary. */
1840 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1841 There are a few non-descriptor entries in the vtable at offsets below
1842 zero. If these entries must be padded (say, to preserve the alignment
1843 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1844 of words in each data entry.
1848 @section Register Usage
1849 @cindex register usage
1851 This section explains how to describe what registers the target machine
1852 has, and how (in general) they can be used.
1854 The description of which registers a specific instruction can use is
1855 done with register classes; see @ref{Register Classes}. For information
1856 on using registers to access a stack frame, see @ref{Frame Registers}.
1857 For passing values in registers, see @ref{Register Arguments}.
1858 For returning values in registers, see @ref{Scalar Return}.
1861 * Register Basics:: Number and kinds of registers.
1862 * Allocation Order:: Order in which registers are allocated.
1863 * Values in Registers:: What kinds of values each reg can hold.
1864 * Leaf Functions:: Renumbering registers for leaf functions.
1865 * Stack Registers:: Handling a register stack such as 80387.
1868 @node Register Basics
1869 @subsection Basic Characteristics of Registers
1871 @c prevent bad page break with this line
1872 Registers have various characteristics.
1874 @defmac FIRST_PSEUDO_REGISTER
1875 Number of hardware registers known to the compiler. They receive
1876 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1877 pseudo register's number really is assigned the number
1878 @code{FIRST_PSEUDO_REGISTER}.
1881 @defmac FIXED_REGISTERS
1882 @cindex fixed register
1883 An initializer that says which registers are used for fixed purposes
1884 all throughout the compiled code and are therefore not available for
1885 general allocation. These would include the stack pointer, the frame
1886 pointer (except on machines where that can be used as a general
1887 register when no frame pointer is needed), the program counter on
1888 machines where that is considered one of the addressable registers,
1889 and any other numbered register with a standard use.
1891 This information is expressed as a sequence of numbers, separated by
1892 commas and surrounded by braces. The @var{n}th number is 1 if
1893 register @var{n} is fixed, 0 otherwise.
1895 The table initialized from this macro, and the table initialized by
1896 the following one, may be overridden at run time either automatically,
1897 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1898 the user with the command options @option{-ffixed-@var{reg}},
1899 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1902 @defmac CALL_USED_REGISTERS
1903 @cindex call-used register
1904 @cindex call-clobbered register
1905 @cindex call-saved register
1906 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1907 clobbered (in general) by function calls as well as for fixed
1908 registers. This macro therefore identifies the registers that are not
1909 available for general allocation of values that must live across
1912 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1913 automatically saves it on function entry and restores it on function
1914 exit, if the register is used within the function.
1917 @defmac CALL_REALLY_USED_REGISTERS
1918 @cindex call-used register
1919 @cindex call-clobbered register
1920 @cindex call-saved register
1921 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1922 that the entire set of @code{FIXED_REGISTERS} be included.
1923 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1924 This macro is optional. If not specified, it defaults to the value
1925 of @code{CALL_USED_REGISTERS}.
1928 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1929 @cindex call-used register
1930 @cindex call-clobbered register
1931 @cindex call-saved register
1932 A C expression that is nonzero if it is not permissible to store a
1933 value of mode @var{mode} in hard register number @var{regno} across a
1934 call without some part of it being clobbered. For most machines this
1935 macro need not be defined. It is only required for machines that do not
1936 preserve the entire contents of a register across a call.
1940 @findex call_used_regs
1943 @findex reg_class_contents
1944 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1945 This hook may conditionally modify five variables
1946 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1947 @code{reg_names}, and @code{reg_class_contents}, to take into account
1948 any dependence of these register sets on target flags. The first three
1949 of these are of type @code{char []} (interpreted as Boolean vectors).
1950 @code{global_regs} is a @code{const char *[]}, and
1951 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1952 called, @code{fixed_regs}, @code{call_used_regs},
1953 @code{reg_class_contents}, and @code{reg_names} have been initialized
1954 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1955 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1956 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1957 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1958 command options have been applied.
1960 @cindex disabling certain registers
1961 @cindex controlling register usage
1962 If the usage of an entire class of registers depends on the target
1963 flags, you may indicate this to GCC by using this macro to modify
1964 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1965 registers in the classes which should not be used by GCC@. Also define
1966 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1967 to return @code{NO_REGS} if it
1968 is called with a letter for a class that shouldn't be used.
1970 (However, if this class is not included in @code{GENERAL_REGS} and all
1971 of the insn patterns whose constraints permit this class are
1972 controlled by target switches, then GCC will automatically avoid using
1973 these registers when the target switches are opposed to them.)
1976 @defmac INCOMING_REGNO (@var{out})
1977 Define this macro if the target machine has register windows. This C
1978 expression returns the register number as seen by the called function
1979 corresponding to the register number @var{out} as seen by the calling
1980 function. Return @var{out} if register number @var{out} is not an
1984 @defmac OUTGOING_REGNO (@var{in})
1985 Define this macro if the target machine has register windows. This C
1986 expression returns the register number as seen by the calling function
1987 corresponding to the register number @var{in} as seen by the called
1988 function. Return @var{in} if register number @var{in} is not an inbound
1992 @defmac LOCAL_REGNO (@var{regno})
1993 Define this macro if the target machine has register windows. This C
1994 expression returns true if the register is call-saved but is in the
1995 register window. Unlike most call-saved registers, such registers
1996 need not be explicitly restored on function exit or during non-local
2001 If the program counter has a register number, define this as that
2002 register number. Otherwise, do not define it.
2005 @node Allocation Order
2006 @subsection Order of Allocation of Registers
2007 @cindex order of register allocation
2008 @cindex register allocation order
2010 @c prevent bad page break with this line
2011 Registers are allocated in order.
2013 @defmac REG_ALLOC_ORDER
2014 If defined, an initializer for a vector of integers, containing the
2015 numbers of hard registers in the order in which GCC should prefer
2016 to use them (from most preferred to least).
2018 If this macro is not defined, registers are used lowest numbered first
2019 (all else being equal).
2021 One use of this macro is on machines where the highest numbered
2022 registers must always be saved and the save-multiple-registers
2023 instruction supports only sequences of consecutive registers. On such
2024 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2025 the highest numbered allocable register first.
2028 @defmac ADJUST_REG_ALLOC_ORDER
2029 A C statement (sans semicolon) to choose the order in which to allocate
2030 hard registers for pseudo-registers local to a basic block.
2032 Store the desired register order in the array @code{reg_alloc_order}.
2033 Element 0 should be the register to allocate first; element 1, the next
2034 register; and so on.
2036 The macro body should not assume anything about the contents of
2037 @code{reg_alloc_order} before execution of the macro.
2039 On most machines, it is not necessary to define this macro.
2042 @defmac HONOR_REG_ALLOC_ORDER
2043 Normally, IRA tries to estimate the costs for saving a register in the
2044 prologue and restoring it in the epilogue. This discourages it from
2045 using call-saved registers. If a machine wants to ensure that IRA
2046 allocates registers in the order given by REG_ALLOC_ORDER even if some
2047 call-saved registers appear earlier than call-used ones, this macro
2051 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2052 In some case register allocation order is not enough for the
2053 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2054 If this macro is defined, it should return a floating point value
2055 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2056 be increased by approximately the pseudo's usage frequency times the
2057 value returned by this macro. Not defining this macro is equivalent
2058 to having it always return @code{0.0}.
2060 On most machines, it is not necessary to define this macro.
2063 @node Values in Registers
2064 @subsection How Values Fit in Registers
2066 This section discusses the macros that describe which kinds of values
2067 (specifically, which machine modes) each register can hold, and how many
2068 consecutive registers are needed for a given mode.
2070 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2071 A C expression for the number of consecutive hard registers, starting
2072 at register number @var{regno}, required to hold a value of mode
2073 @var{mode}. This macro must never return zero, even if a register
2074 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2075 and/or CANNOT_CHANGE_MODE_CLASS instead.
2077 On a machine where all registers are exactly one word, a suitable
2078 definition of this macro is
2081 #define HARD_REGNO_NREGS(REGNO, MODE) \
2082 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2087 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2088 A C expression that is nonzero if a value of mode @var{mode}, stored
2089 in memory, ends with padding that causes it to take up more space than
2090 in registers starting at register number @var{regno} (as determined by
2091 multiplying GCC's notion of the size of the register when containing
2092 this mode by the number of registers returned by
2093 @code{HARD_REGNO_NREGS}). By default this is zero.
2095 For example, if a floating-point value is stored in three 32-bit
2096 registers but takes up 128 bits in memory, then this would be
2099 This macros only needs to be defined if there are cases where
2100 @code{subreg_get_info}
2101 would otherwise wrongly determine that a @code{subreg} can be
2102 represented by an offset to the register number, when in fact such a
2103 @code{subreg} would contain some of the padding not stored in
2104 registers and so not be representable.
2107 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2108 For values of @var{regno} and @var{mode} for which
2109 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2110 returning the greater number of registers required to hold the value
2111 including any padding. In the example above, the value would be four.
2114 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2115 Define this macro if the natural size of registers that hold values
2116 of mode @var{mode} is not the word size. It is a C expression that
2117 should give the natural size in bytes for the specified mode. It is
2118 used by the register allocator to try to optimize its results. This
2119 happens for example on SPARC 64-bit where the natural size of
2120 floating-point registers is still 32-bit.
2123 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2124 A C expression that is nonzero if it is permissible to store a value
2125 of mode @var{mode} in hard register number @var{regno} (or in several
2126 registers starting with that one). For a machine where all registers
2127 are equivalent, a suitable definition is
2130 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2133 You need not include code to check for the numbers of fixed registers,
2134 because the allocation mechanism considers them to be always occupied.
2136 @cindex register pairs
2137 On some machines, double-precision values must be kept in even/odd
2138 register pairs. You can implement that by defining this macro to reject
2139 odd register numbers for such modes.
2141 The minimum requirement for a mode to be OK in a register is that the
2142 @samp{mov@var{mode}} instruction pattern support moves between the
2143 register and other hard register in the same class and that moving a
2144 value into the register and back out not alter it.
2146 Since the same instruction used to move @code{word_mode} will work for
2147 all narrower integer modes, it is not necessary on any machine for
2148 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2149 you define patterns @samp{movhi}, etc., to take advantage of this. This
2150 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2151 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2154 Many machines have special registers for floating point arithmetic.
2155 Often people assume that floating point machine modes are allowed only
2156 in floating point registers. This is not true. Any registers that
2157 can hold integers can safely @emph{hold} a floating point machine
2158 mode, whether or not floating arithmetic can be done on it in those
2159 registers. Integer move instructions can be used to move the values.
2161 On some machines, though, the converse is true: fixed-point machine
2162 modes may not go in floating registers. This is true if the floating
2163 registers normalize any value stored in them, because storing a
2164 non-floating value there would garble it. In this case,
2165 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2166 floating registers. But if the floating registers do not automatically
2167 normalize, if you can store any bit pattern in one and retrieve it
2168 unchanged without a trap, then any machine mode may go in a floating
2169 register, so you can define this macro to say so.
2171 The primary significance of special floating registers is rather that
2172 they are the registers acceptable in floating point arithmetic
2173 instructions. However, this is of no concern to
2174 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2175 constraints for those instructions.
2177 On some machines, the floating registers are especially slow to access,
2178 so that it is better to store a value in a stack frame than in such a
2179 register if floating point arithmetic is not being done. As long as the
2180 floating registers are not in class @code{GENERAL_REGS}, they will not
2181 be used unless some pattern's constraint asks for one.
2184 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2185 A C expression that is nonzero if it is OK to rename a hard register
2186 @var{from} to another hard register @var{to}.
2188 One common use of this macro is to prevent renaming of a register to
2189 another register that is not saved by a prologue in an interrupt
2192 The default is always nonzero.
2195 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2196 A C expression that is nonzero if a value of mode
2197 @var{mode1} is accessible in mode @var{mode2} without copying.
2199 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2200 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2201 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2202 should be nonzero. If they differ for any @var{r}, you should define
2203 this macro to return zero unless some other mechanism ensures the
2204 accessibility of the value in a narrower mode.
2206 You should define this macro to return nonzero in as many cases as
2207 possible since doing so will allow GCC to perform better register
2211 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2212 This target hook should return @code{true} if it is OK to use a hard register
2213 @var{regno} as scratch reg in peephole2.
2215 One common use of this macro is to prevent using of a register that
2216 is not saved by a prologue in an interrupt handler.
2218 The default version of this hook always returns @code{true}.
2221 @defmac AVOID_CCMODE_COPIES
2222 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2223 registers. You should only define this macro if support for copying to/from
2224 @code{CCmode} is incomplete.
2227 @node Leaf Functions
2228 @subsection Handling Leaf Functions
2230 @cindex leaf functions
2231 @cindex functions, leaf
2232 On some machines, a leaf function (i.e., one which makes no calls) can run
2233 more efficiently if it does not make its own register window. Often this
2234 means it is required to receive its arguments in the registers where they
2235 are passed by the caller, instead of the registers where they would
2238 The special treatment for leaf functions generally applies only when
2239 other conditions are met; for example, often they may use only those
2240 registers for its own variables and temporaries. We use the term ``leaf
2241 function'' to mean a function that is suitable for this special
2242 handling, so that functions with no calls are not necessarily ``leaf
2245 GCC assigns register numbers before it knows whether the function is
2246 suitable for leaf function treatment. So it needs to renumber the
2247 registers in order to output a leaf function. The following macros
2250 @defmac LEAF_REGISTERS
2251 Name of a char vector, indexed by hard register number, which
2252 contains 1 for a register that is allowable in a candidate for leaf
2255 If leaf function treatment involves renumbering the registers, then the
2256 registers marked here should be the ones before renumbering---those that
2257 GCC would ordinarily allocate. The registers which will actually be
2258 used in the assembler code, after renumbering, should not be marked with 1
2261 Define this macro only if the target machine offers a way to optimize
2262 the treatment of leaf functions.
2265 @defmac LEAF_REG_REMAP (@var{regno})
2266 A C expression whose value is the register number to which @var{regno}
2267 should be renumbered, when a function is treated as a leaf function.
2269 If @var{regno} is a register number which should not appear in a leaf
2270 function before renumbering, then the expression should yield @minus{}1, which
2271 will cause the compiler to abort.
2273 Define this macro only if the target machine offers a way to optimize the
2274 treatment of leaf functions, and registers need to be renumbered to do
2278 @findex current_function_is_leaf
2279 @findex current_function_uses_only_leaf_regs
2280 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2281 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2282 specially. They can test the C variable @code{current_function_is_leaf}
2283 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2284 set prior to local register allocation and is valid for the remaining
2285 compiler passes. They can also test the C variable
2286 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2287 functions which only use leaf registers.
2288 @code{current_function_uses_only_leaf_regs} is valid after all passes
2289 that modify the instructions have been run and is only useful if
2290 @code{LEAF_REGISTERS} is defined.
2291 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2292 @c of the next paragraph?! --mew 2feb93
2294 @node Stack Registers
2295 @subsection Registers That Form a Stack
2297 There are special features to handle computers where some of the
2298 ``registers'' form a stack. Stack registers are normally written by
2299 pushing onto the stack, and are numbered relative to the top of the
2302 Currently, GCC can only handle one group of stack-like registers, and
2303 they must be consecutively numbered. Furthermore, the existing
2304 support for stack-like registers is specific to the 80387 floating
2305 point coprocessor. If you have a new architecture that uses
2306 stack-like registers, you will need to do substantial work on
2307 @file{reg-stack.c} and write your machine description to cooperate
2308 with it, as well as defining these macros.
2311 Define this if the machine has any stack-like registers.
2314 @defmac STACK_REG_COVER_CLASS
2315 This is a cover class containing the stack registers. Define this if
2316 the machine has any stack-like registers.
2319 @defmac FIRST_STACK_REG
2320 The number of the first stack-like register. This one is the top
2324 @defmac LAST_STACK_REG
2325 The number of the last stack-like register. This one is the bottom of
2329 @node Register Classes
2330 @section Register Classes
2331 @cindex register class definitions
2332 @cindex class definitions, register
2334 On many machines, the numbered registers are not all equivalent.
2335 For example, certain registers may not be allowed for indexed addressing;
2336 certain registers may not be allowed in some instructions. These machine
2337 restrictions are described to the compiler using @dfn{register classes}.
2339 You define a number of register classes, giving each one a name and saying
2340 which of the registers belong to it. Then you can specify register classes
2341 that are allowed as operands to particular instruction patterns.
2345 In general, each register will belong to several classes. In fact, one
2346 class must be named @code{ALL_REGS} and contain all the registers. Another
2347 class must be named @code{NO_REGS} and contain no registers. Often the
2348 union of two classes will be another class; however, this is not required.
2350 @findex GENERAL_REGS
2351 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2352 terribly special about the name, but the operand constraint letters
2353 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2354 the same as @code{ALL_REGS}, just define it as a macro which expands
2357 Order the classes so that if class @var{x} is contained in class @var{y}
2358 then @var{x} has a lower class number than @var{y}.
2360 The way classes other than @code{GENERAL_REGS} are specified in operand
2361 constraints is through machine-dependent operand constraint letters.
2362 You can define such letters to correspond to various classes, then use
2363 them in operand constraints.
2365 You must define the narrowest register classes for allocatable
2366 registers, so that each class either has no subclasses, or that for
2367 some mode, the move cost between registers within the class is
2368 cheaper than moving a register in the class to or from memory
2371 You should define a class for the union of two classes whenever some
2372 instruction allows both classes. For example, if an instruction allows
2373 either a floating point (coprocessor) register or a general register for a
2374 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2375 which includes both of them. Otherwise you will get suboptimal code,
2376 or even internal compiler errors when reload cannot find a register in the
2377 class computed via @code{reg_class_subunion}.
2379 You must also specify certain redundant information about the register
2380 classes: for each class, which classes contain it and which ones are
2381 contained in it; for each pair of classes, the largest class contained
2384 When a value occupying several consecutive registers is expected in a
2385 certain class, all the registers used must belong to that class.
2386 Therefore, register classes cannot be used to enforce a requirement for
2387 a register pair to start with an even-numbered register. The way to
2388 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2390 Register classes used for input-operands of bitwise-and or shift
2391 instructions have a special requirement: each such class must have, for
2392 each fixed-point machine mode, a subclass whose registers can transfer that
2393 mode to or from memory. For example, on some machines, the operations for
2394 single-byte values (@code{QImode}) are limited to certain registers. When
2395 this is so, each register class that is used in a bitwise-and or shift
2396 instruction must have a subclass consisting of registers from which
2397 single-byte values can be loaded or stored. This is so that
2398 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2400 @deftp {Data type} {enum reg_class}
2401 An enumerated type that must be defined with all the register class names
2402 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2403 must be the last register class, followed by one more enumerated value,
2404 @code{LIM_REG_CLASSES}, which is not a register class but rather
2405 tells how many classes there are.
2407 Each register class has a number, which is the value of casting
2408 the class name to type @code{int}. The number serves as an index
2409 in many of the tables described below.
2412 @defmac N_REG_CLASSES
2413 The number of distinct register classes, defined as follows:
2416 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2420 @defmac REG_CLASS_NAMES
2421 An initializer containing the names of the register classes as C string
2422 constants. These names are used in writing some of the debugging dumps.
2425 @defmac REG_CLASS_CONTENTS
2426 An initializer containing the contents of the register classes, as integers
2427 which are bit masks. The @var{n}th integer specifies the contents of class
2428 @var{n}. The way the integer @var{mask} is interpreted is that
2429 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2431 When the machine has more than 32 registers, an integer does not suffice.
2432 Then the integers are replaced by sub-initializers, braced groupings containing
2433 several integers. Each sub-initializer must be suitable as an initializer
2434 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2435 In this situation, the first integer in each sub-initializer corresponds to
2436 registers 0 through 31, the second integer to registers 32 through 63, and
2440 @defmac REGNO_REG_CLASS (@var{regno})
2441 A C expression whose value is a register class containing hard register
2442 @var{regno}. In general there is more than one such class; choose a class
2443 which is @dfn{minimal}, meaning that no smaller class also contains the
2447 @defmac BASE_REG_CLASS
2448 A macro whose definition is the name of the class to which a valid
2449 base register must belong. A base register is one used in an address
2450 which is the register value plus a displacement.
2453 @defmac MODE_BASE_REG_CLASS (@var{mode})
2454 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2455 the selection of a base register in a mode dependent manner. If
2456 @var{mode} is VOIDmode then it should return the same value as
2457 @code{BASE_REG_CLASS}.
2460 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2461 A C expression whose value is the register class to which a valid
2462 base register must belong in order to be used in a base plus index
2463 register address. You should define this macro if base plus index
2464 addresses have different requirements than other base register uses.
2467 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2468 A C expression whose value is the register class to which a valid
2469 base register for a memory reference in mode @var{mode} to address
2470 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2471 define the context in which the base register occurs. @var{outer_code} is
2472 the code of the immediately enclosing expression (@code{MEM} for the top level
2473 of an address, @code{ADDRESS} for something that occurs in an
2474 @code{address_operand}). @var{index_code} is the code of the corresponding
2475 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2478 @defmac INDEX_REG_CLASS
2479 A macro whose definition is the name of the class to which a valid
2480 index register must belong. An index register is one used in an
2481 address where its value is either multiplied by a scale factor or
2482 added to another register (as well as added to a displacement).
2485 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2486 A C expression which is nonzero if register number @var{num} is
2487 suitable for use as a base register in operand addresses.
2490 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2491 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2492 that expression may examine the mode of the memory reference in
2493 @var{mode}. You should define this macro if the mode of the memory
2494 reference affects whether a register may be used as a base register. If
2495 you define this macro, the compiler will use it instead of
2496 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2497 addresses that appear outside a @code{MEM}, i.e., as an
2498 @code{address_operand}.
2501 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2502 A C expression which is nonzero if register number @var{num} is suitable for
2503 use as a base register in base plus index operand addresses, accessing
2504 memory in mode @var{mode}. It may be either a suitable hard register or a
2505 pseudo register that has been allocated such a hard register. You should
2506 define this macro if base plus index addresses have different requirements
2507 than other base register uses.
2509 Use of this macro is deprecated; please use the more general
2510 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2513 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2514 A C expression which is nonzero if register number @var{num} is
2515 suitable for use as a base register in operand addresses, accessing
2516 memory in mode @var{mode} in address space @var{address_space}.
2517 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2518 that that expression may examine the context in which the register
2519 appears in the memory reference. @var{outer_code} is the code of the
2520 immediately enclosing expression (@code{MEM} if at the top level of the
2521 address, @code{ADDRESS} for something that occurs in an
2522 @code{address_operand}). @var{index_code} is the code of the
2523 corresponding index expression if @var{outer_code} is @code{PLUS};
2524 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2525 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2528 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2529 A C expression which is nonzero if register number @var{num} is
2530 suitable for use as an index register in operand addresses. It may be
2531 either a suitable hard register or a pseudo register that has been
2532 allocated such a hard register.
2534 The difference between an index register and a base register is that
2535 the index register may be scaled. If an address involves the sum of
2536 two registers, neither one of them scaled, then either one may be
2537 labeled the ``base'' and the other the ``index''; but whichever
2538 labeling is used must fit the machine's constraints of which registers
2539 may serve in each capacity. The compiler will try both labelings,
2540 looking for one that is valid, and will reload one or both registers
2541 only if neither labeling works.
2544 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2545 A target hook that places additional preference on the register class to use when it is necessary to rename a register in class @var{rclass} to another class, or perhaps @var{NO_REGS}, if no preferred register class is found or hook @code{preferred_rename_class} is not implemented. Sometimes returning a more restrictive class makes better code. For example, on ARM, thumb-2 instructions using @code{LO_REGS} may be smaller than instructions using @code{GENERIC_REGS}. By returning @code{LO_REGS} from @code{preferred_rename_class}, code size can be reduced.
2548 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2549 A target hook that places additional restrictions on the register class
2550 to use when it is necessary to copy value @var{x} into a register in class
2551 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2552 another, smaller class.
2554 The default version of this hook always returns value of @code{rclass} argument.
2556 Sometimes returning a more restrictive class makes better code. For
2557 example, on the 68000, when @var{x} is an integer constant that is in range
2558 for a @samp{moveq} instruction, the value of this macro is always
2559 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2560 Requiring a data register guarantees that a @samp{moveq} will be used.
2562 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2563 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2564 loaded into some register class. By returning @code{NO_REGS} you can
2565 force @var{x} into a memory location. For example, rs6000 can load
2566 immediate values into general-purpose registers, but does not have an
2567 instruction for loading an immediate value into a floating-point
2568 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2569 @var{x} is a floating-point constant. If the constant can't be loaded
2570 into any kind of register, code generation will be better if
2571 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2572 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2574 If an insn has pseudos in it after register allocation, reload will go
2575 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2576 to find the best one. Returning @code{NO_REGS}, in this case, makes
2577 reload add a @code{!} in front of the constraint: the x86 back-end uses
2578 this feature to discourage usage of 387 registers when math is done in
2579 the SSE registers (and vice versa).
2582 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2583 A C expression that places additional restrictions on the register class
2584 to use when it is necessary to copy value @var{x} into a register in class
2585 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2586 another, smaller class. On many machines, the following definition is
2590 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2593 Sometimes returning a more restrictive class makes better code. For
2594 example, on the 68000, when @var{x} is an integer constant that is in range
2595 for a @samp{moveq} instruction, the value of this macro is always
2596 @code{DATA_REGS} as long as @var{class} includes the data registers.
2597 Requiring a data register guarantees that a @samp{moveq} will be used.
2599 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2600 @var{class} is if @var{x} is a legitimate constant which cannot be
2601 loaded into some register class. By returning @code{NO_REGS} you can
2602 force @var{x} into a memory location. For example, rs6000 can load
2603 immediate values into general-purpose registers, but does not have an
2604 instruction for loading an immediate value into a floating-point
2605 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2606 @var{x} is a floating-point constant. If the constant can't be loaded
2607 into any kind of register, code generation will be better if
2608 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2609 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2611 If an insn has pseudos in it after register allocation, reload will go
2612 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2613 to find the best one. Returning @code{NO_REGS}, in this case, makes
2614 reload add a @code{!} in front of the constraint: the x86 back-end uses
2615 this feature to discourage usage of 387 registers when math is done in
2616 the SSE registers (and vice versa).
2619 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2620 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2623 The default version of this hook always returns value of @code{rclass}
2626 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2627 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2630 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2631 A C expression that places additional restrictions on the register class
2632 to use when it is necessary to be able to hold a value of mode
2633 @var{mode} in a reload register for which class @var{class} would
2636 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2637 there are certain modes that simply can't go in certain reload classes.
2639 The value is a register class; perhaps @var{class}, or perhaps another,
2642 Don't define this macro unless the target machine has limitations which
2643 require the macro to do something nontrivial.
2646 @deftypefn {Target Hook} reg_class_t TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, reg_class_t @var{reload_class}, enum machine_mode @var{reload_mode}, secondary_reload_info *@var{sri})
2647 Many machines have some registers that cannot be copied directly to or
2648 from memory or even from other types of registers. An example is the
2649 @samp{MQ} register, which on most machines, can only be copied to or
2650 from general registers, but not memory. Below, we shall be using the
2651 term 'intermediate register' when a move operation cannot be performed
2652 directly, but has to be done by copying the source into the intermediate
2653 register first, and then copying the intermediate register to the
2654 destination. An intermediate register always has the same mode as
2655 source and destination. Since it holds the actual value being copied,
2656 reload might apply optimizations to re-use an intermediate register
2657 and eliding the copy from the source when it can determine that the
2658 intermediate register still holds the required value.
2660 Another kind of secondary reload is required on some machines which
2661 allow copying all registers to and from memory, but require a scratch
2662 register for stores to some memory locations (e.g., those with symbolic
2663 address on the RT, and those with certain symbolic address on the SPARC
2664 when compiling PIC)@. Scratch registers need not have the same mode
2665 as the value being copied, and usually hold a different value than
2666 that being copied. Special patterns in the md file are needed to
2667 describe how the copy is performed with the help of the scratch register;
2668 these patterns also describe the number, register class(es) and mode(s)
2669 of the scratch register(s).
2671 In some cases, both an intermediate and a scratch register are required.
2673 For input reloads, this target hook is called with nonzero @var{in_p},
2674 and @var{x} is an rtx that needs to be copied to a register of class
2675 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2676 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2677 needs to be copied to rtx @var{x} in @var{reload_mode}.
2679 If copying a register of @var{reload_class} from/to @var{x} requires
2680 an intermediate register, the hook @code{secondary_reload} should
2681 return the register class required for this intermediate register.
2682 If no intermediate register is required, it should return NO_REGS.
2683 If more than one intermediate register is required, describe the one
2684 that is closest in the copy chain to the reload register.
2686 If scratch registers are needed, you also have to describe how to
2687 perform the copy from/to the reload register to/from this
2688 closest intermediate register. Or if no intermediate register is
2689 required, but still a scratch register is needed, describe the
2690 copy from/to the reload register to/from the reload operand @var{x}.
2692 You do this by setting @code{sri->icode} to the instruction code of a pattern
2693 in the md file which performs the move. Operands 0 and 1 are the output
2694 and input of this copy, respectively. Operands from operand 2 onward are
2695 for scratch operands. These scratch operands must have a mode, and a
2696 single-register-class
2697 @c [later: or memory]
2700 When an intermediate register is used, the @code{secondary_reload}
2701 hook will be called again to determine how to copy the intermediate
2702 register to/from the reload operand @var{x}, so your hook must also
2703 have code to handle the register class of the intermediate operand.
2705 @c [For later: maybe we'll allow multi-alternative reload patterns -
2706 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2707 @c and match the constraints of input and output to determine the required
2708 @c alternative. A restriction would be that constraints used to match
2709 @c against reloads registers would have to be written as register class
2710 @c constraints, or we need a new target macro / hook that tells us if an
2711 @c arbitrary constraint can match an unknown register of a given class.
2712 @c Such a macro / hook would also be useful in other places.]
2715 @var{x} might be a pseudo-register or a @code{subreg} of a
2716 pseudo-register, which could either be in a hard register or in memory.
2717 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2718 in memory and the hard register number if it is in a register.
2720 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2721 currently not supported. For the time being, you will have to continue
2722 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2724 @code{copy_cost} also uses this target hook to find out how values are
2725 copied. If you want it to include some extra cost for the need to allocate
2726 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2727 Or if two dependent moves are supposed to have a lower cost than the sum
2728 of the individual moves due to expected fortuitous scheduling and/or special
2729 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2732 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2733 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2734 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2735 These macros are obsolete, new ports should use the target hook
2736 @code{TARGET_SECONDARY_RELOAD} instead.
2738 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2739 target hook. Older ports still define these macros to indicate to the
2740 reload phase that it may
2741 need to allocate at least one register for a reload in addition to the
2742 register to contain the data. Specifically, if copying @var{x} to a
2743 register @var{class} in @var{mode} requires an intermediate register,
2744 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2745 largest register class all of whose registers can be used as
2746 intermediate registers or scratch registers.
2748 If copying a register @var{class} in @var{mode} to @var{x} requires an
2749 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2750 was supposed to be defined be defined to return the largest register
2751 class required. If the
2752 requirements for input and output reloads were the same, the macro
2753 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2756 The values returned by these macros are often @code{GENERAL_REGS}.
2757 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2758 can be directly copied to or from a register of @var{class} in
2759 @var{mode} without requiring a scratch register. Do not define this
2760 macro if it would always return @code{NO_REGS}.
2762 If a scratch register is required (either with or without an
2763 intermediate register), you were supposed to define patterns for
2764 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2765 (@pxref{Standard Names}. These patterns, which were normally
2766 implemented with a @code{define_expand}, should be similar to the
2767 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2770 These patterns need constraints for the reload register and scratch
2772 contain a single register class. If the original reload register (whose
2773 class is @var{class}) can meet the constraint given in the pattern, the
2774 value returned by these macros is used for the class of the scratch
2775 register. Otherwise, two additional reload registers are required.
2776 Their classes are obtained from the constraints in the insn pattern.
2778 @var{x} might be a pseudo-register or a @code{subreg} of a
2779 pseudo-register, which could either be in a hard register or in memory.
2780 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2781 in memory and the hard register number if it is in a register.
2783 These macros should not be used in the case where a particular class of
2784 registers can only be copied to memory and not to another class of
2785 registers. In that case, secondary reload registers are not needed and
2786 would not be helpful. Instead, a stack location must be used to perform
2787 the copy and the @code{mov@var{m}} pattern should use memory as an
2788 intermediate storage. This case often occurs between floating-point and
2792 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2793 Certain machines have the property that some registers cannot be copied
2794 to some other registers without using memory. Define this macro on
2795 those machines to be a C expression that is nonzero if objects of mode
2796 @var{m} in registers of @var{class1} can only be copied to registers of
2797 class @var{class2} by storing a register of @var{class1} into memory
2798 and loading that memory location into a register of @var{class2}.
2800 Do not define this macro if its value would always be zero.
2803 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2804 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2805 allocates a stack slot for a memory location needed for register copies.
2806 If this macro is defined, the compiler instead uses the memory location
2807 defined by this macro.
2809 Do not define this macro if you do not define
2810 @code{SECONDARY_MEMORY_NEEDED}.
2813 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2814 When the compiler needs a secondary memory location to copy between two
2815 registers of mode @var{mode}, it normally allocates sufficient memory to
2816 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2817 load operations in a mode that many bits wide and whose class is the
2818 same as that of @var{mode}.
2820 This is right thing to do on most machines because it ensures that all
2821 bits of the register are copied and prevents accesses to the registers
2822 in a narrower mode, which some machines prohibit for floating-point
2825 However, this default behavior is not correct on some machines, such as
2826 the DEC Alpha, that store short integers in floating-point registers
2827 differently than in integer registers. On those machines, the default
2828 widening will not work correctly and you must define this macro to
2829 suppress that widening in some cases. See the file @file{alpha.h} for
2832 Do not define this macro if you do not define
2833 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2834 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2837 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2838 A target hook which returns @code{true} if pseudos that have been assigned
2839 to registers of class @var{rclass} would likely be spilled because
2840 registers of @var{rclass} are needed for spill registers.
2842 The default version of this target hook returns @code{true} if @var{rclass}
2843 has exactly one register and @code{false} otherwise. On most machines, this
2844 default should be used. For generally register-starved machines, such as
2845 i386, or machines with right register constraints, such as SH, this hook
2846 can be used to avoid excessive spilling.
2848 This hook is also used by some of the global intra-procedural code
2849 transformations to throtle code motion, to avoid increasing register
2853 @deftypefn {Target Hook} {unsigned char} TARGET_CLASS_MAX_NREGS (reg_class_t @var{rclass}, enum machine_mode @var{mode})
2854 A target hook returns the maximum number of consecutive registers
2855 of class @var{rclass} needed to hold a value of mode @var{mode}.
2857 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2858 the value returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass},
2859 @var{mode})} target hook should be the maximum value of
2860 @code{HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno}
2861 values in the class @var{rclass}.
2863 This target hook helps control the handling of multiple-word values
2866 The default version of this target hook returns the size of @var{mode}
2870 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2871 A C expression for the maximum number of consecutive registers
2872 of class @var{class} needed to hold a value of mode @var{mode}.
2874 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2875 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2876 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2877 @var{mode})} for all @var{regno} values in the class @var{class}.
2879 This macro helps control the handling of multiple-word values
2883 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2884 If defined, a C expression that returns nonzero for a @var{class} for which
2885 a change from mode @var{from} to mode @var{to} is invalid.
2887 For the example, loading 32-bit integer or floating-point objects into
2888 floating-point registers on the Alpha extends them to 64 bits.
2889 Therefore loading a 64-bit object and then storing it as a 32-bit object
2890 does not store the low-order 32 bits, as would be the case for a normal
2891 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2895 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2896 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2897 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2901 @deftypefn {Target Hook} bool TARGET_LRA_P (void)
2902 A target hook which returns true if we use LRA instead of reload pass. It means that LRA was ported to the target. The default version of this target hook returns always false.
2905 @deftypefn {Target Hook} int TARGET_REGISTER_PRIORITY (int)
2906 A target hook which returns the register priority number to which the register @var{hard_regno} belongs to. The bigger the number, the more preferable the hard register usage (when all other conditions are the same). This hook can be used to prefer some hard register over others in LRA. For example, some x86-64 register usage needs additional prefix which makes instructions longer. The hook can return lower priority number for such registers make them less favorable and as result making the generated code smaller. The default version of this target hook returns always zero.
2909 @deftypefn {Target Hook} bool TARGET_REGISTER_USAGE_LEVELING_P (void)
2910 A target hook which returns true if we need register usage leveling. That means if a few hard registers are equally good for the assignment, we choose the least used hard register. The register usage leveling may be profitable for some targets. Don't use the usage leveling for targets with conditional execution or targets with big register files as it hurts if-conversion and cross-jumping optimizations. The default version of this target hook returns always false.
2913 @deftypefn {Target Hook} bool TARGET_DIFFERENT_ADDR_DISPLACEMENT_P (void)
2914 A target hook which returns true if an address with the same structure can have different maximal legitimate displacement. For example, the displacement can depend on memory mode or on operand combinations in the insn. The default version of this target hook returns always false.
2917 @deftypefn {Target Hook} reg_class_t TARGET_SPILL_CLASS (reg_class_t, enum @var{machine_mode})
2918 This hook defines a class of registers which could be used for spilling pseudos of the given mode and class, or @code{NO_REGS} if only memory should be used. Not defining this hook is equivalent to returning @code{NO_REGS} for all inputs.
2921 @deftypefn {Target Hook} {enum machine_mode} TARGET_CSTORE_MODE (enum insn_code @var{icode})
2922 This hook defines the machine mode to use for the boolean result of conditional store patterns. The ICODE argument is the instruction code for the cstore being performed. Not definiting this hook is the same as accepting the mode encoded into operand 0 of the cstore expander patterns.
2925 @node Old Constraints
2926 @section Obsolete Macros for Defining Constraints
2927 @cindex defining constraints, obsolete method
2928 @cindex constraints, defining, obsolete method
2930 Machine-specific constraints can be defined with these macros instead
2931 of the machine description constructs described in @ref{Define
2932 Constraints}. This mechanism is obsolete. New ports should not use
2933 it; old ports should convert to the new mechanism.
2935 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2936 For the constraint at the start of @var{str}, which starts with the letter
2937 @var{c}, return the length. This allows you to have register class /
2938 constant / extra constraints that are longer than a single letter;
2939 you don't need to define this macro if you can do with single-letter
2940 constraints only. The definition of this macro should use
2941 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2942 to handle specially.
2943 There are some sanity checks in genoutput.c that check the constraint lengths
2944 for the md file, so you can also use this macro to help you while you are
2945 transitioning from a byzantine single-letter-constraint scheme: when you
2946 return a negative length for a constraint you want to re-use, genoutput
2947 will complain about every instance where it is used in the md file.
2950 @defmac REG_CLASS_FROM_LETTER (@var{char})
2951 A C expression which defines the machine-dependent operand constraint
2952 letters for register classes. If @var{char} is such a letter, the
2953 value should be the register class corresponding to it. Otherwise,
2954 the value should be @code{NO_REGS}. The register letter @samp{r},
2955 corresponding to class @code{GENERAL_REGS}, will not be passed
2956 to this macro; you do not need to handle it.
2959 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2960 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2961 passed in @var{str}, so that you can use suffixes to distinguish between
2965 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2966 A C expression that defines the machine-dependent operand constraint
2967 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2968 particular ranges of integer values. If @var{c} is one of those
2969 letters, the expression should check that @var{value}, an integer, is in
2970 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2971 not one of those letters, the value should be 0 regardless of
2975 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2976 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2977 string passed in @var{str}, so that you can use suffixes to distinguish
2978 between different variants.
2981 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2982 A C expression that defines the machine-dependent operand constraint
2983 letters that specify particular ranges of @code{const_double} values
2984 (@samp{G} or @samp{H}).
2986 If @var{c} is one of those letters, the expression should check that
2987 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2988 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2989 letters, the value should be 0 regardless of @var{value}.
2991 @code{const_double} is used for all floating-point constants and for
2992 @code{DImode} fixed-point constants. A given letter can accept either
2993 or both kinds of values. It can use @code{GET_MODE} to distinguish
2994 between these kinds.
2997 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2998 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2999 string passed in @var{str}, so that you can use suffixes to distinguish
3000 between different variants.
3003 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
3004 A C expression that defines the optional machine-dependent constraint
3005 letters that can be used to segregate specific types of operands, usually
3006 memory references, for the target machine. Any letter that is not
3007 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
3008 @code{REG_CLASS_FROM_CONSTRAINT}
3009 may be used. Normally this macro will not be defined.
3011 If it is required for a particular target machine, it should return 1
3012 if @var{value} corresponds to the operand type represented by the
3013 constraint letter @var{c}. If @var{c} is not defined as an extra
3014 constraint, the value returned should be 0 regardless of @var{value}.
3016 For example, on the ROMP, load instructions cannot have their output
3017 in r0 if the memory reference contains a symbolic address. Constraint
3018 letter @samp{Q} is defined as representing a memory address that does
3019 @emph{not} contain a symbolic address. An alternative is specified with
3020 a @samp{Q} constraint on the input and @samp{r} on the output. The next
3021 alternative specifies @samp{m} on the input and a register class that
3022 does not include r0 on the output.
3025 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
3026 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
3027 in @var{str}, so that you can use suffixes to distinguish between different
3031 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
3032 A C expression that defines the optional machine-dependent constraint
3033 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
3034 be treated like memory constraints by the reload pass.
3036 It should return 1 if the operand type represented by the constraint
3037 at the start of @var{str}, the first letter of which is the letter @var{c},
3038 comprises a subset of all memory references including
3039 all those whose address is simply a base register. This allows the reload
3040 pass to reload an operand, if it does not directly correspond to the operand
3041 type of @var{c}, by copying its address into a base register.
3043 For example, on the S/390, some instructions do not accept arbitrary
3044 memory references, but only those that do not make use of an index
3045 register. The constraint letter @samp{Q} is defined via
3046 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
3047 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
3048 a @samp{Q} constraint can handle any memory operand, because the
3049 reload pass knows it can be reloaded by copying the memory address
3050 into a base register if required. This is analogous to the way
3051 an @samp{o} constraint can handle any memory operand.
3054 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
3055 A C expression that defines the optional machine-dependent constraint
3056 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
3057 @code{EXTRA_CONSTRAINT_STR}, that should
3058 be treated like address constraints by the reload pass.
3060 It should return 1 if the operand type represented by the constraint
3061 at the start of @var{str}, which starts with the letter @var{c}, comprises
3062 a subset of all memory addresses including
3063 all those that consist of just a base register. This allows the reload
3064 pass to reload an operand, if it does not directly correspond to the operand
3065 type of @var{str}, by copying it into a base register.
3067 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3068 be used with the @code{address_operand} predicate. It is treated
3069 analogously to the @samp{p} constraint.
3072 @node Stack and Calling
3073 @section Stack Layout and Calling Conventions
3074 @cindex calling conventions
3076 @c prevent bad page break with this line
3077 This describes the stack layout and calling conventions.
3081 * Exception Handling::
3086 * Register Arguments::
3088 * Aggregate Return::
3093 * Stack Smashing Protection::
3097 @subsection Basic Stack Layout
3098 @cindex stack frame layout
3099 @cindex frame layout
3101 @c prevent bad page break with this line
3102 Here is the basic stack layout.
3104 @defmac STACK_GROWS_DOWNWARD
3105 Define this macro if pushing a word onto the stack moves the stack
3106 pointer to a smaller address.
3108 When we say, ``define this macro if @dots{}'', it means that the
3109 compiler checks this macro only with @code{#ifdef} so the precise
3110 definition used does not matter.
3113 @defmac STACK_PUSH_CODE
3114 This macro defines the operation used when something is pushed
3115 on the stack. In RTL, a push operation will be
3116 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3118 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3119 and @code{POST_INC}. Which of these is correct depends on
3120 the stack direction and on whether the stack pointer points
3121 to the last item on the stack or whether it points to the
3122 space for the next item on the stack.
3124 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3125 defined, which is almost always right, and @code{PRE_INC} otherwise,
3126 which is often wrong.
3129 @defmac FRAME_GROWS_DOWNWARD
3130 Define this macro to nonzero value if the addresses of local variable slots
3131 are at negative offsets from the frame pointer.
3134 @defmac ARGS_GROW_DOWNWARD
3135 Define this macro if successive arguments to a function occupy decreasing
3136 addresses on the stack.
3139 @defmac STARTING_FRAME_OFFSET
3140 Offset from the frame pointer to the first local variable slot to be allocated.
3142 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3143 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3144 Otherwise, it is found by adding the length of the first slot to the
3145 value @code{STARTING_FRAME_OFFSET}.
3146 @c i'm not sure if the above is still correct.. had to change it to get
3147 @c rid of an overfull. --mew 2feb93
3150 @defmac STACK_ALIGNMENT_NEEDED
3151 Define to zero to disable final alignment of the stack during reload.
3152 The nonzero default for this macro is suitable for most ports.
3154 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3155 is a register save block following the local block that doesn't require
3156 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3157 stack alignment and do it in the backend.
3160 @defmac STACK_POINTER_OFFSET
3161 Offset from the stack pointer register to the first location at which
3162 outgoing arguments are placed. If not specified, the default value of
3163 zero is used. This is the proper value for most machines.
3165 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3166 the first location at which outgoing arguments are placed.
3169 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3170 Offset from the argument pointer register to the first argument's
3171 address. On some machines it may depend on the data type of the
3174 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3175 the first argument's address.
3178 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3179 Offset from the stack pointer register to an item dynamically allocated
3180 on the stack, e.g., by @code{alloca}.
3182 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3183 length of the outgoing arguments. The default is correct for most
3184 machines. See @file{function.c} for details.
3187 @defmac INITIAL_FRAME_ADDRESS_RTX
3188 A C expression whose value is RTL representing the address of the initial
3189 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3190 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3191 default value will be used. Define this macro in order to make frame pointer
3192 elimination work in the presence of @code{__builtin_frame_address (count)} and
3193 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3196 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3197 A C expression whose value is RTL representing the address in a stack
3198 frame where the pointer to the caller's frame is stored. Assume that
3199 @var{frameaddr} is an RTL expression for the address of the stack frame
3202 If you don't define this macro, the default is to return the value
3203 of @var{frameaddr}---that is, the stack frame address is also the
3204 address of the stack word that points to the previous frame.
3207 @defmac SETUP_FRAME_ADDRESSES
3208 If defined, a C expression that produces the machine-specific code to
3209 setup the stack so that arbitrary frames can be accessed. For example,
3210 on the SPARC, we must flush all of the register windows to the stack
3211 before we can access arbitrary stack frames. You will seldom need to
3215 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3216 This target hook should return an rtx that is used to store
3217 the address of the current frame into the built in @code{setjmp} buffer.
3218 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3219 machines. One reason you may need to define this target hook is if
3220 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3223 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3224 A C expression whose value is RTL representing the value of the frame
3225 address for the current frame. @var{frameaddr} is the frame pointer
3226 of the current frame. This is used for __builtin_frame_address.
3227 You need only define this macro if the frame address is not the same
3228 as the frame pointer. Most machines do not need to define it.
3231 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3232 A C expression whose value is RTL representing the value of the return
3233 address for the frame @var{count} steps up from the current frame, after
3234 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3235 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3236 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3238 The value of the expression must always be the correct address when
3239 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3240 determine the return address of other frames.
3243 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3244 Define this if the return address of a particular stack frame is accessed
3245 from the frame pointer of the previous stack frame.
3248 @defmac INCOMING_RETURN_ADDR_RTX
3249 A C expression whose value is RTL representing the location of the
3250 incoming return address at the beginning of any function, before the
3251 prologue. This RTL is either a @code{REG}, indicating that the return
3252 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3255 You only need to define this macro if you want to support call frame
3256 debugging information like that provided by DWARF 2.
3258 If this RTL is a @code{REG}, you should also define
3259 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3262 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3263 A C expression whose value is an integer giving a DWARF 2 column
3264 number that may be used as an alternative return column. The column
3265 must not correspond to any gcc hard register (that is, it must not
3266 be in the range of @code{DWARF_FRAME_REGNUM}).
3268 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3269 general register, but an alternative column needs to be used for signal
3270 frames. Some targets have also used different frame return columns
3274 @defmac DWARF_ZERO_REG
3275 A C expression whose value is an integer giving a DWARF 2 register
3276 number that is considered to always have the value zero. This should
3277 only be defined if the target has an architected zero register, and
3278 someone decided it was a good idea to use that register number to
3279 terminate the stack backtrace. New ports should avoid this.
3282 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3283 This target hook allows the backend to emit frame-related insns that
3284 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3285 info engine will invoke it on insns of the form
3287 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3291 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3293 to let the backend emit the call frame instructions. @var{label} is
3294 the CFI label attached to the insn, @var{pattern} is the pattern of
3295 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3298 @defmac INCOMING_FRAME_SP_OFFSET
3299 A C expression whose value is an integer giving the offset, in bytes,
3300 from the value of the stack pointer register to the top of the stack
3301 frame at the beginning of any function, before the prologue. The top of
3302 the frame is defined to be the value of the stack pointer in the
3303 previous frame, just before the call instruction.
3305 You only need to define this macro if you want to support call frame
3306 debugging information like that provided by DWARF 2.
3309 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3310 A C expression whose value is an integer giving the offset, in bytes,
3311 from the argument pointer to the canonical frame address (cfa). The
3312 final value should coincide with that calculated by
3313 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3314 during virtual register instantiation.
3316 The default value for this macro is
3317 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3318 which is correct for most machines; in general, the arguments are found
3319 immediately before the stack frame. Note that this is not the case on
3320 some targets that save registers into the caller's frame, such as SPARC
3321 and rs6000, and so such targets need to define this macro.
3323 You only need to define this macro if the default is incorrect, and you
3324 want to support call frame debugging information like that provided by
3328 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3329 If defined, a C expression whose value is an integer giving the offset
3330 in bytes from the frame pointer to the canonical frame address (cfa).
3331 The final value should coincide with that calculated by
3332 @code{INCOMING_FRAME_SP_OFFSET}.
3334 Normally the CFA is calculated as an offset from the argument pointer,
3335 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3336 variable due to the ABI, this may not be possible. If this macro is
3337 defined, it implies that the virtual register instantiation should be
3338 based on the frame pointer instead of the argument pointer. Only one
3339 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3343 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3344 If defined, a C expression whose value is an integer giving the offset
3345 in bytes from the canonical frame address (cfa) to the frame base used
3346 in DWARF 2 debug information. The default is zero. A different value
3347 may reduce the size of debug information on some ports.
3350 @node Exception Handling
3351 @subsection Exception Handling Support
3352 @cindex exception handling
3354 @defmac EH_RETURN_DATA_REGNO (@var{N})
3355 A C expression whose value is the @var{N}th register number used for
3356 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3357 @var{N} registers are usable.
3359 The exception handling library routines communicate with the exception
3360 handlers via a set of agreed upon registers. Ideally these registers
3361 should be call-clobbered; it is possible to use call-saved registers,
3362 but may negatively impact code size. The target must support at least
3363 2 data registers, but should define 4 if there are enough free registers.
3365 You must define this macro if you want to support call frame exception
3366 handling like that provided by DWARF 2.
3369 @defmac EH_RETURN_STACKADJ_RTX
3370 A C expression whose value is RTL representing a location in which
3371 to store a stack adjustment to be applied before function return.
3372 This is used to unwind the stack to an exception handler's call frame.
3373 It will be assigned zero on code paths that return normally.
3375 Typically this is a call-clobbered hard register that is otherwise
3376 untouched by the epilogue, but could also be a stack slot.
3378 Do not define this macro if the stack pointer is saved and restored
3379 by the regular prolog and epilog code in the call frame itself; in
3380 this case, the exception handling library routines will update the
3381 stack location to be restored in place. Otherwise, you must define
3382 this macro if you want to support call frame exception handling like
3383 that provided by DWARF 2.
3386 @defmac EH_RETURN_HANDLER_RTX
3387 A C expression whose value is RTL representing a location in which
3388 to store the address of an exception handler to which we should
3389 return. It will not be assigned on code paths that return normally.
3391 Typically this is the location in the call frame at which the normal
3392 return address is stored. For targets that return by popping an
3393 address off the stack, this might be a memory address just below
3394 the @emph{target} call frame rather than inside the current call
3395 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3396 been assigned, so it may be used to calculate the location of the
3399 Some targets have more complex requirements than storing to an
3400 address calculable during initial code generation. In that case
3401 the @code{eh_return} instruction pattern should be used instead.
3403 If you want to support call frame exception handling, you must
3404 define either this macro or the @code{eh_return} instruction pattern.
3407 @defmac RETURN_ADDR_OFFSET
3408 If defined, an integer-valued C expression for which rtl will be generated
3409 to add it to the exception handler address before it is searched in the
3410 exception handling tables, and to subtract it again from the address before
3411 using it to return to the exception handler.
3414 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3415 This macro chooses the encoding of pointers embedded in the exception
3416 handling sections. If at all possible, this should be defined such
3417 that the exception handling section will not require dynamic relocations,
3418 and so may be read-only.
3420 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3421 @var{global} is true if the symbol may be affected by dynamic relocations.
3422 The macro should return a combination of the @code{DW_EH_PE_*} defines
3423 as found in @file{dwarf2.h}.
3425 If this macro is not defined, pointers will not be encoded but
3426 represented directly.
3429 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3430 This macro allows the target to emit whatever special magic is required
3431 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3432 Generic code takes care of pc-relative and indirect encodings; this must
3433 be defined if the target uses text-relative or data-relative encodings.
3435 This is a C statement that branches to @var{done} if the format was
3436 handled. @var{encoding} is the format chosen, @var{size} is the number
3437 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3441 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3442 This macro allows the target to add CPU and operating system specific
3443 code to the call-frame unwinder for use when there is no unwind data
3444 available. The most common reason to implement this macro is to unwind
3445 through signal frames.
3447 This macro is called from @code{uw_frame_state_for} in
3448 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3449 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3450 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3451 for the address of the code being executed and @code{context->cfa} for
3452 the stack pointer value. If the frame can be decoded, the register
3453 save addresses should be updated in @var{fs} and the macro should
3454 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3455 the macro should evaluate to @code{_URC_END_OF_STACK}.
3457 For proper signal handling in Java this macro is accompanied by
3458 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3461 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3462 This macro allows the target to add operating system specific code to the
3463 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3464 usually used for signal or interrupt frames.
3466 This macro is called from @code{uw_update_context} in libgcc's
3467 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3468 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3469 for the abi and context in the @code{.unwabi} directive. If the
3470 @code{.unwabi} directive can be handled, the register save addresses should
3471 be updated in @var{fs}.
3474 @defmac TARGET_USES_WEAK_UNWIND_INFO
3475 A C expression that evaluates to true if the target requires unwind
3476 info to be given comdat linkage. Define it to be @code{1} if comdat
3477 linkage is necessary. The default is @code{0}.
3480 @node Stack Checking
3481 @subsection Specifying How Stack Checking is Done
3483 GCC will check that stack references are within the boundaries of the
3484 stack, if the option @option{-fstack-check} is specified, in one of
3489 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3490 will assume that you have arranged for full stack checking to be done
3491 at appropriate places in the configuration files. GCC will not do
3492 other special processing.
3495 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3496 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3497 that you have arranged for static stack checking (checking of the
3498 static stack frame of functions) to be done at appropriate places
3499 in the configuration files. GCC will only emit code to do dynamic
3500 stack checking (checking on dynamic stack allocations) using the third
3504 If neither of the above are true, GCC will generate code to periodically
3505 ``probe'' the stack pointer using the values of the macros defined below.
3508 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3509 GCC will change its allocation strategy for large objects if the option
3510 @option{-fstack-check} is specified: they will always be allocated
3511 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3513 @defmac STACK_CHECK_BUILTIN
3514 A nonzero value if stack checking is done by the configuration files in a
3515 machine-dependent manner. You should define this macro if stack checking
3516 is required by the ABI of your machine or if you would like to do stack
3517 checking in some more efficient way than the generic approach. The default
3518 value of this macro is zero.
3521 @defmac STACK_CHECK_STATIC_BUILTIN
3522 A nonzero value if static stack checking is done by the configuration files
3523 in a machine-dependent manner. You should define this macro if you would
3524 like to do static stack checking in some more efficient way than the generic
3525 approach. The default value of this macro is zero.
3528 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3529 An integer specifying the interval at which GCC must generate stack probe
3530 instructions, defined as 2 raised to this integer. You will normally
3531 define this macro so that the interval be no larger than the size of
3532 the ``guard pages'' at the end of a stack area. The default value
3533 of 12 (4096-byte interval) is suitable for most systems.
3536 @defmac STACK_CHECK_MOVING_SP
3537 An integer which is nonzero if GCC should move the stack pointer page by page
3538 when doing probes. This can be necessary on systems where the stack pointer
3539 contains the bottom address of the memory area accessible to the executing
3540 thread at any point in time. In this situation an alternate signal stack
3541 is required in order to be able to recover from a stack overflow. The
3542 default value of this macro is zero.
3545 @defmac STACK_CHECK_PROTECT
3546 The number of bytes of stack needed to recover from a stack overflow, for
3547 languages where such a recovery is supported. The default value of 75 words
3548 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3549 8192 bytes with other exception handling mechanisms should be adequate for
3553 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3554 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3555 in the opposite case.
3557 @defmac STACK_CHECK_MAX_FRAME_SIZE
3558 The maximum size of a stack frame, in bytes. GCC will generate probe
3559 instructions in non-leaf functions to ensure at least this many bytes of
3560 stack are available. If a stack frame is larger than this size, stack
3561 checking will not be reliable and GCC will issue a warning. The
3562 default is chosen so that GCC only generates one instruction on most
3563 systems. You should normally not change the default value of this macro.
3566 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3567 GCC uses this value to generate the above warning message. It
3568 represents the amount of fixed frame used by a function, not including
3569 space for any callee-saved registers, temporaries and user variables.
3570 You need only specify an upper bound for this amount and will normally
3571 use the default of four words.
3574 @defmac STACK_CHECK_MAX_VAR_SIZE
3575 The maximum size, in bytes, of an object that GCC will place in the
3576 fixed area of the stack frame when the user specifies
3577 @option{-fstack-check}.
3578 GCC computed the default from the values of the above macros and you will
3579 normally not need to override that default.
3583 @node Frame Registers
3584 @subsection Registers That Address the Stack Frame
3586 @c prevent bad page break with this line
3587 This discusses registers that address the stack frame.
3589 @defmac STACK_POINTER_REGNUM
3590 The register number of the stack pointer register, which must also be a
3591 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3592 the hardware determines which register this is.
3595 @defmac FRAME_POINTER_REGNUM
3596 The register number of the frame pointer register, which is used to
3597 access automatic variables in the stack frame. On some machines, the
3598 hardware determines which register this is. On other machines, you can
3599 choose any register you wish for this purpose.
3602 @defmac HARD_FRAME_POINTER_REGNUM
3603 On some machines the offset between the frame pointer and starting
3604 offset of the automatic variables is not known until after register
3605 allocation has been done (for example, because the saved registers are
3606 between these two locations). On those machines, define
3607 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3608 be used internally until the offset is known, and define
3609 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3610 used for the frame pointer.
3612 You should define this macro only in the very rare circumstances when it
3613 is not possible to calculate the offset between the frame pointer and
3614 the automatic variables until after register allocation has been
3615 completed. When this macro is defined, you must also indicate in your
3616 definition of @code{ELIMINABLE_REGS} how to eliminate
3617 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3618 or @code{STACK_POINTER_REGNUM}.
3620 Do not define this macro if it would be the same as
3621 @code{FRAME_POINTER_REGNUM}.
3624 @defmac ARG_POINTER_REGNUM
3625 The register number of the arg pointer register, which is used to access
3626 the function's argument list. On some machines, this is the same as the
3627 frame pointer register. On some machines, the hardware determines which
3628 register this is. On other machines, you can choose any register you
3629 wish for this purpose. If this is not the same register as the frame
3630 pointer register, then you must mark it as a fixed register according to
3631 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3632 (@pxref{Elimination}).
3635 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3636 Define this to a preprocessor constant that is nonzero if
3637 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3638 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3639 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3640 definition is not suitable for use in preprocessor conditionals.
3643 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3644 Define this to a preprocessor constant that is nonzero if
3645 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3646 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3647 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3648 definition is not suitable for use in preprocessor conditionals.
3651 @defmac RETURN_ADDRESS_POINTER_REGNUM
3652 The register number of the return address pointer register, which is used to
3653 access the current function's return address from the stack. On some
3654 machines, the return address is not at a fixed offset from the frame
3655 pointer or stack pointer or argument pointer. This register can be defined
3656 to point to the return address on the stack, and then be converted by
3657 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3659 Do not define this macro unless there is no other way to get the return
3660 address from the stack.
3663 @defmac STATIC_CHAIN_REGNUM
3664 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3665 Register numbers used for passing a function's static chain pointer. If
3666 register windows are used, the register number as seen by the called
3667 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3668 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3669 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3672 The static chain register need not be a fixed register.
3674 If the static chain is passed in memory, these macros should not be
3675 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3678 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl}, bool @var{incoming_p})
3679 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3680 targets that may use different static chain locations for different
3681 nested functions. This may be required if the target has function
3682 attributes that affect the calling conventions of the function and
3683 those calling conventions use different static chain locations.
3685 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3687 If the static chain is passed in memory, this hook should be used to
3688 provide rtx giving @code{mem} expressions that denote where they are stored.
3689 Often the @code{mem} expression as seen by the caller will be at an offset
3690 from the stack pointer and the @code{mem} expression as seen by the callee
3691 will be at an offset from the frame pointer.
3692 @findex stack_pointer_rtx
3693 @findex frame_pointer_rtx
3694 @findex arg_pointer_rtx
3695 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3696 @code{arg_pointer_rtx} will have been initialized and should be used
3697 to refer to those items.
3700 @defmac DWARF_FRAME_REGISTERS
3701 This macro specifies the maximum number of hard registers that can be
3702 saved in a call frame. This is used to size data structures used in
3703 DWARF2 exception handling.
3705 Prior to GCC 3.0, this macro was needed in order to establish a stable
3706 exception handling ABI in the face of adding new hard registers for ISA
3707 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3708 in the number of hard registers. Nevertheless, this macro can still be
3709 used to reduce the runtime memory requirements of the exception handling
3710 routines, which can be substantial if the ISA contains a lot of
3711 registers that are not call-saved.
3713 If this macro is not defined, it defaults to
3714 @code{FIRST_PSEUDO_REGISTER}.
3717 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3719 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3720 for backward compatibility in pre GCC 3.0 compiled code.
3722 If this macro is not defined, it defaults to
3723 @code{DWARF_FRAME_REGISTERS}.
3726 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3728 Define this macro if the target's representation for dwarf registers
3729 is different than the internal representation for unwind column.
3730 Given a dwarf register, this macro should return the internal unwind
3731 column number to use instead.
3733 See the PowerPC's SPE target for an example.
3736 @defmac DWARF_FRAME_REGNUM (@var{regno})
3738 Define this macro if the target's representation for dwarf registers
3739 used in .eh_frame or .debug_frame is different from that used in other
3740 debug info sections. Given a GCC hard register number, this macro
3741 should return the .eh_frame register number. The default is
3742 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3746 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3748 Define this macro to map register numbers held in the call frame info
3749 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3750 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3751 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3752 return @code{@var{regno}}.
3756 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3758 Define this macro if the target stores register values as
3759 @code{_Unwind_Word} type in unwind context. It should be defined if
3760 target register size is larger than the size of @code{void *}. The
3761 default is to store register values as @code{void *} type.
3765 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3767 Define this macro to be 1 if the target always uses extended unwind
3768 context with version, args_size and by_value fields. If it is undefined,
3769 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3770 defined and 0 otherwise.
3775 @subsection Eliminating Frame Pointer and Arg Pointer
3777 @c prevent bad page break with this line
3778 This is about eliminating the frame pointer and arg pointer.
3780 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3781 This target hook should return @code{true} if a function must have and use
3782 a frame pointer. This target hook is called in the reload pass. If its return
3783 value is @code{true} the function will have a frame pointer.
3785 This target hook can in principle examine the current function and decide
3786 according to the facts, but on most machines the constant @code{false} or the
3787 constant @code{true} suffices. Use @code{false} when the machine allows code
3788 to be generated with no frame pointer, and doing so saves some time or space.
3789 Use @code{true} when there is no possible advantage to avoiding a frame
3792 In certain cases, the compiler does not know how to produce valid code
3793 without a frame pointer. The compiler recognizes those cases and
3794 automatically gives the function a frame pointer regardless of what
3795 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3798 In a function that does not require a frame pointer, the frame pointer
3799 register can be allocated for ordinary usage, unless you mark it as a
3800 fixed register. See @code{FIXED_REGISTERS} for more information.
3802 Default return value is @code{false}.
3805 @findex get_frame_size
3806 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3807 A C statement to store in the variable @var{depth-var} the difference
3808 between the frame pointer and the stack pointer values immediately after
3809 the function prologue. The value would be computed from information
3810 such as the result of @code{get_frame_size ()} and the tables of
3811 registers @code{regs_ever_live} and @code{call_used_regs}.
3813 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3814 need not be defined. Otherwise, it must be defined even if
3815 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3816 case, you may set @var{depth-var} to anything.
3819 @defmac ELIMINABLE_REGS
3820 If defined, this macro specifies a table of register pairs used to
3821 eliminate unneeded registers that point into the stack frame. If it is not
3822 defined, the only elimination attempted by the compiler is to replace
3823 references to the frame pointer with references to the stack pointer.
3825 The definition of this macro is a list of structure initializations, each
3826 of which specifies an original and replacement register.
3828 On some machines, the position of the argument pointer is not known until
3829 the compilation is completed. In such a case, a separate hard register
3830 must be used for the argument pointer. This register can be eliminated by
3831 replacing it with either the frame pointer or the argument pointer,
3832 depending on whether or not the frame pointer has been eliminated.
3834 In this case, you might specify:
3836 #define ELIMINABLE_REGS \
3837 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3838 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3839 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3842 Note that the elimination of the argument pointer with the stack pointer is
3843 specified first since that is the preferred elimination.
3846 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3847 This target hook should returns @code{true} if the compiler is allowed to
3848 try to replace register number @var{from_reg} with register number
3849 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3850 is defined, and will usually be @code{true}, since most of the cases
3851 preventing register elimination are things that the compiler already
3854 Default return value is @code{true}.
3857 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3858 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3859 specifies the initial difference between the specified pair of
3860 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3864 @node Stack Arguments
3865 @subsection Passing Function Arguments on the Stack
3866 @cindex arguments on stack
3867 @cindex stack arguments
3869 The macros in this section control how arguments are passed
3870 on the stack. See the following section for other macros that
3871 control passing certain arguments in registers.
3873 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3874 This target hook returns @code{true} if an argument declared in a
3875 prototype as an integral type smaller than @code{int} should actually be
3876 passed as an @code{int}. In addition to avoiding errors in certain
3877 cases of mismatch, it also makes for better code on certain machines.
3878 The default is to not promote prototypes.
3882 A C expression. If nonzero, push insns will be used to pass
3884 If the target machine does not have a push instruction, set it to zero.
3885 That directs GCC to use an alternate strategy: to
3886 allocate the entire argument block and then store the arguments into
3887 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3890 @defmac PUSH_ARGS_REVERSED
3891 A C expression. If nonzero, function arguments will be evaluated from
3892 last to first, rather than from first to last. If this macro is not
3893 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3894 and args grow in opposite directions, and 0 otherwise.
3897 @defmac PUSH_ROUNDING (@var{npushed})
3898 A C expression that is the number of bytes actually pushed onto the
3899 stack when an instruction attempts to push @var{npushed} bytes.
3901 On some machines, the definition
3904 #define PUSH_ROUNDING(BYTES) (BYTES)
3908 will suffice. But on other machines, instructions that appear
3909 to push one byte actually push two bytes in an attempt to maintain
3910 alignment. Then the definition should be
3913 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3916 If the value of this macro has a type, it should be an unsigned type.
3919 @findex outgoing_args_size
3920 @findex crtl->outgoing_args_size
3921 @defmac ACCUMULATE_OUTGOING_ARGS
3922 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3923 will be computed and placed into
3924 @code{crtl->outgoing_args_size}. No space will be pushed
3925 onto the stack for each call; instead, the function prologue should
3926 increase the stack frame size by this amount.
3928 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3932 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3933 Define this macro if functions should assume that stack space has been
3934 allocated for arguments even when their values are passed in
3937 The value of this macro is the size, in bytes, of the area reserved for
3938 arguments passed in registers for the function represented by @var{fndecl},
3939 which can be zero if GCC is calling a library function.
3940 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3943 This space can be allocated by the caller, or be a part of the
3944 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3947 @c above is overfull. not sure what to do. --mew 5feb93 did
3948 @c something, not sure if it looks good. --mew 10feb93
3950 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3951 Define this to a nonzero value if it is the responsibility of the
3952 caller to allocate the area reserved for arguments passed in registers
3953 when calling a function of @var{fntype}. @var{fntype} may be NULL
3954 if the function called is a library function.
3956 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3957 whether the space for these arguments counts in the value of
3958 @code{crtl->outgoing_args_size}.
3961 @defmac STACK_PARMS_IN_REG_PARM_AREA
3962 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3963 stack parameters don't skip the area specified by it.
3964 @c i changed this, makes more sens and it should have taken care of the
3965 @c overfull.. not as specific, tho. --mew 5feb93
3967 Normally, when a parameter is not passed in registers, it is placed on the
3968 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3969 suppresses this behavior and causes the parameter to be passed on the
3970 stack in its natural location.
3973 @deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size})
3974 This target hook returns the number of bytes of its own arguments that
3975 a function pops on returning, or 0 if the function pops no arguments
3976 and the caller must therefore pop them all after the function returns.
3978 @var{fundecl} is a C variable whose value is a tree node that describes
3979 the function in question. Normally it is a node of type
3980 @code{FUNCTION_DECL} that describes the declaration of the function.
3981 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3983 @var{funtype} is a C variable whose value is a tree node that
3984 describes the function in question. Normally it is a node of type
3985 @code{FUNCTION_TYPE} that describes the data type of the function.
3986 From this it is possible to obtain the data types of the value and
3987 arguments (if known).
3989 When a call to a library function is being considered, @var{fundecl}
3990 will contain an identifier node for the library function. Thus, if
3991 you need to distinguish among various library functions, you can do so
3992 by their names. Note that ``library function'' in this context means
3993 a function used to perform arithmetic, whose name is known specially
3994 in the compiler and was not mentioned in the C code being compiled.
3996 @var{size} is the number of bytes of arguments passed on the
3997 stack. If a variable number of bytes is passed, it is zero, and
3998 argument popping will always be the responsibility of the calling function.
4000 On the VAX, all functions always pop their arguments, so the definition
4001 of this macro is @var{size}. On the 68000, using the standard
4002 calling convention, no functions pop their arguments, so the value of
4003 the macro is always 0 in this case. But an alternative calling
4004 convention is available in which functions that take a fixed number of
4005 arguments pop them but other functions (such as @code{printf}) pop
4006 nothing (the caller pops all). When this convention is in use,
4007 @var{funtype} is examined to determine whether a function takes a fixed
4008 number of arguments.
4011 @defmac CALL_POPS_ARGS (@var{cum})
4012 A C expression that should indicate the number of bytes a call sequence
4013 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
4014 when compiling a function call.
4016 @var{cum} is the variable in which all arguments to the called function
4017 have been accumulated.
4019 On certain architectures, such as the SH5, a call trampoline is used
4020 that pops certain registers off the stack, depending on the arguments
4021 that have been passed to the function. Since this is a property of the
4022 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
4026 @node Register Arguments
4027 @subsection Passing Arguments in Registers
4028 @cindex arguments in registers
4029 @cindex registers arguments
4031 This section describes the macros which let you control how various
4032 types of arguments are passed in registers or how they are arranged in
4035 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4036 Return an RTX indicating whether a function argument is passed in a
4037 register and if so, which register.
4039 The arguments are @var{ca}, which summarizes all the previous
4040 arguments; @var{mode}, the machine mode of the argument; @var{type},
4041 the data type of the argument as a tree node or 0 if that is not known
4042 (which happens for C support library functions); and @var{named},
4043 which is @code{true} for an ordinary argument and @code{false} for
4044 nameless arguments that correspond to @samp{@dots{}} in the called
4045 function's prototype. @var{type} can be an incomplete type if a
4046 syntax error has previously occurred.
4048 The return value is usually either a @code{reg} RTX for the hard
4049 register in which to pass the argument, or zero to pass the argument
4052 The value of the expression can also be a @code{parallel} RTX@. This is
4053 used when an argument is passed in multiple locations. The mode of the
4054 @code{parallel} should be the mode of the entire argument. The
4055 @code{parallel} holds any number of @code{expr_list} pairs; each one
4056 describes where part of the argument is passed. In each
4057 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
4058 register in which to pass this part of the argument, and the mode of the
4059 register RTX indicates how large this part of the argument is. The
4060 second operand of the @code{expr_list} is a @code{const_int} which gives
4061 the offset in bytes into the entire argument of where this part starts.
4062 As a special exception the first @code{expr_list} in the @code{parallel}
4063 RTX may have a first operand of zero. This indicates that the entire
4064 argument is also stored on the stack.
4066 The last time this hook is called, it is called with @code{MODE ==
4067 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
4068 pattern as operands 2 and 3 respectively.
4070 @cindex @file{stdarg.h} and register arguments
4071 The usual way to make the ISO library @file{stdarg.h} work on a
4072 machine where some arguments are usually passed in registers, is to
4073 cause nameless arguments to be passed on the stack instead. This is
4074 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
4075 @var{named} is @code{false}.
4077 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
4078 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
4079 You may use the hook @code{targetm.calls.must_pass_in_stack}
4080 in the definition of this macro to determine if this argument is of a
4081 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4082 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
4083 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4084 defined, the argument will be computed in the stack and then loaded into
4088 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, const_tree @var{type})
4089 This target hook should return @code{true} if we should not pass @var{type}
4090 solely in registers. The file @file{expr.h} defines a
4091 definition that is usually appropriate, refer to @file{expr.h} for additional
4095 @deftypefn {Target Hook} rtx TARGET_FUNCTION_INCOMING_ARG (cumulative_args_t @var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4096 Define this hook if the target machine has ``register windows'', so
4097 that the register in which a function sees an arguments is not
4098 necessarily the same as the one in which the caller passed the
4101 For such machines, @code{TARGET_FUNCTION_ARG} computes the register in
4102 which the caller passes the value, and
4103 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4104 fashion to tell the function being called where the arguments will
4107 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4108 @code{TARGET_FUNCTION_ARG} serves both purposes.
4111 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4112 This target hook returns the number of bytes at the beginning of an
4113 argument that must be put in registers. The value must be zero for
4114 arguments that are passed entirely in registers or that are entirely
4115 pushed on the stack.
4117 On some machines, certain arguments must be passed partially in
4118 registers and partially in memory. On these machines, typically the
4119 first few words of arguments are passed in registers, and the rest
4120 on the stack. If a multi-word argument (a @code{double} or a
4121 structure) crosses that boundary, its first few words must be passed
4122 in registers and the rest must be pushed. This macro tells the
4123 compiler when this occurs, and how many bytes should go in registers.
4125 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4126 register to be used by the caller for this argument; likewise
4127 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4130 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4131 This target hook should return @code{true} if an argument at the
4132 position indicated by @var{cum} should be passed by reference. This
4133 predicate is queried after target independent reasons for being
4134 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4136 If the hook returns true, a copy of that argument is made in memory and a
4137 pointer to the argument is passed instead of the argument itself.
4138 The pointer is passed in whatever way is appropriate for passing a pointer
4142 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4143 The function argument described by the parameters to this hook is
4144 known to be passed by reference. The hook should return true if the
4145 function argument should be copied by the callee instead of copied
4148 For any argument for which the hook returns true, if it can be
4149 determined that the argument is not modified, then a copy need
4152 The default version of this hook always returns false.
4155 @defmac CUMULATIVE_ARGS
4156 A C type for declaring a variable that is used as the first argument
4157 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4158 target machines, the type @code{int} suffices and can hold the number
4159 of bytes of argument so far.
4161 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4162 arguments that have been passed on the stack. The compiler has other
4163 variables to keep track of that. For target machines on which all
4164 arguments are passed on the stack, there is no need to store anything in
4165 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4166 should not be empty, so use @code{int}.
4169 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4170 If defined, this macro is called before generating any code for a
4171 function, but after the @var{cfun} descriptor for the function has been
4172 created. The back end may use this macro to update @var{cfun} to
4173 reflect an ABI other than that which would normally be used by default.
4174 If the compiler is generating code for a compiler-generated function,
4175 @var{fndecl} may be @code{NULL}.
4178 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4179 A C statement (sans semicolon) for initializing the variable
4180 @var{cum} for the state at the beginning of the argument list. The
4181 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4182 is the tree node for the data type of the function which will receive
4183 the args, or 0 if the args are to a compiler support library function.
4184 For direct calls that are not libcalls, @var{fndecl} contain the
4185 declaration node of the function. @var{fndecl} is also set when
4186 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4187 being compiled. @var{n_named_args} is set to the number of named
4188 arguments, including a structure return address if it is passed as a
4189 parameter, when making a call. When processing incoming arguments,
4190 @var{n_named_args} is set to @minus{}1.
4192 When processing a call to a compiler support library function,
4193 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4194 contains the name of the function, as a string. @var{libname} is 0 when
4195 an ordinary C function call is being processed. Thus, each time this
4196 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4197 never both of them at once.
4200 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4201 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4202 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4203 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4204 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4205 0)} is used instead.
4208 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4209 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4210 finding the arguments for the function being compiled. If this macro is
4211 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4213 The value passed for @var{libname} is always 0, since library routines
4214 with special calling conventions are never compiled with GCC@. The
4215 argument @var{libname} exists for symmetry with
4216 @code{INIT_CUMULATIVE_ARGS}.
4217 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4218 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4221 @deftypefn {Target Hook} void TARGET_FUNCTION_ARG_ADVANCE (cumulative_args_t @var{ca}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4222 This hook updates the summarizer variable pointed to by @var{ca} to
4223 advance past an argument in the argument list. The values @var{mode},
4224 @var{type} and @var{named} describe that argument. Once this is done,
4225 the variable @var{cum} is suitable for analyzing the @emph{following}
4226 argument with @code{TARGET_FUNCTION_ARG}, etc.
4228 This hook need not do anything if the argument in question was passed
4229 on the stack. The compiler knows how to track the amount of stack space
4230 used for arguments without any special help.
4233 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4234 If defined, a C expression that is the number of bytes to add to the
4235 offset of the argument passed in memory. This is needed for the SPU,
4236 which passes @code{char} and @code{short} arguments in the preferred
4237 slot that is in the middle of the quad word instead of starting at the
4241 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4242 If defined, a C expression which determines whether, and in which direction,
4243 to pad out an argument with extra space. The value should be of type
4244 @code{enum direction}: either @code{upward} to pad above the argument,
4245 @code{downward} to pad below, or @code{none} to inhibit padding.
4247 The @emph{amount} of padding is not controlled by this macro, but by the
4248 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
4249 always just enough to reach the next multiple of that boundary.
4251 This macro has a default definition which is right for most systems.
4252 For little-endian machines, the default is to pad upward. For
4253 big-endian machines, the default is to pad downward for an argument of
4254 constant size shorter than an @code{int}, and upward otherwise.
4257 @defmac PAD_VARARGS_DOWN
4258 If defined, a C expression which determines whether the default
4259 implementation of va_arg will attempt to pad down before reading the
4260 next argument, if that argument is smaller than its aligned space as
4261 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4262 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4265 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4266 Specify padding for the last element of a block move between registers and
4267 memory. @var{first} is nonzero if this is the only element. Defining this
4268 macro allows better control of register function parameters on big-endian
4269 machines, without using @code{PARALLEL} rtl. In particular,
4270 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4271 registers, as there is no longer a "wrong" part of a register; For example,
4272 a three byte aggregate may be passed in the high part of a register if so
4276 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4277 This hook returns the alignment boundary, in bits, of an argument
4278 with the specified mode and type. The default hook returns
4279 @code{PARM_BOUNDARY} for all arguments.
4282 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_ROUND_BOUNDARY (enum machine_mode @var{mode}, const_tree @var{type})
4283 Normally, the size of an argument is rounded up to @code{PARM_BOUNDARY},
4284 which is the default value for this hook. You can define this hook to
4285 return a different value if an argument size must be rounded to a larger
4289 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4290 A C expression that is nonzero if @var{regno} is the number of a hard
4291 register in which function arguments are sometimes passed. This does
4292 @emph{not} include implicit arguments such as the static chain and
4293 the structure-value address. On many machines, no registers can be
4294 used for this purpose since all function arguments are pushed on the
4298 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4299 This hook should return true if parameter of type @var{type} are passed
4300 as two scalar parameters. By default, GCC will attempt to pack complex
4301 arguments into the target's word size. Some ABIs require complex arguments
4302 to be split and treated as their individual components. For example, on
4303 AIX64, complex floats should be passed in a pair of floating point
4304 registers, even though a complex float would fit in one 64-bit floating
4307 The default value of this hook is @code{NULL}, which is treated as always
4311 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4312 This hook returns a type node for @code{va_list} for the target.
4313 The default version of the hook returns @code{void*}.
4316 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4317 This target hook is used in function @code{c_common_nodes_and_builtins}
4318 to iterate through the target specific builtin types for va_list. The
4319 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4320 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4322 The arguments @var{pname} and @var{ptree} are used to store the result of
4323 this macro and are set to the name of the va_list builtin type and its
4325 If the return value of this macro is zero, then there is no more element.
4326 Otherwise the @var{IDX} should be increased for the next call of this
4327 macro to iterate through all types.
4330 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4331 This hook returns the va_list type of the calling convention specified by
4333 The default version of this hook returns @code{va_list_type_node}.
4336 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4337 This hook returns the va_list type of the calling convention specified by the
4338 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4342 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, gimple_seq *@var{pre_p}, gimple_seq *@var{post_p})
4343 This hook performs target-specific gimplification of
4344 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4345 arguments to @code{va_arg}; the latter two are as in
4346 @code{gimplify.c:gimplify_expr}.
4349 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4350 Define this to return nonzero if the port can handle pointers
4351 with machine mode @var{mode}. The default version of this
4352 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4355 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (struct ao_ref *@var{ref})
4356 Define this to return nonzero if the memory reference @var{ref} may alias with the system C library errno location. The default version of this hook assumes the system C library errno location is either a declaration of type int or accessed by dereferencing a pointer to int.
4359 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4360 Define this to return nonzero if the port is prepared to handle
4361 insns involving scalar mode @var{mode}. For a scalar mode to be
4362 considered supported, all the basic arithmetic and comparisons
4365 The default version of this hook returns true for any mode
4366 required to handle the basic C types (as defined by the port).
4367 Included here are the double-word arithmetic supported by the
4368 code in @file{optabs.c}.
4371 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4372 Define this to return nonzero if the port is prepared to handle
4373 insns involving vector mode @var{mode}. At the very least, it
4374 must have move patterns for this mode.
4377 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (enum machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4378 Return true if GCC should try to use a scalar mode to store an array
4379 of @var{nelems} elements, given that each element has mode @var{mode}.
4380 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4381 and allows GCC to use any defined integer mode.
4383 One use of this hook is to support vector load and store operations
4384 that operate on several homogeneous vectors. For example, ARM NEON
4385 has operations like:
4388 int8x8x3_t vld3_s8 (const int8_t *)
4391 where the return type is defined as:
4394 typedef struct int8x8x3_t
4400 If this hook allows @code{val} to have a scalar mode, then
4401 @code{int8x8x3_t} can have the same mode. GCC can then store
4402 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4405 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (enum machine_mode @var{mode})
4406 Define this to return nonzero for machine modes for which the port has
4407 small register classes. If this target hook returns nonzero for a given
4408 @var{mode}, the compiler will try to minimize the lifetime of registers
4409 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4410 In this case, the hook is expected to return nonzero if it returns nonzero
4413 On some machines, it is risky to let hard registers live across arbitrary
4414 insns. Typically, these machines have instructions that require values
4415 to be in specific registers (like an accumulator), and reload will fail
4416 if the required hard register is used for another purpose across such an
4419 Passes before reload do not know which hard registers will be used
4420 in an instruction, but the machine modes of the registers set or used in
4421 the instruction are already known. And for some machines, register
4422 classes are small for, say, integer registers but not for floating point
4423 registers. For example, the AMD x86-64 architecture requires specific
4424 registers for the legacy x86 integer instructions, but there are many
4425 SSE registers for floating point operations. On such targets, a good
4426 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4427 machine modes but zero for the SSE register classes.
4429 The default version of this hook returns false for any mode. It is always
4430 safe to redefine this hook to return with a nonzero value. But if you
4431 unnecessarily define it, you will reduce the amount of optimizations
4432 that can be performed in some cases. If you do not define this hook
4433 to return a nonzero value when it is required, the compiler will run out
4434 of spill registers and print a fatal error message.
4437 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
4438 If the target has a dedicated flags register, and it needs to use the post-reload comparison elimination pass, then this value should be set appropriately.
4442 @subsection How Scalar Function Values Are Returned
4443 @cindex return values in registers
4444 @cindex values, returned by functions
4445 @cindex scalars, returned as values
4447 This section discusses the macros that control returning scalars as
4448 values---values that can fit in registers.
4450 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4452 Define this to return an RTX representing the place where a function
4453 returns or receives a value of data type @var{ret_type}, a tree node
4454 representing a data type. @var{fn_decl_or_type} is a tree node
4455 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4456 function being called. If @var{outgoing} is false, the hook should
4457 compute the register in which the caller will see the return value.
4458 Otherwise, the hook should return an RTX representing the place where
4459 a function returns a value.
4461 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4462 (Actually, on most machines, scalar values are returned in the same
4463 place regardless of mode.) The value of the expression is usually a
4464 @code{reg} RTX for the hard register where the return value is stored.
4465 The value can also be a @code{parallel} RTX, if the return value is in
4466 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4467 @code{parallel} form. Note that the callee will populate every
4468 location specified in the @code{parallel}, but if the first element of
4469 the @code{parallel} contains the whole return value, callers will use
4470 that element as the canonical location and ignore the others. The m68k
4471 port uses this type of @code{parallel} to return pointers in both
4472 @samp{%a0} (the canonical location) and @samp{%d0}.
4474 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4475 the same promotion rules specified in @code{PROMOTE_MODE} if
4476 @var{valtype} is a scalar type.
4478 If the precise function being called is known, @var{func} is a tree
4479 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4480 pointer. This makes it possible to use a different value-returning
4481 convention for specific functions when all their calls are
4484 Some target machines have ``register windows'' so that the register in
4485 which a function returns its value is not the same as the one in which
4486 the caller sees the value. For such machines, you should return
4487 different RTX depending on @var{outgoing}.
4489 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4490 aggregate data types, because these are returned in another way. See
4491 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4494 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4495 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4496 a new target instead.
4499 @defmac LIBCALL_VALUE (@var{mode})
4500 A C expression to create an RTX representing the place where a library
4501 function returns a value of mode @var{mode}.
4503 Note that ``library function'' in this context means a compiler
4504 support routine, used to perform arithmetic, whose name is known
4505 specially by the compiler and was not mentioned in the C code being
4509 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (enum machine_mode @var{mode}, const_rtx @var{fun})
4510 Define this hook if the back-end needs to know the name of the libcall
4511 function in order to determine where the result should be returned.
4513 The mode of the result is given by @var{mode} and the name of the called
4514 library function is given by @var{fun}. The hook should return an RTX
4515 representing the place where the library function result will be returned.
4517 If this hook is not defined, then LIBCALL_VALUE will be used.
4520 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4521 A C expression that is nonzero if @var{regno} is the number of a hard
4522 register in which the values of called function may come back.
4524 A register whose use for returning values is limited to serving as the
4525 second of a pair (for a value of type @code{double}, say) need not be
4526 recognized by this macro. So for most machines, this definition
4530 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4533 If the machine has register windows, so that the caller and the called
4534 function use different registers for the return value, this macro
4535 should recognize only the caller's register numbers.
4537 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4538 for a new target instead.
4541 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4542 A target hook that return @code{true} if @var{regno} is the number of a hard
4543 register in which the values of called function may come back.
4545 A register whose use for returning values is limited to serving as the
4546 second of a pair (for a value of type @code{double}, say) need not be
4547 recognized by this target hook.
4549 If the machine has register windows, so that the caller and the called
4550 function use different registers for the return value, this target hook
4551 should recognize only the caller's register numbers.
4553 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4556 @defmac APPLY_RESULT_SIZE
4557 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4558 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4559 saving and restoring an arbitrary return value.
4562 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4563 This hook should return true if values of type @var{type} are returned
4564 at the most significant end of a register (in other words, if they are
4565 padded at the least significant end). You can assume that @var{type}
4566 is returned in a register; the caller is required to check this.
4568 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4569 be able to hold the complete return value. For example, if a 1-, 2-
4570 or 3-byte structure is returned at the most significant end of a
4571 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4575 @node Aggregate Return
4576 @subsection How Large Values Are Returned
4577 @cindex aggregates as return values
4578 @cindex large return values
4579 @cindex returning aggregate values
4580 @cindex structure value address
4582 When a function value's mode is @code{BLKmode} (and in some other
4583 cases), the value is not returned according to
4584 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4585 caller passes the address of a block of memory in which the value
4586 should be stored. This address is called the @dfn{structure value
4589 This section describes how to control returning structure values in
4592 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4593 This target hook should return a nonzero value to say to return the
4594 function value in memory, just as large structures are always returned.
4595 Here @var{type} will be the data type of the value, and @var{fntype}
4596 will be the type of the function doing the returning, or @code{NULL} for
4599 Note that values of mode @code{BLKmode} must be explicitly handled
4600 by this function. Also, the option @option{-fpcc-struct-return}
4601 takes effect regardless of this macro. On most systems, it is
4602 possible to leave the hook undefined; this causes a default
4603 definition to be used, whose value is the constant 1 for @code{BLKmode}
4604 values, and 0 otherwise.
4606 Do not use this hook to indicate that structures and unions should always
4607 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4611 @defmac DEFAULT_PCC_STRUCT_RETURN
4612 Define this macro to be 1 if all structure and union return values must be
4613 in memory. Since this results in slower code, this should be defined
4614 only if needed for compatibility with other compilers or with an ABI@.
4615 If you define this macro to be 0, then the conventions used for structure
4616 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4619 If not defined, this defaults to the value 1.
4622 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4623 This target hook should return the location of the structure value
4624 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4625 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4626 be @code{NULL}, for libcalls. You do not need to define this target
4627 hook if the address is always passed as an ``invisible'' first
4630 On some architectures the place where the structure value address
4631 is found by the called function is not the same place that the
4632 caller put it. This can be due to register windows, or it could
4633 be because the function prologue moves it to a different place.
4634 @var{incoming} is @code{1} or @code{2} when the location is needed in
4635 the context of the called function, and @code{0} in the context of
4638 If @var{incoming} is nonzero and the address is to be found on the
4639 stack, return a @code{mem} which refers to the frame pointer. If
4640 @var{incoming} is @code{2}, the result is being used to fetch the
4641 structure value address at the beginning of a function. If you need
4642 to emit adjusting code, you should do it at this point.
4645 @defmac PCC_STATIC_STRUCT_RETURN
4646 Define this macro if the usual system convention on the target machine
4647 for returning structures and unions is for the called function to return
4648 the address of a static variable containing the value.
4650 Do not define this if the usual system convention is for the caller to
4651 pass an address to the subroutine.
4653 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4654 nothing when you use @option{-freg-struct-return} mode.
4657 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4658 This target hook returns the mode to be used when accessing raw return registers in @code{__builtin_return}. Define this macro if the value in @var{reg_raw_mode} is not correct.
4661 @deftypefn {Target Hook} {enum machine_mode} TARGET_GET_RAW_ARG_MODE (int @var{regno})
4662 This target hook returns the mode to be used when accessing raw argument registers in @code{__builtin_apply_args}. Define this macro if the value in @var{reg_raw_mode} is not correct.
4666 @subsection Caller-Saves Register Allocation
4668 If you enable it, GCC can save registers around function calls. This
4669 makes it possible to use call-clobbered registers to hold variables that
4670 must live across calls.
4672 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4673 A C expression to determine whether it is worthwhile to consider placing
4674 a pseudo-register in a call-clobbered hard register and saving and
4675 restoring it around each function call. The expression should be 1 when
4676 this is worth doing, and 0 otherwise.
4678 If you don't define this macro, a default is used which is good on most
4679 machines: @code{4 * @var{calls} < @var{refs}}.
4682 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4683 A C expression specifying which mode is required for saving @var{nregs}
4684 of a pseudo-register in call-clobbered hard register @var{regno}. If
4685 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4686 returned. For most machines this macro need not be defined since GCC
4687 will select the smallest suitable mode.
4690 @node Function Entry
4691 @subsection Function Entry and Exit
4692 @cindex function entry and exit
4696 This section describes the macros that output function entry
4697 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4699 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4700 If defined, a function that outputs the assembler code for entry to a
4701 function. The prologue is responsible for setting up the stack frame,
4702 initializing the frame pointer register, saving registers that must be
4703 saved, and allocating @var{size} additional bytes of storage for the
4704 local variables. @var{size} is an integer. @var{file} is a stdio
4705 stream to which the assembler code should be output.
4707 The label for the beginning of the function need not be output by this
4708 macro. That has already been done when the macro is run.
4710 @findex regs_ever_live
4711 To determine which registers to save, the macro can refer to the array
4712 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4713 @var{r} is used anywhere within the function. This implies the function
4714 prologue should save register @var{r}, provided it is not one of the
4715 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4716 @code{regs_ever_live}.)
4718 On machines that have ``register windows'', the function entry code does
4719 not save on the stack the registers that are in the windows, even if
4720 they are supposed to be preserved by function calls; instead it takes
4721 appropriate steps to ``push'' the register stack, if any non-call-used
4722 registers are used in the function.
4724 @findex frame_pointer_needed
4725 On machines where functions may or may not have frame-pointers, the
4726 function entry code must vary accordingly; it must set up the frame
4727 pointer if one is wanted, and not otherwise. To determine whether a
4728 frame pointer is in wanted, the macro can refer to the variable
4729 @code{frame_pointer_needed}. The variable's value will be 1 at run
4730 time in a function that needs a frame pointer. @xref{Elimination}.
4732 The function entry code is responsible for allocating any stack space
4733 required for the function. This stack space consists of the regions
4734 listed below. In most cases, these regions are allocated in the
4735 order listed, with the last listed region closest to the top of the
4736 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4737 the highest address if it is not defined). You can use a different order
4738 for a machine if doing so is more convenient or required for
4739 compatibility reasons. Except in cases where required by standard
4740 or by a debugger, there is no reason why the stack layout used by GCC
4741 need agree with that used by other compilers for a machine.
4744 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4745 If defined, a function that outputs assembler code at the end of a
4746 prologue. This should be used when the function prologue is being
4747 emitted as RTL, and you have some extra assembler that needs to be
4748 emitted. @xref{prologue instruction pattern}.
4751 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4752 If defined, a function that outputs assembler code at the start of an
4753 epilogue. This should be used when the function epilogue is being
4754 emitted as RTL, and you have some extra assembler that needs to be
4755 emitted. @xref{epilogue instruction pattern}.
4758 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4759 If defined, a function that outputs the assembler code for exit from a
4760 function. The epilogue is responsible for restoring the saved
4761 registers and stack pointer to their values when the function was
4762 called, and returning control to the caller. This macro takes the
4763 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4764 registers to restore are determined from @code{regs_ever_live} and
4765 @code{CALL_USED_REGISTERS} in the same way.
4767 On some machines, there is a single instruction that does all the work
4768 of returning from the function. On these machines, give that
4769 instruction the name @samp{return} and do not define the macro
4770 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4772 Do not define a pattern named @samp{return} if you want the
4773 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4774 switches to control whether return instructions or epilogues are used,
4775 define a @samp{return} pattern with a validity condition that tests the
4776 target switches appropriately. If the @samp{return} pattern's validity
4777 condition is false, epilogues will be used.
4779 On machines where functions may or may not have frame-pointers, the
4780 function exit code must vary accordingly. Sometimes the code for these
4781 two cases is completely different. To determine whether a frame pointer
4782 is wanted, the macro can refer to the variable
4783 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4784 a function that needs a frame pointer.
4786 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4787 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4788 The C variable @code{current_function_is_leaf} is nonzero for such a
4789 function. @xref{Leaf Functions}.
4791 On some machines, some functions pop their arguments on exit while
4792 others leave that for the caller to do. For example, the 68020 when
4793 given @option{-mrtd} pops arguments in functions that take a fixed
4794 number of arguments.
4797 @findex crtl->args.pops_args
4798 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4799 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4800 needs to know what was decided. The number of bytes of the current
4801 function's arguments that this function should pop is available in
4802 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4807 @findex pretend_args_size
4808 @findex crtl->args.pretend_args_size
4809 A region of @code{crtl->args.pretend_args_size} bytes of
4810 uninitialized space just underneath the first argument arriving on the
4811 stack. (This may not be at the very start of the allocated stack region
4812 if the calling sequence has pushed anything else since pushing the stack
4813 arguments. But usually, on such machines, nothing else has been pushed
4814 yet, because the function prologue itself does all the pushing.) This
4815 region is used on machines where an argument may be passed partly in
4816 registers and partly in memory, and, in some cases to support the
4817 features in @code{<stdarg.h>}.
4820 An area of memory used to save certain registers used by the function.
4821 The size of this area, which may also include space for such things as
4822 the return address and pointers to previous stack frames, is
4823 machine-specific and usually depends on which registers have been used
4824 in the function. Machines with register windows often do not require
4828 A region of at least @var{size} bytes, possibly rounded up to an allocation
4829 boundary, to contain the local variables of the function. On some machines,
4830 this region and the save area may occur in the opposite order, with the
4831 save area closer to the top of the stack.
4834 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4835 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4836 @code{crtl->outgoing_args_size} bytes to be used for outgoing
4837 argument lists of the function. @xref{Stack Arguments}.
4840 @defmac EXIT_IGNORE_STACK
4841 Define this macro as a C expression that is nonzero if the return
4842 instruction or the function epilogue ignores the value of the stack
4843 pointer; in other words, if it is safe to delete an instruction to
4844 adjust the stack pointer before a return from the function. The
4847 Note that this macro's value is relevant only for functions for which
4848 frame pointers are maintained. It is never safe to delete a final
4849 stack adjustment in a function that has no frame pointer, and the
4850 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4853 @defmac EPILOGUE_USES (@var{regno})
4854 Define this macro as a C expression that is nonzero for registers that are
4855 used by the epilogue or the @samp{return} pattern. The stack and frame
4856 pointer registers are already assumed to be used as needed.
4859 @defmac EH_USES (@var{regno})
4860 Define this macro as a C expression that is nonzero for registers that are
4861 used by the exception handling mechanism, and so should be considered live
4862 on entry to an exception edge.
4865 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function})
4866 A function that outputs the assembler code for a thunk
4867 function, used to implement C++ virtual function calls with multiple
4868 inheritance. The thunk acts as a wrapper around a virtual function,
4869 adjusting the implicit object parameter before handing control off to
4872 First, emit code to add the integer @var{delta} to the location that
4873 contains the incoming first argument. Assume that this argument
4874 contains a pointer, and is the one used to pass the @code{this} pointer
4875 in C++. This is the incoming argument @emph{before} the function prologue,
4876 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4877 all other incoming arguments.
4879 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4880 made after adding @code{delta}. In particular, if @var{p} is the
4881 adjusted pointer, the following adjustment should be made:
4884 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4887 After the additions, emit code to jump to @var{function}, which is a
4888 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4889 not touch the return address. Hence returning from @var{FUNCTION} will
4890 return to whoever called the current @samp{thunk}.
4892 The effect must be as if @var{function} had been called directly with
4893 the adjusted first argument. This macro is responsible for emitting all
4894 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4895 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4897 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4898 have already been extracted from it.) It might possibly be useful on
4899 some targets, but probably not.
4901 If you do not define this macro, the target-independent code in the C++
4902 front end will generate a less efficient heavyweight thunk that calls
4903 @var{function} instead of jumping to it. The generic approach does
4904 not support varargs.
4907 @deftypefn {Target Hook} bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (const_tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, const_tree @var{function})
4908 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4909 to output the assembler code for the thunk function specified by the
4910 arguments it is passed, and false otherwise. In the latter case, the
4911 generic approach will be used by the C++ front end, with the limitations
4916 @subsection Generating Code for Profiling
4917 @cindex profiling, code generation
4919 These macros will help you generate code for profiling.
4921 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4922 A C statement or compound statement to output to @var{file} some
4923 assembler code to call the profiling subroutine @code{mcount}.
4926 The details of how @code{mcount} expects to be called are determined by
4927 your operating system environment, not by GCC@. To figure them out,
4928 compile a small program for profiling using the system's installed C
4929 compiler and look at the assembler code that results.
4931 Older implementations of @code{mcount} expect the address of a counter
4932 variable to be loaded into some register. The name of this variable is
4933 @samp{LP} followed by the number @var{labelno}, so you would generate
4934 the name using @samp{LP%d} in a @code{fprintf}.
4937 @defmac PROFILE_HOOK
4938 A C statement or compound statement to output to @var{file} some assembly
4939 code to call the profiling subroutine @code{mcount} even the target does
4940 not support profiling.
4943 @defmac NO_PROFILE_COUNTERS
4944 Define this macro to be an expression with a nonzero value if the
4945 @code{mcount} subroutine on your system does not need a counter variable
4946 allocated for each function. This is true for almost all modern
4947 implementations. If you define this macro, you must not use the
4948 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4951 @defmac PROFILE_BEFORE_PROLOGUE
4952 Define this macro if the code for function profiling should come before
4953 the function prologue. Normally, the profiling code comes after.
4957 @subsection Permitting tail calls
4960 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4961 True if it is OK to do sibling call optimization for the specified
4962 call expression @var{exp}. @var{decl} will be the called function,
4963 or @code{NULL} if this is an indirect call.
4965 It is not uncommon for limitations of calling conventions to prevent
4966 tail calls to functions outside the current unit of translation, or
4967 during PIC compilation. The hook is used to enforce these restrictions,
4968 as the @code{sibcall} md pattern can not fail, or fall over to a
4969 ``normal'' call. The criteria for successful sibling call optimization
4970 may vary greatly between different architectures.
4973 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4974 Add any hard registers to @var{regs} that are live on entry to the
4975 function. This hook only needs to be defined to provide registers that
4976 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4977 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4978 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4979 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4982 @deftypefn {Target Hook} void TARGET_SET_UP_BY_PROLOGUE (struct hard_reg_set_container *@var{})
4983 This hook should add additional registers that are computed by the prologue to the hard regset for shrink-wrapping optimization purposes.
4986 @deftypefn {Target Hook} bool TARGET_WARN_FUNC_RETURN (tree)
4987 True if a function's return statements should be checked for matching the function's return type. This includes checking for falling off the end of a non-void function. Return false if no such check should be made.
4990 @node Stack Smashing Protection
4991 @subsection Stack smashing protection
4992 @cindex stack smashing protection
4994 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4995 This hook returns a @code{DECL} node for the external variable to use
4996 for the stack protection guard. This variable is initialized by the
4997 runtime to some random value and is used to initialize the guard value
4998 that is placed at the top of the local stack frame. The type of this
4999 variable must be @code{ptr_type_node}.
5001 The default version of this hook creates a variable called
5002 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
5005 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
5006 This hook returns a @code{CALL_EXPR} that alerts the runtime that the
5007 stack protect guard variable has been modified. This expression should
5008 involve a call to a @code{noreturn} function.
5010 The default version of this hook invokes a function called
5011 @samp{__stack_chk_fail}, taking no arguments. This function is
5012 normally defined in @file{libgcc2.c}.
5015 @deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
5016 Whether this target supports splitting the stack when the options described in @var{opts} have been passed. This is called after options have been parsed, so the target may reject splitting the stack in some configurations. The default version of this hook returns false. If @var{report} is true, this function may issue a warning or error; if @var{report} is false, it must simply return a value
5020 @section Implementing the Varargs Macros
5021 @cindex varargs implementation
5023 GCC comes with an implementation of @code{<varargs.h>} and
5024 @code{<stdarg.h>} that work without change on machines that pass arguments
5025 on the stack. Other machines require their own implementations of
5026 varargs, and the two machine independent header files must have
5027 conditionals to include it.
5029 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
5030 the calling convention for @code{va_start}. The traditional
5031 implementation takes just one argument, which is the variable in which
5032 to store the argument pointer. The ISO implementation of
5033 @code{va_start} takes an additional second argument. The user is
5034 supposed to write the last named argument of the function here.
5036 However, @code{va_start} should not use this argument. The way to find
5037 the end of the named arguments is with the built-in functions described
5040 @defmac __builtin_saveregs ()
5041 Use this built-in function to save the argument registers in memory so
5042 that the varargs mechanism can access them. Both ISO and traditional
5043 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
5044 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
5046 On some machines, @code{__builtin_saveregs} is open-coded under the
5047 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
5048 other machines, it calls a routine written in assembler language,
5049 found in @file{libgcc2.c}.
5051 Code generated for the call to @code{__builtin_saveregs} appears at the
5052 beginning of the function, as opposed to where the call to
5053 @code{__builtin_saveregs} is written, regardless of what the code is.
5054 This is because the registers must be saved before the function starts
5055 to use them for its own purposes.
5056 @c i rewrote the first sentence above to fix an overfull hbox. --mew
5060 @defmac __builtin_next_arg (@var{lastarg})
5061 This builtin returns the address of the first anonymous stack
5062 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
5063 returns the address of the location above the first anonymous stack
5064 argument. Use it in @code{va_start} to initialize the pointer for
5065 fetching arguments from the stack. Also use it in @code{va_start} to
5066 verify that the second parameter @var{lastarg} is the last named argument
5067 of the current function.
5070 @defmac __builtin_classify_type (@var{object})
5071 Since each machine has its own conventions for which data types are
5072 passed in which kind of register, your implementation of @code{va_arg}
5073 has to embody these conventions. The easiest way to categorize the
5074 specified data type is to use @code{__builtin_classify_type} together
5075 with @code{sizeof} and @code{__alignof__}.
5077 @code{__builtin_classify_type} ignores the value of @var{object},
5078 considering only its data type. It returns an integer describing what
5079 kind of type that is---integer, floating, pointer, structure, and so on.
5081 The file @file{typeclass.h} defines an enumeration that you can use to
5082 interpret the values of @code{__builtin_classify_type}.
5085 These machine description macros help implement varargs:
5087 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5088 If defined, this hook produces the machine-specific code for a call to
5089 @code{__builtin_saveregs}. This code will be moved to the very
5090 beginning of the function, before any parameter access are made. The
5091 return value of this function should be an RTX that contains the value
5092 to use as the return of @code{__builtin_saveregs}.
5095 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (cumulative_args_t @var{args_so_far}, enum machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
5096 This target hook offers an alternative to using
5097 @code{__builtin_saveregs} and defining the hook
5098 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5099 register arguments into the stack so that all the arguments appear to
5100 have been passed consecutively on the stack. Once this is done, you can
5101 use the standard implementation of varargs that works for machines that
5102 pass all their arguments on the stack.
5104 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5105 structure, containing the values that are obtained after processing the
5106 named arguments. The arguments @var{mode} and @var{type} describe the
5107 last named argument---its machine mode and its data type as a tree node.
5109 The target hook should do two things: first, push onto the stack all the
5110 argument registers @emph{not} used for the named arguments, and second,
5111 store the size of the data thus pushed into the @code{int}-valued
5112 variable pointed to by @var{pretend_args_size}. The value that you
5113 store here will serve as additional offset for setting up the stack
5116 Because you must generate code to push the anonymous arguments at
5117 compile time without knowing their data types,
5118 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5119 have just a single category of argument register and use it uniformly
5122 If the argument @var{second_time} is nonzero, it means that the
5123 arguments of the function are being analyzed for the second time. This
5124 happens for an inline function, which is not actually compiled until the
5125 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5126 not generate any instructions in this case.
5129 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca})
5130 Define this hook to return @code{true} if the location where a function
5131 argument is passed depends on whether or not it is a named argument.
5133 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5134 is set for varargs and stdarg functions. If this hook returns
5135 @code{true}, the @var{named} argument is always true for named
5136 arguments, and false for unnamed arguments. If it returns @code{false},
5137 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5138 then all arguments are treated as named. Otherwise, all named arguments
5139 except the last are treated as named.
5141 You need not define this hook if it always returns @code{false}.
5144 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca})
5145 If you need to conditionally change ABIs so that one works with
5146 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5147 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5148 defined, then define this hook to return @code{true} if
5149 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5150 Otherwise, you should not define this hook.
5154 @section Trampolines for Nested Functions
5155 @cindex trampolines for nested functions
5156 @cindex nested functions, trampolines for
5158 A @dfn{trampoline} is a small piece of code that is created at run time
5159 when the address of a nested function is taken. It normally resides on
5160 the stack, in the stack frame of the containing function. These macros
5161 tell GCC how to generate code to allocate and initialize a
5164 The instructions in the trampoline must do two things: load a constant
5165 address into the static chain register, and jump to the real address of
5166 the nested function. On CISC machines such as the m68k, this requires
5167 two instructions, a move immediate and a jump. Then the two addresses
5168 exist in the trampoline as word-long immediate operands. On RISC
5169 machines, it is often necessary to load each address into a register in
5170 two parts. Then pieces of each address form separate immediate
5173 The code generated to initialize the trampoline must store the variable
5174 parts---the static chain value and the function address---into the
5175 immediate operands of the instructions. On a CISC machine, this is
5176 simply a matter of copying each address to a memory reference at the
5177 proper offset from the start of the trampoline. On a RISC machine, it
5178 may be necessary to take out pieces of the address and store them
5181 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5182 This hook is called by @code{assemble_trampoline_template} to output,
5183 on the stream @var{f}, assembler code for a block of data that contains
5184 the constant parts of a trampoline. This code should not include a
5185 label---the label is taken care of automatically.
5187 If you do not define this hook, it means no template is needed
5188 for the target. Do not define this hook on systems where the block move
5189 code to copy the trampoline into place would be larger than the code
5190 to generate it on the spot.
5193 @defmac TRAMPOLINE_SECTION
5194 Return the section into which the trampoline template is to be placed
5195 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5198 @defmac TRAMPOLINE_SIZE
5199 A C expression for the size in bytes of the trampoline, as an integer.
5202 @defmac TRAMPOLINE_ALIGNMENT
5203 Alignment required for trampolines, in bits.
5205 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5206 is used for aligning trampolines.
5209 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5210 This hook is called to initialize a trampoline.
5211 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5212 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5213 RTX for the static chain value that should be passed to the function
5216 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5217 first thing this hook should do is emit a block move into @var{m_tramp}
5218 from the memory block returned by @code{assemble_trampoline_template}.
5219 Note that the block move need only cover the constant parts of the
5220 trampoline. If the target isolates the variable parts of the trampoline
5221 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5223 If the target requires any other actions, such as flushing caches or
5224 enabling stack execution, these actions should be performed after
5225 initializing the trampoline proper.
5228 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5229 This hook should perform any machine-specific adjustment in
5230 the address of the trampoline. Its argument contains the address of the
5231 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5232 the address to be used for a function call should be different from the
5233 address at which the template was stored, the different address should
5234 be returned; otherwise @var{addr} should be returned unchanged.
5235 If this hook is not defined, @var{addr} will be used for function calls.
5238 Implementing trampolines is difficult on many machines because they have
5239 separate instruction and data caches. Writing into a stack location
5240 fails to clear the memory in the instruction cache, so when the program
5241 jumps to that location, it executes the old contents.
5243 Here are two possible solutions. One is to clear the relevant parts of
5244 the instruction cache whenever a trampoline is set up. The other is to
5245 make all trampolines identical, by having them jump to a standard
5246 subroutine. The former technique makes trampoline execution faster; the
5247 latter makes initialization faster.
5249 To clear the instruction cache when a trampoline is initialized, define
5250 the following macro.
5252 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5253 If defined, expands to a C expression clearing the @emph{instruction
5254 cache} in the specified interval. The definition of this macro would
5255 typically be a series of @code{asm} statements. Both @var{beg} and
5256 @var{end} are both pointer expressions.
5259 To use a standard subroutine, define the following macro. In addition,
5260 you must make sure that the instructions in a trampoline fill an entire
5261 cache line with identical instructions, or else ensure that the
5262 beginning of the trampoline code is always aligned at the same point in
5263 its cache line. Look in @file{m68k.h} as a guide.
5265 @defmac TRANSFER_FROM_TRAMPOLINE
5266 Define this macro if trampolines need a special subroutine to do their
5267 work. The macro should expand to a series of @code{asm} statements
5268 which will be compiled with GCC@. They go in a library function named
5269 @code{__transfer_from_trampoline}.
5271 If you need to avoid executing the ordinary prologue code of a compiled
5272 C function when you jump to the subroutine, you can do so by placing a
5273 special label of your own in the assembler code. Use one @code{asm}
5274 statement to generate an assembler label, and another to make the label
5275 global. Then trampolines can use that label to jump directly to your
5276 special assembler code.
5280 @section Implicit Calls to Library Routines
5281 @cindex library subroutine names
5282 @cindex @file{libgcc.a}
5284 @c prevent bad page break with this line
5285 Here is an explanation of implicit calls to library routines.
5287 @defmac DECLARE_LIBRARY_RENAMES
5288 This macro, if defined, should expand to a piece of C code that will get
5289 expanded when compiling functions for libgcc.a. It can be used to
5290 provide alternate names for GCC's internal library functions if there
5291 are ABI-mandated names that the compiler should provide.
5294 @findex set_optab_libfunc
5295 @findex init_one_libfunc
5296 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5297 This hook should declare additional library routines or rename
5298 existing ones, using the functions @code{set_optab_libfunc} and
5299 @code{init_one_libfunc} defined in @file{optabs.c}.
5300 @code{init_optabs} calls this macro after initializing all the normal
5303 The default is to do nothing. Most ports don't need to define this hook.
5306 @deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX
5307 If false (the default), internal library routines start with two
5308 underscores. If set to true, these routines start with @code{__gnu_}
5309 instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This
5310 currently only affects functions defined in @file{libgcc2.c}. If this
5311 is set to true, the @file{tm.h} file must also
5312 @code{#define LIBGCC2_GNU_PREFIX}.
5315 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5316 This macro should return @code{true} if the library routine that
5317 implements the floating point comparison operator @var{comparison} in
5318 mode @var{mode} will return a boolean, and @var{false} if it will
5321 GCC's own floating point libraries return tristates from the
5322 comparison operators, so the default returns false always. Most ports
5323 don't need to define this macro.
5326 @defmac TARGET_LIB_INT_CMP_BIASED
5327 This macro should evaluate to @code{true} if the integer comparison
5328 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5329 operand is smaller than the second, 1 to indicate that they are equal,
5330 and 2 to indicate that the first operand is greater than the second.
5331 If this macro evaluates to @code{false} the comparison functions return
5332 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5333 in @file{libgcc.a}, you do not need to define this macro.
5336 @defmac TARGET_HAS_NO_HW_DIVIDE
5337 This macro should be defined if the target has no hardware divide
5338 instructions. If this macro is defined, GCC will use an algorithm which
5339 make use of simple logical and arithmetic operations for 64-bit
5340 division. If the macro is not defined, GCC will use an algorithm which
5341 make use of a 64-bit by 32-bit divide primitive.
5344 @cindex @code{EDOM}, implicit usage
5347 The value of @code{EDOM} on the target machine, as a C integer constant
5348 expression. If you don't define this macro, GCC does not attempt to
5349 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5350 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5353 If you do not define @code{TARGET_EDOM}, then compiled code reports
5354 domain errors by calling the library function and letting it report the
5355 error. If mathematical functions on your system use @code{matherr} when
5356 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5357 that @code{matherr} is used normally.
5360 @cindex @code{errno}, implicit usage
5361 @defmac GEN_ERRNO_RTX
5362 Define this macro as a C expression to create an rtl expression that
5363 refers to the global ``variable'' @code{errno}. (On certain systems,
5364 @code{errno} may not actually be a variable.) If you don't define this
5365 macro, a reasonable default is used.
5368 @deftypefn {Target Hook} bool TARGET_LIBC_HAS_FUNCTION (enum function_class @var{fn_class})
5369 This hook determines whether a function from a class of functions
5370 @var{fn_class} is present at the runtime.
5373 @defmac NEXT_OBJC_RUNTIME
5374 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
5375 by default. This calling convention involves passing the object, the selector
5376 and the method arguments all at once to the method-lookup library function.
5377 This is the usual setting when targeting Darwin/Mac OS X systems, which have
5378 the NeXT runtime installed.
5380 If the macro is set to 0, the "GNU" Objective-C message sending convention
5381 will be used by default. This convention passes just the object and the
5382 selector to the method-lookup function, which returns a pointer to the method.
5384 In either case, it remains possible to select code-generation for the alternate
5385 scheme, by means of compiler command line switches.
5388 @node Addressing Modes
5389 @section Addressing Modes
5390 @cindex addressing modes
5392 @c prevent bad page break with this line
5393 This is about addressing modes.
5395 @defmac HAVE_PRE_INCREMENT
5396 @defmacx HAVE_PRE_DECREMENT
5397 @defmacx HAVE_POST_INCREMENT
5398 @defmacx HAVE_POST_DECREMENT
5399 A C expression that is nonzero if the machine supports pre-increment,
5400 pre-decrement, post-increment, or post-decrement addressing respectively.
5403 @defmac HAVE_PRE_MODIFY_DISP
5404 @defmacx HAVE_POST_MODIFY_DISP
5405 A C expression that is nonzero if the machine supports pre- or
5406 post-address side-effect generation involving constants other than
5407 the size of the memory operand.
5410 @defmac HAVE_PRE_MODIFY_REG
5411 @defmacx HAVE_POST_MODIFY_REG
5412 A C expression that is nonzero if the machine supports pre- or
5413 post-address side-effect generation involving a register displacement.
5416 @defmac CONSTANT_ADDRESS_P (@var{x})
5417 A C expression that is 1 if the RTX @var{x} is a constant which
5418 is a valid address. On most machines the default definition of
5419 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5420 is acceptable, but a few machines are more restrictive as to which
5421 constant addresses are supported.
5424 @defmac CONSTANT_P (@var{x})
5425 @code{CONSTANT_P}, which is defined by target-independent code,
5426 accepts integer-values expressions whose values are not explicitly
5427 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5428 expressions and @code{const} arithmetic expressions, in addition to
5429 @code{const_int} and @code{const_double} expressions.
5432 @defmac MAX_REGS_PER_ADDRESS
5433 A number, the maximum number of registers that can appear in a valid
5434 memory address. Note that it is up to you to specify a value equal to
5435 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5439 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5440 A function that returns whether @var{x} (an RTX) is a legitimate memory
5441 address on the target machine for a memory operand of mode @var{mode}.
5443 Legitimate addresses are defined in two variants: a strict variant and a
5444 non-strict one. The @var{strict} parameter chooses which variant is
5445 desired by the caller.
5447 The strict variant is used in the reload pass. It must be defined so
5448 that any pseudo-register that has not been allocated a hard register is
5449 considered a memory reference. This is because in contexts where some
5450 kind of register is required, a pseudo-register with no hard register
5451 must be rejected. For non-hard registers, the strict variant should look
5452 up the @code{reg_renumber} array; it should then proceed using the hard
5453 register number in the array, or treat the pseudo as a memory reference
5454 if the array holds @code{-1}.
5456 The non-strict variant is used in other passes. It must be defined to
5457 accept all pseudo-registers in every context where some kind of
5458 register is required.
5460 Normally, constant addresses which are the sum of a @code{symbol_ref}
5461 and an integer are stored inside a @code{const} RTX to mark them as
5462 constant. Therefore, there is no need to recognize such sums
5463 specifically as legitimate addresses. Normally you would simply
5464 recognize any @code{const} as legitimate.
5466 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5467 sums that are not marked with @code{const}. It assumes that a naked
5468 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5469 naked constant sums as illegitimate addresses, so that none of them will
5470 be given to @code{PRINT_OPERAND_ADDRESS}.
5472 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5473 On some machines, whether a symbolic address is legitimate depends on
5474 the section that the address refers to. On these machines, define the
5475 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5476 into the @code{symbol_ref}, and then check for it here. When you see a
5477 @code{const}, you will have to look inside it to find the
5478 @code{symbol_ref} in order to determine the section. @xref{Assembler
5481 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5482 Some ports are still using a deprecated legacy substitute for
5483 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5487 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5491 and should @code{goto @var{label}} if the address @var{x} is a valid
5492 address on the target machine for a memory operand of mode @var{mode}.
5494 @findex REG_OK_STRICT
5495 Compiler source files that want to use the strict variant of this
5496 macro define the macro @code{REG_OK_STRICT}. You should use an
5497 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5498 that case and the non-strict variant otherwise.
5500 Using the hook is usually simpler because it limits the number of
5501 files that are recompiled when changes are made.
5504 @defmac TARGET_MEM_CONSTRAINT
5505 A single character to be used instead of the default @code{'m'}
5506 character for general memory addresses. This defines the constraint
5507 letter which matches the memory addresses accepted by
5508 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5509 support new address formats in your back end without changing the
5510 semantics of the @code{'m'} constraint. This is necessary in order to
5511 preserve functionality of inline assembly constructs using the
5512 @code{'m'} constraint.
5515 @defmac FIND_BASE_TERM (@var{x})
5516 A C expression to determine the base term of address @var{x},
5517 or to provide a simplified version of @var{x} from which @file{alias.c}
5518 can easily find the base term. This macro is used in only two places:
5519 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5521 It is always safe for this macro to not be defined. It exists so
5522 that alias analysis can understand machine-dependent addresses.
5524 The typical use of this macro is to handle addresses containing
5525 a label_ref or symbol_ref within an UNSPEC@.
5528 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode})
5529 This hook is given an invalid memory address @var{x} for an
5530 operand of mode @var{mode} and should try to return a valid memory
5533 @findex break_out_memory_refs
5534 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5535 and @var{oldx} will be the operand that was given to that function to produce
5538 The code of the hook should not alter the substructure of
5539 @var{x}. If it transforms @var{x} into a more legitimate form, it
5540 should return the new @var{x}.
5542 It is not necessary for this hook to come up with a legitimate address,
5543 with the exception of native TLS addresses (@pxref{Emulated TLS}).
5544 The compiler has standard ways of doing so in all cases. In fact, if
5545 the target supports only emulated TLS, it
5546 is safe to omit this hook or make it return @var{x} if it cannot find
5547 a valid way to legitimize the address. But often a machine-dependent
5548 strategy can generate better code.
5551 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5552 A C compound statement that attempts to replace @var{x}, which is an address
5553 that needs reloading, with a valid memory address for an operand of mode
5554 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5555 It is not necessary to define this macro, but it might be useful for
5556 performance reasons.
5558 For example, on the i386, it is sometimes possible to use a single
5559 reload register instead of two by reloading a sum of two pseudo
5560 registers into a register. On the other hand, for number of RISC
5561 processors offsets are limited so that often an intermediate address
5562 needs to be generated in order to address a stack slot. By defining
5563 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5564 generated for adjacent some stack slots can be made identical, and thus
5567 @emph{Note}: This macro should be used with caution. It is necessary
5568 to know something of how reload works in order to effectively use this,
5569 and it is quite easy to produce macros that build in too much knowledge
5570 of reload internals.
5572 @emph{Note}: This macro must be able to reload an address created by a
5573 previous invocation of this macro. If it fails to handle such addresses
5574 then the compiler may generate incorrect code or abort.
5577 The macro definition should use @code{push_reload} to indicate parts that
5578 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5579 suitable to be passed unaltered to @code{push_reload}.
5581 The code generated by this macro must not alter the substructure of
5582 @var{x}. If it transforms @var{x} into a more legitimate form, it
5583 should assign @var{x} (which will always be a C variable) a new value.
5584 This also applies to parts that you change indirectly by calling
5587 @findex strict_memory_address_p
5588 The macro definition may use @code{strict_memory_address_p} to test if
5589 the address has become legitimate.
5592 If you want to change only a part of @var{x}, one standard way of doing
5593 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5594 single level of rtl. Thus, if the part to be changed is not at the
5595 top level, you'll need to replace first the top level.
5596 It is not necessary for this macro to come up with a legitimate
5597 address; but often a machine-dependent strategy can generate better code.
5600 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr}, addr_space_t @var{addrspace})
5601 This hook returns @code{true} if memory address @var{addr} in address
5602 space @var{addrspace} can have
5603 different meanings depending on the machine mode of the memory
5604 reference it is used for or if the address is valid for some modes
5607 Autoincrement and autodecrement addresses typically have mode-dependent
5608 effects because the amount of the increment or decrement is the size
5609 of the operand being addressed. Some machines have other mode-dependent
5610 addresses. Many RISC machines have no mode-dependent addresses.
5612 You may assume that @var{addr} is a valid address for the machine.
5614 The default version of this hook returns @code{false}.
5617 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5618 This hook returns true if @var{x} is a legitimate constant for a
5619 @var{mode}-mode immediate operand on the target machine. You can assume that
5620 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5622 The default definition returns true.
5625 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5626 This hook is used to undo the possibly obfuscating effects of the
5627 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5628 macros. Some backend implementations of these macros wrap symbol
5629 references inside an @code{UNSPEC} rtx to represent PIC or similar
5630 addressing modes. This target hook allows GCC's optimizers to understand
5631 the semantics of these opaque @code{UNSPEC}s by converting them back
5632 into their original form.
5635 @deftypefn {Target Hook} bool TARGET_CONST_NOT_OK_FOR_DEBUG_P (rtx @var{x})
5636 This hook should return true if @var{x} should not be emitted into
5640 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (enum machine_mode @var{mode}, rtx @var{x})
5641 This hook should return true if @var{x} is of a form that cannot (or
5642 should not) be spilled to the constant pool. @var{mode} is the mode
5645 The default version of this hook returns false.
5647 The primary reason to define this hook is to prevent reload from
5648 deciding that a non-legitimate constant would be better reloaded
5649 from the constant pool instead of spilling and reloading a register
5650 holding the constant. This restriction is often true of addresses
5651 of TLS symbols for various targets.
5654 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, const_rtx @var{x})
5655 This hook should return true if pool entries for constant @var{x} can
5656 be placed in an @code{object_block} structure. @var{mode} is the mode
5659 The default version returns false for all constants.
5662 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_DECL_P (const_tree @var{decl})
5663 This hook should return true if pool entries for @var{decl} should
5664 be placed in an @code{object_block} structure.
5666 The default version returns true for all decls.
5669 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (unsigned @var{fn}, bool @var{md_fn}, bool @var{sqrt})
5670 This hook should return the DECL of a function that implements reciprocal of
5671 the builtin function with builtin function code @var{fn}, or
5672 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5673 when @var{fn} is a code of a machine-dependent builtin function. When
5674 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5675 of a square root function are performed, and only reciprocals of @code{sqrt}
5679 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5680 This hook should return the DECL of a function @var{f} that given an
5681 address @var{addr} as an argument returns a mask @var{m} that can be
5682 used to extract from two vectors the relevant data that resides in
5683 @var{addr} in case @var{addr} is not properly aligned.
5685 The autovectorizer, when vectorizing a load operation from an address
5686 @var{addr} that may be unaligned, will generate two vector loads from
5687 the two aligned addresses around @var{addr}. It then generates a
5688 @code{REALIGN_LOAD} operation to extract the relevant data from the
5689 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5690 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5691 the third argument, @var{OFF}, defines how the data will be extracted
5692 from these two vectors: if @var{OFF} is 0, then the returned vector is
5693 @var{v2}; otherwise, the returned vector is composed from the last
5694 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5695 @var{OFF} elements of @var{v2}.
5697 If this hook is defined, the autovectorizer will generate a call
5698 to @var{f} (using the DECL tree that this hook returns) and will
5699 use the return value of @var{f} as the argument @var{OFF} to
5700 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5701 should comply with the semantics expected by @code{REALIGN_LOAD}
5703 If this hook is not defined, then @var{addr} will be used as
5704 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5705 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5708 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5709 Returns cost of different scalar or vector statements for vectorization cost model.
5710 For vector memory operations the cost may depend on type (@var{vectype}) and
5711 misalignment value (@var{misalign}).
5714 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5715 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5718 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VEC_PERM_CONST_OK (enum @var{machine_mode}, const unsigned char *@var{sel})
5719 Return true if a vector created for @code{vec_perm_const} is valid.
5722 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5723 This hook should return the DECL of a function that implements conversion of the
5724 input vector of type @var{src_type} to type @var{dest_type}.
5725 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5726 specifies how the conversion is to be applied
5727 (truncation, rounding, etc.).
5729 If this hook is defined, the autovectorizer will use the
5730 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5731 conversion. Otherwise, it will return @code{NULL_TREE}.
5734 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5735 This hook should return the decl of a function that implements the
5736 vectorized variant of the builtin function with builtin function code
5737 @var{code} or @code{NULL_TREE} if such a function is not available.
5738 The value of @var{fndecl} is the builtin function declaration. The
5739 return type of the vectorized function shall be of vector type
5740 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5743 @deftypefn {Target Hook} bool TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT (enum machine_mode @var{mode}, const_tree @var{type}, int @var{misalignment}, bool @var{is_packed})
5744 This hook should return true if the target supports misaligned vector
5745 store/load of a specific factor denoted in the @var{misalignment}
5746 parameter. The vector store/load should be of machine mode @var{mode} and
5747 the elements in the vectors should be of type @var{type}. @var{is_packed}
5748 parameter is true if the memory access is defined in a packed struct.
5751 @deftypefn {Target Hook} {enum machine_mode} TARGET_VECTORIZE_PREFERRED_SIMD_MODE (enum machine_mode @var{mode})
5752 This hook should return the preferred mode for vectorizing scalar
5753 mode @var{mode}. The default is
5754 equal to @code{word_mode}, because the vectorizer can do some
5755 transformations even in absence of specialized @acronym{SIMD} hardware.
5758 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void)
5759 This hook should return a mask of sizes that should be iterated over
5760 after trying to autovectorize using the vector size derived from the
5761 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5762 The default is zero which means to not iterate over other vector sizes.
5765 @deftypefn {Target Hook} {void *} TARGET_VECTORIZE_INIT_COST (struct loop *@var{loop_info})
5766 This hook should initialize target-specific data structures in preparation for modeling the costs of vectorizing a loop or basic block. The default allocates three unsigned integers for accumulating costs for the prologue, body, and epilogue of the loop or basic block. If @var{loop_info} is non-NULL, it identifies the loop being vectorized; otherwise a single block is being vectorized.
5769 @deftypefn {Target Hook} unsigned TARGET_VECTORIZE_ADD_STMT_COST (void *@var{data}, int @var{count}, enum vect_cost_for_stmt @var{kind}, struct _stmt_vec_info *@var{stmt_info}, int @var{misalign}, enum vect_cost_model_location @var{where})
5770 This hook should update the target-specific @var{data} in response to adding @var{count} copies of the given @var{kind} of statement to a loop or basic block. The default adds the builtin vectorizer cost for the copies of the statement to the accumulator specified by @var{where}, (the prologue, body, or epilogue) and returns the amount added. The return value should be viewed as a tentative cost that may later be revised.
5773 @deftypefn {Target Hook} void TARGET_VECTORIZE_FINISH_COST (void *@var{data}, unsigned *@var{prologue_cost}, unsigned *@var{body_cost}, unsigned *@var{epilogue_cost})
5774 This hook should complete calculations of the cost of vectorizing a loop or basic block based on @var{data}, and return the prologue, body, and epilogue costs as unsigned integers. The default returns the value of the three accumulators.
5777 @deftypefn {Target Hook} void TARGET_VECTORIZE_DESTROY_COST_DATA (void *@var{data})
5778 This hook should release @var{data} and any related data structures allocated by TARGET_VECTORIZE_INIT_COST. The default releases the accumulator.
5781 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_TM_LOAD (tree)
5782 This hook should return the built-in decl needed to load a vector of the given type within a transaction.
5785 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_TM_STORE (tree)
5786 This hook should return the built-in decl needed to store a vector of the given type within a transaction.
5789 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_GATHER (const_tree @var{mem_vectype}, const_tree @var{index_type}, int @var{scale})
5790 Target builtin that implements vector gather operation. @var{mem_vectype}
5791 is the vector type of the load and @var{index_type} is scalar type of
5792 the index, scaled by @var{scale}.
5793 The default is @code{NULL_TREE} which means to not vectorize gather
5797 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN (struct cgraph_node *@var{}, struct cgraph_simd_clone *@var{}, @var{tree}, @var{int})
5798 This hook should set @var{vecsize_mangle}, @var{vecsize_int}, @var{vecsize_float}
5799 fields in @var{simd_clone} structure pointed by @var{clone_info} argument and also
5800 @var{simdlen} field if it was previously 0.
5801 The hook should return 0 if SIMD clones shouldn't be emitted,
5802 or number of @var{vecsize_mangle} variants that should be emitted.
5805 @deftypefn {Target Hook} void TARGET_SIMD_CLONE_ADJUST (struct cgraph_node *@var{})
5806 This hook should add implicit @code{attribute(target("..."))} attribute
5807 to SIMD clone @var{node} if needed.
5810 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_USABLE (struct cgraph_node *@var{})
5811 This hook should return -1 if SIMD clone @var{node} shouldn't be used
5812 in vectorized loops in current function, or non-negative number if it is
5813 usable. In that case, the smaller the number is, the more desirable it is
5817 @node Anchored Addresses
5818 @section Anchored Addresses
5819 @cindex anchored addresses
5820 @cindex @option{-fsection-anchors}
5822 GCC usually addresses every static object as a separate entity.
5823 For example, if we have:
5827 int foo (void) @{ return a + b + c; @}
5830 the code for @code{foo} will usually calculate three separate symbolic
5831 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5832 it would be better to calculate just one symbolic address and access
5833 the three variables relative to it. The equivalent pseudocode would
5839 register int *xr = &x;
5840 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5844 (which isn't valid C). We refer to shared addresses like @code{x} as
5845 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5847 The hooks below describe the target properties that GCC needs to know
5848 in order to make effective use of section anchors. It won't use
5849 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5850 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5852 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5853 The minimum offset that should be applied to a section anchor.
5854 On most targets, it should be the smallest offset that can be
5855 applied to a base register while still giving a legitimate address
5856 for every mode. The default value is 0.
5859 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5860 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5861 offset that should be applied to section anchors. The default
5865 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5866 Write the assembly code to define section anchor @var{x}, which is a
5867 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5868 The hook is called with the assembly output position set to the beginning
5869 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5871 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5872 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5873 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5874 is @code{NULL}, which disables the use of section anchors altogether.
5877 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
5878 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5879 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5880 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5882 The default version is correct for most targets, but you might need to
5883 intercept this hook to handle things like target-specific attributes
5884 or target-specific sections.
5887 @node Condition Code
5888 @section Condition Code Status
5889 @cindex condition code status
5891 The macros in this section can be split in two families, according to the
5892 two ways of representing condition codes in GCC.
5894 The first representation is the so called @code{(cc0)} representation
5895 (@pxref{Jump Patterns}), where all instructions can have an implicit
5896 clobber of the condition codes. The second is the condition code
5897 register representation, which provides better schedulability for
5898 architectures that do have a condition code register, but on which
5899 most instructions do not affect it. The latter category includes
5902 The implicit clobbering poses a strong restriction on the placement of
5903 the definition and use of the condition code, which need to be in adjacent
5904 insns for machines using @code{(cc0)}. This can prevent important
5905 optimizations on some machines. For example, on the IBM RS/6000, there
5906 is a delay for taken branches unless the condition code register is set
5907 three instructions earlier than the conditional branch. The instruction
5908 scheduler cannot perform this optimization if it is not permitted to
5909 separate the definition and use of the condition code register.
5911 For this reason, it is possible and suggested to use a register to
5912 represent the condition code for new ports. If there is a specific
5913 condition code register in the machine, use a hard register. If the
5914 condition code or comparison result can be placed in any general register,
5915 or if there are multiple condition registers, use a pseudo register.
5916 Registers used to store the condition code value will usually have a mode
5917 that is in class @code{MODE_CC}.
5919 Alternatively, you can use @code{BImode} if the comparison operator is
5920 specified already in the compare instruction. In this case, you are not
5921 interested in most macros in this section.
5924 * CC0 Condition Codes:: Old style representation of condition codes.
5925 * MODE_CC Condition Codes:: Modern representation of condition codes.
5928 @node CC0 Condition Codes
5929 @subsection Representation of condition codes using @code{(cc0)}
5933 The file @file{conditions.h} defines a variable @code{cc_status} to
5934 describe how the condition code was computed (in case the interpretation of
5935 the condition code depends on the instruction that it was set by). This
5936 variable contains the RTL expressions on which the condition code is
5937 currently based, and several standard flags.
5939 Sometimes additional machine-specific flags must be defined in the machine
5940 description header file. It can also add additional machine-specific
5941 information by defining @code{CC_STATUS_MDEP}.
5943 @defmac CC_STATUS_MDEP
5944 C code for a data type which is used for declaring the @code{mdep}
5945 component of @code{cc_status}. It defaults to @code{int}.
5947 This macro is not used on machines that do not use @code{cc0}.
5950 @defmac CC_STATUS_MDEP_INIT
5951 A C expression to initialize the @code{mdep} field to ``empty''.
5952 The default definition does nothing, since most machines don't use
5953 the field anyway. If you want to use the field, you should probably
5954 define this macro to initialize it.
5956 This macro is not used on machines that do not use @code{cc0}.
5959 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5960 A C compound statement to set the components of @code{cc_status}
5961 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5962 this macro's responsibility to recognize insns that set the condition
5963 code as a byproduct of other activity as well as those that explicitly
5966 This macro is not used on machines that do not use @code{cc0}.
5968 If there are insns that do not set the condition code but do alter
5969 other machine registers, this macro must check to see whether they
5970 invalidate the expressions that the condition code is recorded as
5971 reflecting. For example, on the 68000, insns that store in address
5972 registers do not set the condition code, which means that usually
5973 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5974 insns. But suppose that the previous insn set the condition code
5975 based on location @samp{a4@@(102)} and the current insn stores a new
5976 value in @samp{a4}. Although the condition code is not changed by
5977 this, it will no longer be true that it reflects the contents of
5978 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5979 @code{cc_status} in this case to say that nothing is known about the
5980 condition code value.
5982 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5983 with the results of peephole optimization: insns whose patterns are
5984 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5985 constants which are just the operands. The RTL structure of these
5986 insns is not sufficient to indicate what the insns actually do. What
5987 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5988 @code{CC_STATUS_INIT}.
5990 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5991 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5992 @samp{cc}. This avoids having detailed information about patterns in
5993 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5996 @node MODE_CC Condition Codes
5997 @subsection Representation of condition codes using registers
6001 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
6002 On many machines, the condition code may be produced by other instructions
6003 than compares, for example the branch can use directly the condition
6004 code set by a subtract instruction. However, on some machines
6005 when the condition code is set this way some bits (such as the overflow
6006 bit) are not set in the same way as a test instruction, so that a different
6007 branch instruction must be used for some conditional branches. When
6008 this happens, use the machine mode of the condition code register to
6009 record different formats of the condition code register. Modes can
6010 also be used to record which compare instruction (e.g. a signed or an
6011 unsigned comparison) produced the condition codes.
6013 If other modes than @code{CCmode} are required, add them to
6014 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
6015 a mode given an operand of a compare. This is needed because the modes
6016 have to be chosen not only during RTL generation but also, for example,
6017 by instruction combination. The result of @code{SELECT_CC_MODE} should
6018 be consistent with the mode used in the patterns; for example to support
6019 the case of the add on the SPARC discussed above, we have the pattern
6023 [(set (reg:CC_NOOV 0)
6025 (plus:SI (match_operand:SI 0 "register_operand" "%r")
6026 (match_operand:SI 1 "arith_operand" "rI"))
6033 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
6034 for comparisons whose argument is a @code{plus}:
6037 #define SELECT_CC_MODE(OP,X,Y) \
6038 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
6039 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
6040 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
6041 || GET_CODE (X) == NEG) \
6042 ? CC_NOOVmode : CCmode))
6045 Another reason to use modes is to retain information on which operands
6046 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
6049 You should define this macro if and only if you define extra CC modes
6050 in @file{@var{machine}-modes.def}.
6053 @deftypefn {Target Hook} void TARGET_CANONICALIZE_COMPARISON (int *@var{code}, rtx *@var{op0}, rtx *@var{op1}, bool @var{op0_preserve_value})
6054 On some machines not all possible comparisons are defined, but you can
6055 convert an invalid comparison into a valid one. For example, the Alpha
6056 does not have a @code{GT} comparison, but you can use an @code{LT}
6057 comparison instead and swap the order of the operands.
6059 On such machines, implement this hook to do any required conversions.
6060 @var{code} is the initial comparison code and @var{op0} and @var{op1}
6061 are the left and right operands of the comparison, respectively. If
6062 @var{op0_preserve_value} is @code{true} the implementation is not
6063 allowed to change the value of @var{op0} since the value might be used
6064 in RTXs which aren't comparisons. E.g. the implementation is not
6065 allowed to swap operands in that case.
6067 GCC will not assume that the comparison resulting from this macro is
6068 valid but will see if the resulting insn matches a pattern in the
6071 You need not to implement this hook if it would never change the
6072 comparison code or operands.
6075 @defmac REVERSIBLE_CC_MODE (@var{mode})
6076 A C expression whose value is one if it is always safe to reverse a
6077 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6078 can ever return @var{mode} for a floating-point inequality comparison,
6079 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6081 You need not define this macro if it would always returns zero or if the
6082 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6083 For example, here is the definition used on the SPARC, where floating-point
6084 inequality comparisons are always given @code{CCFPEmode}:
6087 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
6091 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6092 A C expression whose value is reversed condition code of the @var{code} for
6093 comparison done in CC_MODE @var{mode}. The macro is used only in case
6094 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6095 machine has some non-standard way how to reverse certain conditionals. For
6096 instance in case all floating point conditions are non-trapping, compiler may
6097 freely convert unordered compares to ordered one. Then definition may look
6101 #define REVERSE_CONDITION(CODE, MODE) \
6102 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6103 : reverse_condition_maybe_unordered (CODE))
6107 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6108 On targets which do not use @code{(cc0)}, and which use a hard
6109 register rather than a pseudo-register to hold condition codes, the
6110 regular CSE passes are often not able to identify cases in which the
6111 hard register is set to a common value. Use this hook to enable a
6112 small pass which optimizes such cases. This hook should return true
6113 to enable this pass, and it should set the integers to which its
6114 arguments point to the hard register numbers used for condition codes.
6115 When there is only one such register, as is true on most systems, the
6116 integer pointed to by @var{p2} should be set to
6117 @code{INVALID_REGNUM}.
6119 The default version of this hook returns false.
6122 @deftypefn {Target Hook} {enum machine_mode} TARGET_CC_MODES_COMPATIBLE (enum machine_mode @var{m1}, enum machine_mode @var{m2})
6123 On targets which use multiple condition code modes in class
6124 @code{MODE_CC}, it is sometimes the case that a comparison can be
6125 validly done in more than one mode. On such a system, define this
6126 target hook to take two mode arguments and to return a mode in which
6127 both comparisons may be validly done. If there is no such mode,
6128 return @code{VOIDmode}.
6130 The default version of this hook checks whether the modes are the
6131 same. If they are, it returns that mode. If they are different, it
6132 returns @code{VOIDmode}.
6136 @section Describing Relative Costs of Operations
6137 @cindex costs of instructions
6138 @cindex relative costs
6139 @cindex speed of instructions
6141 These macros let you describe the relative speed of various operations
6142 on the target machine.
6144 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6145 A C expression for the cost of moving data of mode @var{mode} from a
6146 register in class @var{from} to one in class @var{to}. The classes are
6147 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6148 value of 2 is the default; other values are interpreted relative to
6151 It is not required that the cost always equal 2 when @var{from} is the
6152 same as @var{to}; on some machines it is expensive to move between
6153 registers if they are not general registers.
6155 If reload sees an insn consisting of a single @code{set} between two
6156 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6157 classes returns a value of 2, reload does not check to ensure that the
6158 constraints of the insn are met. Setting a cost of other than 2 will
6159 allow reload to verify that the constraints are met. You should do this
6160 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6162 These macros are obsolete, new ports should use the target hook
6163 @code{TARGET_REGISTER_MOVE_COST} instead.
6166 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6167 This target hook should return the cost of moving data of mode @var{mode}
6168 from a register in class @var{from} to one in class @var{to}. The classes
6169 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6170 A value of 2 is the default; other values are interpreted relative to
6173 It is not required that the cost always equal 2 when @var{from} is the
6174 same as @var{to}; on some machines it is expensive to move between
6175 registers if they are not general registers.
6177 If reload sees an insn consisting of a single @code{set} between two
6178 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6179 classes returns a value of 2, reload does not check to ensure that the
6180 constraints of the insn are met. Setting a cost of other than 2 will
6181 allow reload to verify that the constraints are met. You should do this
6182 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6184 The default version of this function returns 2.
6187 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6188 A C expression for the cost of moving data of mode @var{mode} between a
6189 register of class @var{class} and memory; @var{in} is zero if the value
6190 is to be written to memory, nonzero if it is to be read in. This cost
6191 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6192 registers and memory is more expensive than between two registers, you
6193 should define this macro to express the relative cost.
6195 If you do not define this macro, GCC uses a default cost of 4 plus
6196 the cost of copying via a secondary reload register, if one is
6197 needed. If your machine requires a secondary reload register to copy
6198 between memory and a register of @var{class} but the reload mechanism is
6199 more complex than copying via an intermediate, define this macro to
6200 reflect the actual cost of the move.
6202 GCC defines the function @code{memory_move_secondary_cost} if
6203 secondary reloads are needed. It computes the costs due to copying via
6204 a secondary register. If your machine copies from memory using a
6205 secondary register in the conventional way but the default base value of
6206 4 is not correct for your machine, define this macro to add some other
6207 value to the result of that function. The arguments to that function
6208 are the same as to this macro.
6210 These macros are obsolete, new ports should use the target hook
6211 @code{TARGET_MEMORY_MOVE_COST} instead.
6214 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6215 This target hook should return the cost of moving data of mode @var{mode}
6216 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6217 if the value is to be written to memory, @code{true} if it is to be read in.
6218 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6219 If moving between registers and memory is more expensive than between two
6220 registers, you should add this target hook to express the relative cost.
6222 If you do not add this target hook, GCC uses a default cost of 4 plus
6223 the cost of copying via a secondary reload register, if one is
6224 needed. If your machine requires a secondary reload register to copy
6225 between memory and a register of @var{rclass} but the reload mechanism is
6226 more complex than copying via an intermediate, use this target hook to
6227 reflect the actual cost of the move.
6229 GCC defines the function @code{memory_move_secondary_cost} if
6230 secondary reloads are needed. It computes the costs due to copying via
6231 a secondary register. If your machine copies from memory using a
6232 secondary register in the conventional way but the default base value of
6233 4 is not correct for your machine, use this target hook to add some other
6234 value to the result of that function. The arguments to that function
6235 are the same as to this target hook.
6238 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6239 A C expression for the cost of a branch instruction. A value of 1 is
6240 the default; other values are interpreted relative to that. Parameter
6241 @var{speed_p} is true when the branch in question should be optimized
6242 for speed. When it is false, @code{BRANCH_COST} should return a value
6243 optimal for code size rather than performance. @var{predictable_p} is
6244 true for well-predicted branches. On many architectures the
6245 @code{BRANCH_COST} can be reduced then.
6248 Here are additional macros which do not specify precise relative costs,
6249 but only that certain actions are more expensive than GCC would
6252 @defmac SLOW_BYTE_ACCESS
6253 Define this macro as a C expression which is nonzero if accessing less
6254 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6255 faster than accessing a word of memory, i.e., if such access
6256 require more than one instruction or if there is no difference in cost
6257 between byte and (aligned) word loads.
6259 When this macro is not defined, the compiler will access a field by
6260 finding the smallest containing object; when it is defined, a fullword
6261 load will be used if alignment permits. Unless bytes accesses are
6262 faster than word accesses, using word accesses is preferable since it
6263 may eliminate subsequent memory access if subsequent accesses occur to
6264 other fields in the same word of the structure, but to different bytes.
6267 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6268 Define this macro to be the value 1 if memory accesses described by the
6269 @var{mode} and @var{alignment} parameters have a cost many times greater
6270 than aligned accesses, for example if they are emulated in a trap
6273 When this macro is nonzero, the compiler will act as if
6274 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6275 moves. This can cause significantly more instructions to be produced.
6276 Therefore, do not set this macro nonzero if unaligned accesses only add a
6277 cycle or two to the time for a memory access.
6279 If the value of this macro is always zero, it need not be defined. If
6280 this macro is defined, it should produce a nonzero value when
6281 @code{STRICT_ALIGNMENT} is nonzero.
6284 @defmac MOVE_RATIO (@var{speed})
6285 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6286 which a sequence of insns should be generated instead of a
6287 string move insn or a library call. Increasing the value will always
6288 make code faster, but eventually incurs high cost in increased code size.
6290 Note that on machines where the corresponding move insn is a
6291 @code{define_expand} that emits a sequence of insns, this macro counts
6292 the number of such sequences.
6294 The parameter @var{speed} is true if the code is currently being
6295 optimized for speed rather than size.
6297 If you don't define this, a reasonable default is used.
6300 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6301 A C expression used to determine whether @code{move_by_pieces} will be used to
6302 copy a chunk of memory, or whether some other block move mechanism
6303 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6304 than @code{MOVE_RATIO}.
6307 @defmac MOVE_MAX_PIECES
6308 A C expression used by @code{move_by_pieces} to determine the largest unit
6309 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6312 @defmac CLEAR_RATIO (@var{speed})
6313 The threshold of number of scalar move insns, @emph{below} which a sequence
6314 of insns should be generated to clear memory instead of a string clear insn
6315 or a library call. Increasing the value will always make code faster, but
6316 eventually incurs high cost in increased code size.
6318 The parameter @var{speed} is true if the code is currently being
6319 optimized for speed rather than size.
6321 If you don't define this, a reasonable default is used.
6324 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6325 A C expression used to determine whether @code{clear_by_pieces} will be used
6326 to clear a chunk of memory, or whether some other block clear mechanism
6327 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6328 than @code{CLEAR_RATIO}.
6331 @defmac SET_RATIO (@var{speed})
6332 The threshold of number of scalar move insns, @emph{below} which a sequence
6333 of insns should be generated to set memory to a constant value, instead of
6334 a block set insn or a library call.
6335 Increasing the value will always make code faster, but
6336 eventually incurs high cost in increased code size.
6338 The parameter @var{speed} is true if the code is currently being
6339 optimized for speed rather than size.
6341 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6344 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6345 A C expression used to determine whether @code{store_by_pieces} will be
6346 used to set a chunk of memory to a constant value, or whether some
6347 other mechanism will be used. Used by @code{__builtin_memset} when
6348 storing values other than constant zero.
6349 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6350 than @code{SET_RATIO}.
6353 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6354 A C expression used to determine whether @code{store_by_pieces} will be
6355 used to set a chunk of memory to a constant string value, or whether some
6356 other mechanism will be used. Used by @code{__builtin_strcpy} when
6357 called with a constant source string.
6358 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6359 than @code{MOVE_RATIO}.
6362 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6363 A C expression used to determine whether a load postincrement is a good
6364 thing to use for a given mode. Defaults to the value of
6365 @code{HAVE_POST_INCREMENT}.
6368 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6369 A C expression used to determine whether a load postdecrement is a good
6370 thing to use for a given mode. Defaults to the value of
6371 @code{HAVE_POST_DECREMENT}.
6374 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6375 A C expression used to determine whether a load preincrement is a good
6376 thing to use for a given mode. Defaults to the value of
6377 @code{HAVE_PRE_INCREMENT}.
6380 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6381 A C expression used to determine whether a load predecrement is a good
6382 thing to use for a given mode. Defaults to the value of
6383 @code{HAVE_PRE_DECREMENT}.
6386 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6387 A C expression used to determine whether a store postincrement is a good
6388 thing to use for a given mode. Defaults to the value of
6389 @code{HAVE_POST_INCREMENT}.
6392 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6393 A C expression used to determine whether a store postdecrement is a good
6394 thing to use for a given mode. Defaults to the value of
6395 @code{HAVE_POST_DECREMENT}.
6398 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6399 This macro is used to determine whether a store preincrement is a good
6400 thing to use for a given mode. Defaults to the value of
6401 @code{HAVE_PRE_INCREMENT}.
6404 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6405 This macro is used to determine whether a store predecrement is a good
6406 thing to use for a given mode. Defaults to the value of
6407 @code{HAVE_PRE_DECREMENT}.
6410 @defmac NO_FUNCTION_CSE
6411 Define this macro if it is as good or better to call a constant
6412 function address than to call an address kept in a register.
6415 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
6416 Define this macro if a non-short-circuit operation produced by
6417 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6418 @code{BRANCH_COST} is greater than or equal to the value 2.
6421 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int @var{opno}, int *@var{total}, bool @var{speed})
6422 This target hook describes the relative costs of RTL expressions.
6424 The cost may depend on the precise form of the expression, which is
6425 available for examination in @var{x}, and the fact that @var{x} appears
6426 as operand @var{opno} of an expression with rtx code @var{outer_code}.
6427 That is, the hook can assume that there is some rtx @var{y} such
6428 that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that
6429 either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or
6430 (b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}.
6432 @var{code} is @var{x}'s expression code---redundant, since it can be
6433 obtained with @code{GET_CODE (@var{x})}.
6435 In implementing this hook, you can use the construct
6436 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6439 On entry to the hook, @code{*@var{total}} contains a default estimate
6440 for the cost of the expression. The hook should modify this value as
6441 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6442 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6443 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6445 When optimizing for code size, i.e.@: when @code{speed} is
6446 false, this target hook should be used to estimate the relative
6447 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6449 The hook returns true when all subexpressions of @var{x} have been
6450 processed, and false when @code{rtx_cost} should recurse.
6453 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, enum machine_mode @var{mode}, addr_space_t @var{as}, bool @var{speed})
6454 This hook computes the cost of an addressing mode that contains
6455 @var{address}. If not defined, the cost is computed from
6456 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6458 For most CISC machines, the default cost is a good approximation of the
6459 true cost of the addressing mode. However, on RISC machines, all
6460 instructions normally have the same length and execution time. Hence
6461 all addresses will have equal costs.
6463 In cases where more than one form of an address is known, the form with
6464 the lowest cost will be used. If multiple forms have the same, lowest,
6465 cost, the one that is the most complex will be used.
6467 For example, suppose an address that is equal to the sum of a register
6468 and a constant is used twice in the same basic block. When this macro
6469 is not defined, the address will be computed in a register and memory
6470 references will be indirect through that register. On machines where
6471 the cost of the addressing mode containing the sum is no higher than
6472 that of a simple indirect reference, this will produce an additional
6473 instruction and possibly require an additional register. Proper
6474 specification of this macro eliminates this overhead for such machines.
6476 This hook is never called with an invalid address.
6478 On machines where an address involving more than one register is as
6479 cheap as an address computation involving only one register, defining
6480 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6481 be live over a region of code where only one would have been if
6482 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6483 should be considered in the definition of this macro. Equivalent costs
6484 should probably only be given to addresses with different numbers of
6485 registers on machines with lots of registers.
6489 @section Adjusting the Instruction Scheduler
6491 The instruction scheduler may need a fair amount of machine-specific
6492 adjustment in order to produce good code. GCC provides several target
6493 hooks for this purpose. It is usually enough to define just a few of
6494 them: try the first ones in this list first.
6496 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6497 This hook returns the maximum number of instructions that can ever
6498 issue at the same time on the target machine. The default is one.
6499 Although the insn scheduler can define itself the possibility of issue
6500 an insn on the same cycle, the value can serve as an additional
6501 constraint to issue insns on the same simulated processor cycle (see
6502 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6503 This value must be constant over the entire compilation. If you need
6504 it to vary depending on what the instructions are, you must use
6505 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6508 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6509 This hook is executed by the scheduler after it has scheduled an insn
6510 from the ready list. It should return the number of insns which can
6511 still be issued in the current cycle. The default is
6512 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6513 @code{USE}, which normally are not counted against the issue rate.
6514 You should define this hook if some insns take more machine resources
6515 than others, so that fewer insns can follow them in the same cycle.
6516 @var{file} is either a null pointer, or a stdio stream to write any
6517 debug output to. @var{verbose} is the verbose level provided by
6518 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6522 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6523 This function corrects the value of @var{cost} based on the
6524 relationship between @var{insn} and @var{dep_insn} through the
6525 dependence @var{link}. It should return the new value. The default
6526 is to make no adjustment to @var{cost}. This can be used for example
6527 to specify to the scheduler using the traditional pipeline description
6528 that an output- or anti-dependence does not incur the same cost as a
6529 data-dependence. If the scheduler using the automaton based pipeline
6530 description, the cost of anti-dependence is zero and the cost of
6531 output-dependence is maximum of one and the difference of latency
6532 times of the first and the second insns. If these values are not
6533 acceptable, you could use the hook to modify them too. See also
6534 @pxref{Processor pipeline description}.
6537 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6538 This hook adjusts the integer scheduling priority @var{priority} of
6539 @var{insn}. It should return the new priority. Increase the priority to
6540 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6541 later. Do not define this hook if you do not need to adjust the
6542 scheduling priorities of insns.
6545 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6546 This hook is executed by the scheduler after it has scheduled the ready
6547 list, to allow the machine description to reorder it (for example to
6548 combine two small instructions together on @samp{VLIW} machines).
6549 @var{file} is either a null pointer, or a stdio stream to write any
6550 debug output to. @var{verbose} is the verbose level provided by
6551 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6552 list of instructions that are ready to be scheduled. @var{n_readyp} is
6553 a pointer to the number of elements in the ready list. The scheduler
6554 reads the ready list in reverse order, starting with
6555 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6556 is the timer tick of the scheduler. You may modify the ready list and
6557 the number of ready insns. The return value is the number of insns that
6558 can issue this cycle; normally this is just @code{issue_rate}. See also
6559 @samp{TARGET_SCHED_REORDER2}.
6562 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6563 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6564 function is called whenever the scheduler starts a new cycle. This one
6565 is called once per iteration over a cycle, immediately after
6566 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6567 return the number of insns to be scheduled in the same cycle. Defining
6568 this hook can be useful if there are frequent situations where
6569 scheduling one insn causes other insns to become ready in the same
6570 cycle. These other insns can then be taken into account properly.
6573 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_P (void)
6574 This hook is used to check whether target platform supports macro fusion.
6577 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_PAIR_P (rtx @var{condgen}, rtx @var{condjmp})
6578 This hook is used to check whether two insns could be macro fused for
6579 target microarchitecture. If this hook returns true for the given insn pair
6580 (@var{condgen} and @var{condjmp}), scheduler will put them into a sched
6581 group, and they will not be scheduled apart.
6584 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6585 This hook is called after evaluation forward dependencies of insns in
6586 chain given by two parameter values (@var{head} and @var{tail}
6587 correspondingly) but before insns scheduling of the insn chain. For
6588 example, it can be used for better insn classification if it requires
6589 analysis of dependencies. This hook can use backward and forward
6590 dependencies of the insn scheduler because they are already
6594 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6595 This hook is executed by the scheduler at the beginning of each block of
6596 instructions that are to be scheduled. @var{file} is either a null
6597 pointer, or a stdio stream to write any debug output to. @var{verbose}
6598 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6599 @var{max_ready} is the maximum number of insns in the current scheduling
6600 region that can be live at the same time. This can be used to allocate
6601 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6604 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6605 This hook is executed by the scheduler at the end of each block of
6606 instructions that are to be scheduled. It can be used to perform
6607 cleanup of any actions done by the other scheduling hooks. @var{file}
6608 is either a null pointer, or a stdio stream to write any debug output
6609 to. @var{verbose} is the verbose level provided by
6610 @option{-fsched-verbose-@var{n}}.
6613 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6614 This hook is executed by the scheduler after function level initializations.
6615 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6616 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6617 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6620 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6621 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6622 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6623 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6626 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6627 The hook returns an RTL insn. The automaton state used in the
6628 pipeline hazard recognizer is changed as if the insn were scheduled
6629 when the new simulated processor cycle starts. Usage of the hook may
6630 simplify the automaton pipeline description for some @acronym{VLIW}
6631 processors. If the hook is defined, it is used only for the automaton
6632 based pipeline description. The default is not to change the state
6633 when the new simulated processor cycle starts.
6636 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6637 The hook can be used to initialize data used by the previous hook.
6640 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6641 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6642 to changed the state as if the insn were scheduled when the new
6643 simulated processor cycle finishes.
6646 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6647 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6648 used to initialize data used by the previous hook.
6651 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
6652 The hook to notify target that the current simulated cycle is about to finish.
6653 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6654 to change the state in more complicated situations - e.g., when advancing
6655 state on a single insn is not enough.
6658 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
6659 The hook to notify target that new simulated cycle has just started.
6660 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6661 to change the state in more complicated situations - e.g., when advancing
6662 state on a single insn is not enough.
6665 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6666 This hook controls better choosing an insn from the ready insn queue
6667 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6668 chooses the first insn from the queue. If the hook returns a positive
6669 value, an additional scheduler code tries all permutations of
6670 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6671 subsequent ready insns to choose an insn whose issue will result in
6672 maximal number of issued insns on the same cycle. For the
6673 @acronym{VLIW} processor, the code could actually solve the problem of
6674 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6675 rules of @acronym{VLIW} packing are described in the automaton.
6677 This code also could be used for superscalar @acronym{RISC}
6678 processors. Let us consider a superscalar @acronym{RISC} processor
6679 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6680 @var{B}, some insns can be executed only in pipelines @var{B} or
6681 @var{C}, and one insn can be executed in pipeline @var{B}. The
6682 processor may issue the 1st insn into @var{A} and the 2nd one into
6683 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6684 until the next cycle. If the scheduler issues the 3rd insn the first,
6685 the processor could issue all 3 insns per cycle.
6687 Actually this code demonstrates advantages of the automaton based
6688 pipeline hazard recognizer. We try quickly and easy many insn
6689 schedules to choose the best one.
6691 The default is no multipass scheduling.
6694 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx @var{insn})
6696 This hook controls what insns from the ready insn queue will be
6697 considered for the multipass insn scheduling. If the hook returns
6698 zero for @var{insn}, the insn will be not chosen to
6701 The default is that any ready insns can be chosen to be issued.
6704 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN (void *@var{data}, char *@var{ready_try}, int @var{n_ready}, bool @var{first_cycle_insn_p})
6705 This hook prepares the target backend for a new round of multipass
6709 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE (void *@var{data}, char *@var{ready_try}, int @var{n_ready}, rtx @var{insn}, const void *@var{prev_data})
6710 This hook is called when multipass scheduling evaluates instruction INSN.
6713 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, char *@var{ready_try}, int @var{n_ready})
6714 This is called when multipass scheduling backtracks from evaluation of
6718 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
6719 This hook notifies the target about the result of the concluded current
6720 round of multipass scheduling.
6723 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
6724 This hook initializes target-specific data used in multipass scheduling.
6727 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
6728 This hook finalizes target-specific data used in multipass scheduling.
6731 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *@var{dump}, int @var{verbose}, rtx @var{insn}, int @var{last_clock}, int @var{clock}, int *@var{sort_p})
6732 This hook is called by the insn scheduler before issuing @var{insn}
6733 on cycle @var{clock}. If the hook returns nonzero,
6734 @var{insn} is not issued on this processor cycle. Instead,
6735 the processor cycle is advanced. If *@var{sort_p}
6736 is zero, the insn ready queue is not sorted on the new cycle
6737 start as usually. @var{dump} and @var{verbose} specify the file and
6738 verbosity level to use for debugging output.
6739 @var{last_clock} and @var{clock} are, respectively, the
6740 processor cycle on which the previous insn has been issued,
6741 and the current processor cycle.
6744 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
6745 This hook is used to define which dependences are considered costly by
6746 the target, so costly that it is not advisable to schedule the insns that
6747 are involved in the dependence too close to one another. The parameters
6748 to this hook are as follows: The first parameter @var{_dep} is the dependence
6749 being evaluated. The second parameter @var{cost} is the cost of the
6750 dependence as estimated by the scheduler, and the third
6751 parameter @var{distance} is the distance in cycles between the two insns.
6752 The hook returns @code{true} if considering the distance between the two
6753 insns the dependence between them is considered costly by the target,
6754 and @code{false} otherwise.
6756 Defining this hook can be useful in multiple-issue out-of-order machines,
6757 where (a) it's practically hopeless to predict the actual data/resource
6758 delays, however: (b) there's a better chance to predict the actual grouping
6759 that will be formed, and (c) correctly emulating the grouping can be very
6760 important. In such targets one may want to allow issuing dependent insns
6761 closer to one another---i.e., closer than the dependence distance; however,
6762 not in cases of ``costly dependences'', which this hooks allows to define.
6765 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6766 This hook is called by the insn scheduler after emitting a new instruction to
6767 the instruction stream. The hook notifies a target backend to extend its
6768 per instruction data structures.
6771 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6772 Return a pointer to a store large enough to hold target scheduling context.
6775 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6776 Initialize store pointed to by @var{tc} to hold target scheduling context.
6777 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6778 beginning of the block. Otherwise, copy the current context into @var{tc}.
6781 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6782 Copy target scheduling context pointed to by @var{tc} to the current context.
6785 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6786 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6789 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6790 Deallocate a store for target scheduling context pointed to by @var{tc}.
6793 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, unsigned int @var{dep_status}, rtx *@var{new_pat})
6794 This hook is called by the insn scheduler when @var{insn} has only
6795 speculative dependencies and therefore can be scheduled speculatively.
6796 The hook is used to check if the pattern of @var{insn} has a speculative
6797 version and, in case of successful check, to generate that speculative
6798 pattern. The hook should return 1, if the instruction has a speculative form,
6799 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6800 speculation. If the return value equals 1 then @var{new_pat} is assigned
6801 the generated speculative pattern.
6804 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (unsigned int @var{dep_status})
6805 This hook is called by the insn scheduler during generation of recovery code
6806 for @var{insn}. It should return @code{true}, if the corresponding check
6807 instruction should branch to recovery code, or @code{false} otherwise.
6810 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx @var{insn}, rtx @var{label}, unsigned int @var{ds})
6811 This hook is called by the insn scheduler to generate a pattern for recovery
6812 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6813 speculative instruction for which the check should be generated.
6814 @var{label} is either a label of a basic block, where recovery code should
6815 be emitted, or a null pointer, when requested check doesn't branch to
6816 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6817 a pattern for a branchy check corresponding to a simple check denoted by
6818 @var{insn} should be generated. In this case @var{label} can't be null.
6821 @deftypefn {Target Hook} bool TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (const_rtx @var{insn})
6822 This hook is used as a workaround for
6823 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6824 called on the first instruction of the ready list. The hook is used to
6825 discard speculative instructions that stand first in the ready list from
6826 being scheduled on the current cycle. If the hook returns @code{false},
6827 @var{insn} will not be chosen to be issued.
6828 For non-speculative instructions,
6829 the hook should always return @code{true}. For example, in the ia64 backend
6830 the hook is used to cancel data speculative insns when the ALAT table
6834 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
6835 This hook is used by the insn scheduler to find out what features should be
6837 The structure *@var{spec_info} should be filled in by the target.
6838 The structure describes speculation types that can be used in the scheduler.
6841 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6842 This hook is called by the swing modulo scheduler to calculate a
6843 resource-based lower bound which is based on the resources available in
6844 the machine and the resources required by each instruction. The target
6845 backend can use @var{g} to calculate such bound. A very simple lower
6846 bound will be used in case this hook is not implemented: the total number
6847 of instructions divided by the issue rate.
6850 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx @var{insn}, int @var{x})
6851 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6852 is supported in hardware and the condition specified in the parameter is true.
6855 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx @var{insn}, int @var{x})
6856 This hook is called by Haifa Scheduler. It performs the operation specified
6857 in its second parameter.
6860 @deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE
6861 True if the processor has an exposed pipeline, which means that not just
6862 the order of instructions is important for correctness when scheduling, but
6863 also the latencies of operations.
6866 @deftypefn {Target Hook} int TARGET_SCHED_REASSOCIATION_WIDTH (unsigned int @var{opc}, enum machine_mode @var{mode})
6867 This hook is called by tree reassociator to determine a level of
6868 parallelism required in output calculations chain.
6872 @section Dividing the Output into Sections (Texts, Data, @dots{})
6873 @c the above section title is WAY too long. maybe cut the part between
6874 @c the (...)? --mew 10feb93
6876 An object file is divided into sections containing different types of
6877 data. In the most common case, there are three sections: the @dfn{text
6878 section}, which holds instructions and read-only data; the @dfn{data
6879 section}, which holds initialized writable data; and the @dfn{bss
6880 section}, which holds uninitialized data. Some systems have other kinds
6883 @file{varasm.c} provides several well-known sections, such as
6884 @code{text_section}, @code{data_section} and @code{bss_section}.
6885 The normal way of controlling a @code{@var{foo}_section} variable
6886 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6887 as described below. The macros are only read once, when @file{varasm.c}
6888 initializes itself, so their values must be run-time constants.
6889 They may however depend on command-line flags.
6891 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6892 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6893 to be string literals.
6895 Some assemblers require a different string to be written every time a
6896 section is selected. If your assembler falls into this category, you
6897 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6898 @code{get_unnamed_section} to set up the sections.
6900 You must always create a @code{text_section}, either by defining
6901 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6902 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6903 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6904 create a distinct @code{readonly_data_section}, the default is to
6905 reuse @code{text_section}.
6907 All the other @file{varasm.c} sections are optional, and are null
6908 if the target does not provide them.
6910 @defmac TEXT_SECTION_ASM_OP
6911 A C expression whose value is a string, including spacing, containing the
6912 assembler operation that should precede instructions and read-only data.
6913 Normally @code{"\t.text"} is right.
6916 @defmac HOT_TEXT_SECTION_NAME
6917 If defined, a C string constant for the name of the section containing most
6918 frequently executed functions of the program. If not defined, GCC will provide
6919 a default definition if the target supports named sections.
6922 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6923 If defined, a C string constant for the name of the section containing unlikely
6924 executed functions in the program.
6927 @defmac DATA_SECTION_ASM_OP
6928 A C expression whose value is a string, including spacing, containing the
6929 assembler operation to identify the following data as writable initialized
6930 data. Normally @code{"\t.data"} is right.
6933 @defmac SDATA_SECTION_ASM_OP
6934 If defined, a C expression whose value is a string, including spacing,
6935 containing the assembler operation to identify the following data as
6936 initialized, writable small data.
6939 @defmac READONLY_DATA_SECTION_ASM_OP
6940 A C expression whose value is a string, including spacing, containing the
6941 assembler operation to identify the following data as read-only initialized
6945 @defmac BSS_SECTION_ASM_OP
6946 If defined, a C expression whose value is a string, including spacing,
6947 containing the assembler operation to identify the following data as
6948 uninitialized global data. If not defined, and
6949 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
6950 uninitialized global data will be output in the data section if
6951 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6955 @defmac SBSS_SECTION_ASM_OP
6956 If defined, a C expression whose value is a string, including spacing,
6957 containing the assembler operation to identify the following data as
6958 uninitialized, writable small data.
6961 @defmac TLS_COMMON_ASM_OP
6962 If defined, a C expression whose value is a string containing the
6963 assembler operation to identify the following data as thread-local
6964 common data. The default is @code{".tls_common"}.
6967 @defmac TLS_SECTION_ASM_FLAG
6968 If defined, a C expression whose value is a character constant
6969 containing the flag used to mark a section as a TLS section. The
6970 default is @code{'T'}.
6973 @defmac INIT_SECTION_ASM_OP
6974 If defined, a C expression whose value is a string, including spacing,
6975 containing the assembler operation to identify the following data as
6976 initialization code. If not defined, GCC will assume such a section does
6977 not exist. This section has no corresponding @code{init_section}
6978 variable; it is used entirely in runtime code.
6981 @defmac FINI_SECTION_ASM_OP
6982 If defined, a C expression whose value is a string, including spacing,
6983 containing the assembler operation to identify the following data as
6984 finalization code. If not defined, GCC will assume such a section does
6985 not exist. This section has no corresponding @code{fini_section}
6986 variable; it is used entirely in runtime code.
6989 @defmac INIT_ARRAY_SECTION_ASM_OP
6990 If defined, a C expression whose value is a string, including spacing,
6991 containing the assembler operation to identify the following data as
6992 part of the @code{.init_array} (or equivalent) section. If not
6993 defined, GCC will assume such a section does not exist. Do not define
6994 both this macro and @code{INIT_SECTION_ASM_OP}.
6997 @defmac FINI_ARRAY_SECTION_ASM_OP
6998 If defined, a C expression whose value is a string, including spacing,
6999 containing the assembler operation to identify the following data as
7000 part of the @code{.fini_array} (or equivalent) section. If not
7001 defined, GCC will assume such a section does not exist. Do not define
7002 both this macro and @code{FINI_SECTION_ASM_OP}.
7005 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
7006 If defined, an ASM statement that switches to a different section
7007 via @var{section_op}, calls @var{function}, and switches back to
7008 the text section. This is used in @file{crtstuff.c} if
7009 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
7010 to initialization and finalization functions from the init and fini
7011 sections. By default, this macro uses a simple function call. Some
7012 ports need hand-crafted assembly code to avoid dependencies on
7013 registers initialized in the function prologue or to ensure that
7014 constant pools don't end up too far way in the text section.
7017 @defmac TARGET_LIBGCC_SDATA_SECTION
7018 If defined, a string which names the section into which small
7019 variables defined in crtstuff and libgcc should go. This is useful
7020 when the target has options for optimizing access to small data, and
7021 you want the crtstuff and libgcc routines to be conservative in what
7022 they expect of your application yet liberal in what your application
7023 expects. For example, for targets with a @code{.sdata} section (like
7024 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
7025 require small data support from your application, but use this macro
7026 to put small data into @code{.sdata} so that your application can
7027 access these variables whether it uses small data or not.
7030 @defmac FORCE_CODE_SECTION_ALIGN
7031 If defined, an ASM statement that aligns a code section to some
7032 arbitrary boundary. This is used to force all fragments of the
7033 @code{.init} and @code{.fini} sections to have to same alignment
7034 and thus prevent the linker from having to add any padding.
7037 @defmac JUMP_TABLES_IN_TEXT_SECTION
7038 Define this macro to be an expression with a nonzero value if jump
7039 tables (for @code{tablejump} insns) should be output in the text
7040 section, along with the assembler instructions. Otherwise, the
7041 readonly data section is used.
7043 This macro is irrelevant if there is no separate readonly data section.
7046 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
7047 Define this hook if you need to do something special to set up the
7048 @file{varasm.c} sections, or if your target has some special sections
7049 of its own that you need to create.
7051 GCC calls this hook after processing the command line, but before writing
7052 any assembly code, and before calling any of the section-returning hooks
7056 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
7057 Return a mask describing how relocations should be treated when
7058 selecting sections. Bit 1 should be set if global relocations
7059 should be placed in a read-write section; bit 0 should be set if
7060 local relocations should be placed in a read-write section.
7062 The default version of this function returns 3 when @option{-fpic}
7063 is in effect, and 0 otherwise. The hook is typically redefined
7064 when the target cannot support (some kinds of) dynamic relocations
7065 in read-only sections even in executables.
7068 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
7069 Return the section into which @var{exp} should be placed. You can
7070 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
7071 some sort. @var{reloc} indicates whether the initial value of @var{exp}
7072 requires link-time relocations. Bit 0 is set when variable contains
7073 local relocations only, while bit 1 is set for global relocations.
7074 @var{align} is the constant alignment in bits.
7076 The default version of this function takes care of putting read-only
7077 variables in @code{readonly_data_section}.
7079 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7082 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7083 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7084 for @code{FUNCTION_DECL}s as well as for variables and constants.
7086 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7087 function has been determined to be likely to be called, and nonzero if
7088 it is unlikely to be called.
7091 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7092 Build up a unique section name, expressed as a @code{STRING_CST} node,
7093 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7094 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7095 the initial value of @var{exp} requires link-time relocations.
7097 The default version of this function appends the symbol name to the
7098 ELF section name that would normally be used for the symbol. For
7099 example, the function @code{foo} would be placed in @code{.text.foo}.
7100 Whatever the actual target object format, this is often good enough.
7103 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7104 Return the readonly data section associated with
7105 @samp{DECL_SECTION_NAME (@var{decl})}.
7106 The default version of this function selects @code{.gnu.linkonce.r.name} if
7107 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7108 if function is in @code{.text.name}, and the normal readonly-data section
7112 @deftypevr {Target Hook} {const char *} TARGET_ASM_MERGEABLE_RODATA_PREFIX
7113 Usually, the compiler uses the prefix @code{".rodata"} to construct
7114 section names for mergeable constant data. Define this macro to override
7115 the string if a different section name should be used.
7118 @deftypefn {Target Hook} {section *} TARGET_ASM_TM_CLONE_TABLE_SECTION (void)
7119 Return the section that should be used for transactional memory clone tables.
7122 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7123 Return the section into which a constant @var{x}, of mode @var{mode},
7124 should be placed. You can assume that @var{x} is some kind of
7125 constant in RTL@. The argument @var{mode} is redundant except in the
7126 case of a @code{const_int} rtx. @var{align} is the constant alignment
7129 The default version of this function takes care of putting symbolic
7130 constants in @code{flag_pic} mode in @code{data_section} and everything
7131 else in @code{readonly_data_section}.
7134 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7135 Define this hook if you need to postprocess the assembler name generated
7136 by target-independent code. The @var{id} provided to this hook will be
7137 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7138 or the mangled name of the @var{decl} in C++). The return value of the
7139 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7140 your target system. The default implementation of this hook just
7141 returns the @var{id} provided.
7144 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7145 Define this hook if references to a symbol or a constant must be
7146 treated differently depending on something about the variable or
7147 function named by the symbol (such as what section it is in).
7149 The hook is executed immediately after rtl has been created for
7150 @var{decl}, which may be a variable or function declaration or
7151 an entry in the constant pool. In either case, @var{rtl} is the
7152 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7153 in this hook; that field may not have been initialized yet.
7155 In the case of a constant, it is safe to assume that the rtl is
7156 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7157 will also have this form, but that is not guaranteed. Global
7158 register variables, for instance, will have a @code{reg} for their
7159 rtl. (Normally the right thing to do with such unusual rtl is
7162 The @var{new_decl_p} argument will be true if this is the first time
7163 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7164 be false for subsequent invocations, which will happen for duplicate
7165 declarations. Whether or not anything must be done for the duplicate
7166 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7167 @var{new_decl_p} is always true when the hook is called for a constant.
7169 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7170 The usual thing for this hook to do is to record flags in the
7171 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7172 Historically, the name string was modified if it was necessary to
7173 encode more than one bit of information, but this practice is now
7174 discouraged; use @code{SYMBOL_REF_FLAGS}.
7176 The default definition of this hook, @code{default_encode_section_info}
7177 in @file{varasm.c}, sets a number of commonly-useful bits in
7178 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7179 before overriding it.
7182 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7183 Decode @var{name} and return the real name part, sans
7184 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7188 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7189 Returns true if @var{exp} should be placed into a ``small data'' section.
7190 The default version of this hook always returns false.
7193 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7194 Contains the value true if the target places read-only
7195 ``small data'' into a separate section. The default value is false.
7198 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7199 It returns true if target wants profile code emitted before prologue.
7201 The default version of this hook use the target macro
7202 @code{PROFILE_BEFORE_PROLOGUE}.
7205 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7206 Returns true if @var{exp} names an object for which name resolution
7207 rules must resolve to the current ``module'' (dynamic shared library
7208 or executable image).
7210 The default version of this hook implements the name resolution rules
7211 for ELF, which has a looser model of global name binding than other
7212 currently supported object file formats.
7215 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7216 Contains the value true if the target supports thread-local storage.
7217 The default value is false.
7222 @section Position Independent Code
7223 @cindex position independent code
7226 This section describes macros that help implement generation of position
7227 independent code. Simply defining these macros is not enough to
7228 generate valid PIC; you must also add support to the hook
7229 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7230 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7231 must modify the definition of @samp{movsi} to do something appropriate
7232 when the source operand contains a symbolic address. You may also
7233 need to alter the handling of switch statements so that they use
7235 @c i rearranged the order of the macros above to try to force one of
7236 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7238 @defmac PIC_OFFSET_TABLE_REGNUM
7239 The register number of the register used to address a table of static
7240 data addresses in memory. In some cases this register is defined by a
7241 processor's ``application binary interface'' (ABI)@. When this macro
7242 is defined, RTL is generated for this register once, as with the stack
7243 pointer and frame pointer registers. If this macro is not defined, it
7244 is up to the machine-dependent files to allocate such a register (if
7245 necessary). Note that this register must be fixed when in use (e.g.@:
7246 when @code{flag_pic} is true).
7249 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7250 A C expression that is nonzero if the register defined by
7251 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7252 the default is zero. Do not define
7253 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7256 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7257 A C expression that is nonzero if @var{x} is a legitimate immediate
7258 operand on the target machine when generating position independent code.
7259 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7260 check this. You can also assume @var{flag_pic} is true, so you need not
7261 check it either. You need not define this macro if all constants
7262 (including @code{SYMBOL_REF}) can be immediate operands when generating
7263 position independent code.
7266 @node Assembler Format
7267 @section Defining the Output Assembler Language
7269 This section describes macros whose principal purpose is to describe how
7270 to write instructions in assembler language---rather than what the
7274 * File Framework:: Structural information for the assembler file.
7275 * Data Output:: Output of constants (numbers, strings, addresses).
7276 * Uninitialized Data:: Output of uninitialized variables.
7277 * Label Output:: Output and generation of labels.
7278 * Initialization:: General principles of initialization
7279 and termination routines.
7280 * Macros for Initialization::
7281 Specific macros that control the handling of
7282 initialization and termination routines.
7283 * Instruction Output:: Output of actual instructions.
7284 * Dispatch Tables:: Output of jump tables.
7285 * Exception Region Output:: Output of exception region code.
7286 * Alignment Output:: Pseudo ops for alignment and skipping data.
7289 @node File Framework
7290 @subsection The Overall Framework of an Assembler File
7291 @cindex assembler format
7292 @cindex output of assembler code
7294 @c prevent bad page break with this line
7295 This describes the overall framework of an assembly file.
7297 @findex default_file_start
7298 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7299 Output to @code{asm_out_file} any text which the assembler expects to
7300 find at the beginning of a file. The default behavior is controlled
7301 by two flags, documented below. Unless your target's assembler is
7302 quite unusual, if you override the default, you should call
7303 @code{default_file_start} at some point in your target hook. This
7304 lets other target files rely on these variables.
7307 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7308 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7309 printed as the very first line in the assembly file, unless
7310 @option{-fverbose-asm} is in effect. (If that macro has been defined
7311 to the empty string, this variable has no effect.) With the normal
7312 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7313 assembler that it need not bother stripping comments or extra
7314 whitespace from its input. This allows it to work a bit faster.
7316 The default is false. You should not set it to true unless you have
7317 verified that your port does not generate any extra whitespace or
7318 comments that will cause GAS to issue errors in NO_APP mode.
7321 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7322 If this flag is true, @code{output_file_directive} will be called
7323 for the primary source file, immediately after printing
7324 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7325 this to be done. The default is false.
7328 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7329 Output to @code{asm_out_file} any text which the assembler expects
7330 to find at the end of a file. The default is to output nothing.
7333 @deftypefun void file_end_indicate_exec_stack ()
7334 Some systems use a common convention, the @samp{.note.GNU-stack}
7335 special section, to indicate whether or not an object file relies on
7336 the stack being executable. If your system uses this convention, you
7337 should define @code{TARGET_ASM_FILE_END} to this function. If you
7338 need to do other things in that hook, have your hook function call
7342 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7343 Output to @code{asm_out_file} any text which the assembler expects
7344 to find at the start of an LTO section. The default is to output
7348 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7349 Output to @code{asm_out_file} any text which the assembler expects
7350 to find at the end of an LTO section. The default is to output
7354 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7355 Output to @code{asm_out_file} any text which is needed before emitting
7356 unwind info and debug info at the end of a file. Some targets emit
7357 here PIC setup thunks that cannot be emitted at the end of file,
7358 because they couldn't have unwind info then. The default is to output
7362 @defmac ASM_COMMENT_START
7363 A C string constant describing how to begin a comment in the target
7364 assembler language. The compiler assumes that the comment will end at
7365 the end of the line.
7369 A C string constant for text to be output before each @code{asm}
7370 statement or group of consecutive ones. Normally this is
7371 @code{"#APP"}, which is a comment that has no effect on most
7372 assemblers but tells the GNU assembler that it must check the lines
7373 that follow for all valid assembler constructs.
7377 A C string constant for text to be output after each @code{asm}
7378 statement or group of consecutive ones. Normally this is
7379 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7380 time-saving assumptions that are valid for ordinary compiler output.
7383 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7384 A C statement to output COFF information or DWARF debugging information
7385 which indicates that filename @var{name} is the current source file to
7386 the stdio stream @var{stream}.
7388 This macro need not be defined if the standard form of output
7389 for the file format in use is appropriate.
7392 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7393 Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7395 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7398 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_IDENT (const char *@var{name})
7399 Output a string based on @var{name}, suitable for the @samp{#ident} directive, or the equivalent directive or pragma in non-C-family languages. If this hook is not defined, nothing is output for the @samp{#ident} directive.
7402 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7403 A C statement to output the string @var{string} to the stdio stream
7404 @var{stream}. If you do not call the function @code{output_quoted_string}
7405 in your config files, GCC will only call it to output filenames to
7406 the assembler source. So you can use it to canonicalize the format
7407 of the filename using this macro.
7410 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7411 Output assembly directives to switch to section @var{name}. The section
7412 should have attributes as specified by @var{flags}, which is a bit mask
7413 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7414 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7415 this section is associated.
7418 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
7419 Return preferred text (sub)section for function @var{decl}.
7420 Main purpose of this function is to separate cold, normal and hot
7421 functions. @var{startup} is true when function is known to be used only
7422 at startup (from static constructors or it is @code{main()}).
7423 @var{exit} is true when function is known to be used only at exit
7424 (from static destructors).
7425 Return NULL if function should go to default text section.
7428 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
7429 Used by the target to emit any assembler directives or additional labels needed when a function is partitioned between different sections. Output should be written to @var{file}. The function decl is available as @var{decl} and the new section is `cold' if @var{new_is_cold} is @code{true}.
7432 @deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7433 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7434 It must not be modified by command-line option processing.
7437 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7438 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7439 This flag is true if we can create zeroed data by switching to a BSS
7440 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7441 This is true on most ELF targets.
7444 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7445 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7446 based on a variable or function decl, a section name, and whether or not the
7447 declaration's initializer may contain runtime relocations. @var{decl} may be
7448 null, in which case read-write data should be assumed.
7450 The default version of this function handles choosing code vs data,
7451 read-only vs read-write data, and @code{flag_pic}. You should only
7452 need to override this if your target has special flags that might be
7453 set via @code{__attribute__}.
7456 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
7457 Provides the target with the ability to record the gcc command line
7458 switches that have been passed to the compiler, and options that are
7459 enabled. The @var{type} argument specifies what is being recorded.
7460 It can take the following values:
7463 @item SWITCH_TYPE_PASSED
7464 @var{text} is a command line switch that has been set by the user.
7466 @item SWITCH_TYPE_ENABLED
7467 @var{text} is an option which has been enabled. This might be as a
7468 direct result of a command line switch, or because it is enabled by
7469 default or because it has been enabled as a side effect of a different
7470 command line switch. For example, the @option{-O2} switch enables
7471 various different individual optimization passes.
7473 @item SWITCH_TYPE_DESCRIPTIVE
7474 @var{text} is either NULL or some descriptive text which should be
7475 ignored. If @var{text} is NULL then it is being used to warn the
7476 target hook that either recording is starting or ending. The first
7477 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7478 warning is for start up and the second time the warning is for
7479 wind down. This feature is to allow the target hook to make any
7480 necessary preparations before it starts to record switches and to
7481 perform any necessary tidying up after it has finished recording
7484 @item SWITCH_TYPE_LINE_START
7485 This option can be ignored by this target hook.
7487 @item SWITCH_TYPE_LINE_END
7488 This option can be ignored by this target hook.
7491 The hook's return value must be zero. Other return values may be
7492 supported in the future.
7494 By default this hook is set to NULL, but an example implementation is
7495 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7496 it records the switches as ASCII text inside a new, string mergeable
7497 section in the assembler output file. The name of the new section is
7498 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7502 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7503 This is the name of the section that will be created by the example
7504 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7510 @subsection Output of Data
7513 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7514 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7515 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7516 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7517 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7518 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7519 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7520 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7521 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7522 These hooks specify assembly directives for creating certain kinds
7523 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7524 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7525 aligned two-byte object, and so on. Any of the hooks may be
7526 @code{NULL}, indicating that no suitable directive is available.
7528 The compiler will print these strings at the start of a new line,
7529 followed immediately by the object's initial value. In most cases,
7530 the string should contain a tab, a pseudo-op, and then another tab.
7533 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7534 The @code{assemble_integer} function uses this hook to output an
7535 integer object. @var{x} is the object's value, @var{size} is its size
7536 in bytes and @var{aligned_p} indicates whether it is aligned. The
7537 function should return @code{true} if it was able to output the
7538 object. If it returns false, @code{assemble_integer} will try to
7539 split the object into smaller parts.
7541 The default implementation of this hook will use the
7542 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7543 when the relevant string is @code{NULL}.
7546 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
7547 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7548 can't deal with, and output assembly code to @var{file} corresponding to
7549 the pattern @var{x}. This may be used to allow machine-dependent
7550 @code{UNSPEC}s to appear within constants.
7552 If target hook fails to recognize a pattern, it must return @code{false},
7553 so that a standard error message is printed. If it prints an error message
7554 itself, by calling, for example, @code{output_operand_lossage}, it may just
7558 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7559 A C statement to output to the stdio stream @var{stream} an assembler
7560 instruction to assemble a string constant containing the @var{len}
7561 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7562 @code{char *} and @var{len} a C expression of type @code{int}.
7564 If the assembler has a @code{.ascii} pseudo-op as found in the
7565 Berkeley Unix assembler, do not define the macro
7566 @code{ASM_OUTPUT_ASCII}.
7569 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7570 A C statement to output word @var{n} of a function descriptor for
7571 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7572 is defined, and is otherwise unused.
7575 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7576 You may define this macro as a C expression. You should define the
7577 expression to have a nonzero value if GCC should output the constant
7578 pool for a function before the code for the function, or a zero value if
7579 GCC should output the constant pool after the function. If you do
7580 not define this macro, the usual case, GCC will output the constant
7581 pool before the function.
7584 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7585 A C statement to output assembler commands to define the start of the
7586 constant pool for a function. @var{funname} is a string giving
7587 the name of the function. Should the return type of the function
7588 be required, it can be obtained via @var{fundecl}. @var{size}
7589 is the size, in bytes, of the constant pool that will be written
7590 immediately after this call.
7592 If no constant-pool prefix is required, the usual case, this macro need
7596 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7597 A C statement (with or without semicolon) to output a constant in the
7598 constant pool, if it needs special treatment. (This macro need not do
7599 anything for RTL expressions that can be output normally.)
7601 The argument @var{file} is the standard I/O stream to output the
7602 assembler code on. @var{x} is the RTL expression for the constant to
7603 output, and @var{mode} is the machine mode (in case @var{x} is a
7604 @samp{const_int}). @var{align} is the required alignment for the value
7605 @var{x}; you should output an assembler directive to force this much
7608 The argument @var{labelno} is a number to use in an internal label for
7609 the address of this pool entry. The definition of this macro is
7610 responsible for outputting the label definition at the proper place.
7611 Here is how to do this:
7614 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7617 When you output a pool entry specially, you should end with a
7618 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7619 entry from being output a second time in the usual manner.
7621 You need not define this macro if it would do nothing.
7624 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7625 A C statement to output assembler commands to at the end of the constant
7626 pool for a function. @var{funname} is a string giving the name of the
7627 function. Should the return type of the function be required, you can
7628 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7629 constant pool that GCC wrote immediately before this call.
7631 If no constant-pool epilogue is required, the usual case, you need not
7635 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7636 Define this macro as a C expression which is nonzero if @var{C} is
7637 used as a logical line separator by the assembler. @var{STR} points
7638 to the position in the string where @var{C} was found; this can be used if
7639 a line separator uses multiple characters.
7641 If you do not define this macro, the default is that only
7642 the character @samp{;} is treated as a logical line separator.
7645 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7646 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7647 These target hooks are C string constants, describing the syntax in the
7648 assembler for grouping arithmetic expressions. If not overridden, they
7649 default to normal parentheses, which is correct for most assemblers.
7652 These macros are provided by @file{real.h} for writing the definitions
7653 of @code{ASM_OUTPUT_DOUBLE} and the like:
7655 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7656 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7657 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7658 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7659 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7660 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7661 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7662 target's floating point representation, and store its bit pattern in
7663 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7664 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7665 simple @code{long int}. For the others, it should be an array of
7666 @code{long int}. The number of elements in this array is determined
7667 by the size of the desired target floating point data type: 32 bits of
7668 it go in each @code{long int} array element. Each array element holds
7669 32 bits of the result, even if @code{long int} is wider than 32 bits
7670 on the host machine.
7672 The array element values are designed so that you can print them out
7673 using @code{fprintf} in the order they should appear in the target
7677 @node Uninitialized Data
7678 @subsection Output of Uninitialized Variables
7680 Each of the macros in this section is used to do the whole job of
7681 outputting a single uninitialized variable.
7683 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7684 A C statement (sans semicolon) to output to the stdio stream
7685 @var{stream} the assembler definition of a common-label named
7686 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7687 is the size rounded up to whatever alignment the caller wants. It is
7688 possible that @var{size} may be zero, for instance if a struct with no
7689 other member than a zero-length array is defined. In this case, the
7690 backend must output a symbol definition that allocates at least one
7691 byte, both so that the address of the resulting object does not compare
7692 equal to any other, and because some object formats cannot even express
7693 the concept of a zero-sized common symbol, as that is how they represent
7694 an ordinary undefined external.
7696 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7697 output the name itself; before and after that, output the additional
7698 assembler syntax for defining the name, and a newline.
7700 This macro controls how the assembler definitions of uninitialized
7701 common global variables are output.
7704 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7705 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7706 separate, explicit argument. If you define this macro, it is used in
7707 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7708 handling the required alignment of the variable. The alignment is specified
7709 as the number of bits.
7712 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7713 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7714 variable to be output, if there is one, or @code{NULL_TREE} if there
7715 is no corresponding variable. If you define this macro, GCC will use it
7716 in place of both @code{ASM_OUTPUT_COMMON} and
7717 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7718 the variable's decl in order to chose what to output.
7721 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7722 A C statement (sans semicolon) to output to the stdio stream
7723 @var{stream} the assembler definition of uninitialized global @var{decl} named
7724 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
7725 is the alignment specified as the number of bits.
7727 Try to use function @code{asm_output_aligned_bss} defined in file
7728 @file{varasm.c} when defining this macro. If unable, use the expression
7729 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7730 before and after that, output the additional assembler syntax for defining
7731 the name, and a newline.
7733 There are two ways of handling global BSS@. One is to define this macro.
7734 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7735 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7736 You do not need to do both.
7738 Some languages do not have @code{common} data, and require a
7739 non-common form of global BSS in order to handle uninitialized globals
7740 efficiently. C++ is one example of this. However, if the target does
7741 not support global BSS, the front end may choose to make globals
7742 common in order to save space in the object file.
7745 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7746 A C statement (sans semicolon) to output to the stdio stream
7747 @var{stream} the assembler definition of a local-common-label named
7748 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7749 is the size rounded up to whatever alignment the caller wants.
7751 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7752 output the name itself; before and after that, output the additional
7753 assembler syntax for defining the name, and a newline.
7755 This macro controls how the assembler definitions of uninitialized
7756 static variables are output.
7759 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7760 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7761 separate, explicit argument. If you define this macro, it is used in
7762 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7763 handling the required alignment of the variable. The alignment is specified
7764 as the number of bits.
7767 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7768 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7769 variable to be output, if there is one, or @code{NULL_TREE} if there
7770 is no corresponding variable. If you define this macro, GCC will use it
7771 in place of both @code{ASM_OUTPUT_DECL} and
7772 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7773 the variable's decl in order to chose what to output.
7777 @subsection Output and Generation of Labels
7779 @c prevent bad page break with this line
7780 This is about outputting labels.
7782 @findex assemble_name
7783 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7784 A C statement (sans semicolon) to output to the stdio stream
7785 @var{stream} the assembler definition of a label named @var{name}.
7786 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7787 output the name itself; before and after that, output the additional
7788 assembler syntax for defining the name, and a newline. A default
7789 definition of this macro is provided which is correct for most systems.
7792 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7793 A C statement (sans semicolon) to output to the stdio stream
7794 @var{stream} the assembler definition of a label named @var{name} of
7796 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7797 output the name itself; before and after that, output the additional
7798 assembler syntax for defining the name, and a newline. A default
7799 definition of this macro is provided which is correct for most systems.
7801 If this macro is not defined, then the function name is defined in the
7802 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7805 @findex assemble_name_raw
7806 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7807 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7808 to refer to a compiler-generated label. The default definition uses
7809 @code{assemble_name_raw}, which is like @code{assemble_name} except
7810 that it is more efficient.
7814 A C string containing the appropriate assembler directive to specify the
7815 size of a symbol, without any arguments. On systems that use ELF, the
7816 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7817 systems, the default is not to define this macro.
7819 Define this macro only if it is correct to use the default definitions
7820 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7821 for your system. If you need your own custom definitions of those
7822 macros, or if you do not need explicit symbol sizes at all, do not
7826 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7827 A C statement (sans semicolon) to output to the stdio stream
7828 @var{stream} a directive telling the assembler that the size of the
7829 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7830 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7834 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7835 A C statement (sans semicolon) to output to the stdio stream
7836 @var{stream} a directive telling the assembler to calculate the size of
7837 the symbol @var{name} by subtracting its address from the current
7840 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7841 provided. The default assumes that the assembler recognizes a special
7842 @samp{.} symbol as referring to the current address, and can calculate
7843 the difference between this and another symbol. If your assembler does
7844 not recognize @samp{.} or cannot do calculations with it, you will need
7845 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7848 @defmac NO_DOLLAR_IN_LABEL
7849 Define this macro if the assembler does not accept the character
7850 @samp{$} in label names. By default constructors and destructors in
7851 G++ have @samp{$} in the identifiers. If this macro is defined,
7852 @samp{.} is used instead.
7855 @defmac NO_DOT_IN_LABEL
7856 Define this macro if the assembler does not accept the character
7857 @samp{.} in label names. By default constructors and destructors in G++
7858 have names that use @samp{.}. If this macro is defined, these names
7859 are rewritten to avoid @samp{.}.
7863 A C string containing the appropriate assembler directive to specify the
7864 type of a symbol, without any arguments. On systems that use ELF, the
7865 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7866 systems, the default is not to define this macro.
7868 Define this macro only if it is correct to use the default definition of
7869 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7870 custom definition of this macro, or if you do not need explicit symbol
7871 types at all, do not define this macro.
7874 @defmac TYPE_OPERAND_FMT
7875 A C string which specifies (using @code{printf} syntax) the format of
7876 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7877 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7878 the default is not to define this macro.
7880 Define this macro only if it is correct to use the default definition of
7881 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7882 custom definition of this macro, or if you do not need explicit symbol
7883 types at all, do not define this macro.
7886 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7887 A C statement (sans semicolon) to output to the stdio stream
7888 @var{stream} a directive telling the assembler that the type of the
7889 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7890 that string is always either @samp{"function"} or @samp{"object"}, but
7891 you should not count on this.
7893 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7894 definition of this macro is provided.
7897 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7898 A C statement (sans semicolon) to output to the stdio stream
7899 @var{stream} any text necessary for declaring the name @var{name} of a
7900 function which is being defined. This macro is responsible for
7901 outputting the label definition (perhaps using
7902 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7903 @code{FUNCTION_DECL} tree node representing the function.
7905 If this macro is not defined, then the function name is defined in the
7906 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7908 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7912 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7913 A C statement (sans semicolon) to output to the stdio stream
7914 @var{stream} any text necessary for declaring the size of a function
7915 which is being defined. The argument @var{name} is the name of the
7916 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7917 representing the function.
7919 If this macro is not defined, then the function size is not defined.
7921 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7925 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7926 A C statement (sans semicolon) to output to the stdio stream
7927 @var{stream} any text necessary for declaring the name @var{name} of an
7928 initialized variable which is being defined. This macro must output the
7929 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7930 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7932 If this macro is not defined, then the variable name is defined in the
7933 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7935 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7936 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7939 @deftypefn {Target Hook} void TARGET_ASM_DECLARE_CONSTANT_NAME (FILE *@var{file}, const char *@var{name}, const_tree @var{expr}, HOST_WIDE_INT @var{size})
7940 A target hook to output to the stdio stream @var{file} any text necessary
7941 for declaring the name @var{name} of a constant which is being defined. This
7942 target hook is responsible for outputting the label definition (perhaps using
7943 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7944 and @var{size} is the size of the constant in bytes. The @var{name}
7945 will be an internal label.
7947 The default version of this target hook, define the @var{name} in the
7948 usual manner as a label (by means of @code{assemble_label}).
7950 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7953 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7954 A C statement (sans semicolon) to output to the stdio stream
7955 @var{stream} any text necessary for claiming a register @var{regno}
7956 for a global variable @var{decl} with name @var{name}.
7958 If you don't define this macro, that is equivalent to defining it to do
7962 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7963 A C statement (sans semicolon) to finish up declaring a variable name
7964 once the compiler has processed its initializer fully and thus has had a
7965 chance to determine the size of an array when controlled by an
7966 initializer. This is used on systems where it's necessary to declare
7967 something about the size of the object.
7969 If you don't define this macro, that is equivalent to defining it to do
7972 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7973 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7976 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7977 This target hook is a function to output to the stdio stream
7978 @var{stream} some commands that will make the label @var{name} global;
7979 that is, available for reference from other files.
7981 The default implementation relies on a proper definition of
7982 @code{GLOBAL_ASM_OP}.
7985 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7986 This target hook is a function to output to the stdio stream
7987 @var{stream} some commands that will make the name associated with @var{decl}
7988 global; that is, available for reference from other files.
7990 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7993 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7994 A C statement (sans semicolon) to output to the stdio stream
7995 @var{stream} some commands that will make the label @var{name} weak;
7996 that is, available for reference from other files but only used if
7997 no other definition is available. Use the expression
7998 @code{assemble_name (@var{stream}, @var{name})} to output the name
7999 itself; before and after that, output the additional assembler syntax
8000 for making that name weak, and a newline.
8002 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
8003 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
8007 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
8008 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
8009 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
8010 or variable decl. If @var{value} is not @code{NULL}, this C statement
8011 should output to the stdio stream @var{stream} assembler code which
8012 defines (equates) the weak symbol @var{name} to have the value
8013 @var{value}. If @var{value} is @code{NULL}, it should output commands
8014 to make @var{name} weak.
8017 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
8018 Outputs a directive that enables @var{name} to be used to refer to
8019 symbol @var{value} with weak-symbol semantics. @code{decl} is the
8020 declaration of @code{name}.
8023 @defmac SUPPORTS_WEAK
8024 A preprocessor constant expression which evaluates to true if the target
8025 supports weak symbols.
8027 If you don't define this macro, @file{defaults.h} provides a default
8028 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
8029 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
8032 @defmac TARGET_SUPPORTS_WEAK
8033 A C expression which evaluates to true if the target supports weak symbols.
8035 If you don't define this macro, @file{defaults.h} provides a default
8036 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
8037 this macro if you want to control weak symbol support with a compiler
8038 flag such as @option{-melf}.
8041 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
8042 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
8043 public symbol such that extra copies in multiple translation units will
8044 be discarded by the linker. Define this macro if your object file
8045 format provides support for this concept, such as the @samp{COMDAT}
8046 section flags in the Microsoft Windows PE/COFF format, and this support
8047 requires changes to @var{decl}, such as putting it in a separate section.
8050 @defmac SUPPORTS_ONE_ONLY
8051 A C expression which evaluates to true if the target supports one-only
8054 If you don't define this macro, @file{varasm.c} provides a default
8055 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
8056 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
8057 you want to control one-only symbol support with a compiler flag, or if
8058 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
8059 be emitted as one-only.
8062 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
8063 This target hook is a function to output to @var{asm_out_file} some
8064 commands that will make the symbol(s) associated with @var{decl} have
8065 hidden, protected or internal visibility as specified by @var{visibility}.
8068 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
8069 A C expression that evaluates to true if the target's linker expects
8070 that weak symbols do not appear in a static archive's table of contents.
8071 The default is @code{0}.
8073 Leaving weak symbols out of an archive's table of contents means that,
8074 if a symbol will only have a definition in one translation unit and
8075 will have undefined references from other translation units, that
8076 symbol should not be weak. Defining this macro to be nonzero will
8077 thus have the effect that certain symbols that would normally be weak
8078 (explicit template instantiations, and vtables for polymorphic classes
8079 with noninline key methods) will instead be nonweak.
8081 The C++ ABI requires this macro to be zero. Define this macro for
8082 targets where full C++ ABI compliance is impossible and where linker
8083 restrictions require weak symbols to be left out of a static archive's
8087 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8088 A C statement (sans semicolon) to output to the stdio stream
8089 @var{stream} any text necessary for declaring the name of an external
8090 symbol named @var{name} which is referenced in this compilation but
8091 not defined. The value of @var{decl} is the tree node for the
8094 This macro need not be defined if it does not need to output anything.
8095 The GNU assembler and most Unix assemblers don't require anything.
8098 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8099 This target hook is a function to output to @var{asm_out_file} an assembler
8100 pseudo-op to declare a library function name external. The name of the
8101 library function is given by @var{symref}, which is a @code{symbol_ref}.
8104 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8105 This target hook is a function to output to @var{asm_out_file} an assembler
8106 directive to annotate @var{symbol} as used. The Darwin target uses the
8107 .no_dead_code_strip directive.
8110 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8111 A C statement (sans semicolon) to output to the stdio stream
8112 @var{stream} a reference in assembler syntax to a label named
8113 @var{name}. This should add @samp{_} to the front of the name, if that
8114 is customary on your operating system, as it is in most Berkeley Unix
8115 systems. This macro is used in @code{assemble_name}.
8118 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8119 Given a symbol @var{name}, perform same mangling as @code{varasm.c}'s @code{assemble_name}, but in memory rather than to a file stream, returning result as an @code{IDENTIFIER_NODE}. Required for correct LTO symtabs. The default implementation calls the @code{TARGET_STRIP_NAME_ENCODING} hook and then prepends the @code{USER_LABEL_PREFIX}, if any.
8122 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8123 A C statement (sans semicolon) to output a reference to
8124 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8125 will be used to output the name of the symbol. This macro may be used
8126 to modify the way a symbol is referenced depending on information
8127 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8130 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8131 A C statement (sans semicolon) to output a reference to @var{buf}, the
8132 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8133 @code{assemble_name} will be used to output the name of the symbol.
8134 This macro is not used by @code{output_asm_label}, or the @code{%l}
8135 specifier that calls it; the intention is that this macro should be set
8136 when it is necessary to output a label differently when its address is
8140 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8141 A function to output to the stdio stream @var{stream} a label whose
8142 name is made from the string @var{prefix} and the number @var{labelno}.
8144 It is absolutely essential that these labels be distinct from the labels
8145 used for user-level functions and variables. Otherwise, certain programs
8146 will have name conflicts with internal labels.
8148 It is desirable to exclude internal labels from the symbol table of the
8149 object file. Most assemblers have a naming convention for labels that
8150 should be excluded; on many systems, the letter @samp{L} at the
8151 beginning of a label has this effect. You should find out what
8152 convention your system uses, and follow it.
8154 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8157 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8158 A C statement to output to the stdio stream @var{stream} a debug info
8159 label whose name is made from the string @var{prefix} and the number
8160 @var{num}. This is useful for VLIW targets, where debug info labels
8161 may need to be treated differently than branch target labels. On some
8162 systems, branch target labels must be at the beginning of instruction
8163 bundles, but debug info labels can occur in the middle of instruction
8166 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8170 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8171 A C statement to store into the string @var{string} a label whose name
8172 is made from the string @var{prefix} and the number @var{num}.
8174 This string, when output subsequently by @code{assemble_name}, should
8175 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8176 with the same @var{prefix} and @var{num}.
8178 If the string begins with @samp{*}, then @code{assemble_name} will
8179 output the rest of the string unchanged. It is often convenient for
8180 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8181 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8182 to output the string, and may change it. (Of course,
8183 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8184 you should know what it does on your machine.)
8187 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8188 A C expression to assign to @var{outvar} (which is a variable of type
8189 @code{char *}) a newly allocated string made from the string
8190 @var{name} and the number @var{number}, with some suitable punctuation
8191 added. Use @code{alloca} to get space for the string.
8193 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8194 produce an assembler label for an internal static variable whose name is
8195 @var{name}. Therefore, the string must be such as to result in valid
8196 assembler code. The argument @var{number} is different each time this
8197 macro is executed; it prevents conflicts between similarly-named
8198 internal static variables in different scopes.
8200 Ideally this string should not be a valid C identifier, to prevent any
8201 conflict with the user's own symbols. Most assemblers allow periods
8202 or percent signs in assembler symbols; putting at least one of these
8203 between the name and the number will suffice.
8205 If this macro is not defined, a default definition will be provided
8206 which is correct for most systems.
8209 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8210 A C statement to output to the stdio stream @var{stream} assembler code
8211 which defines (equates) the symbol @var{name} to have the value @var{value}.
8214 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8215 correct for most systems.
8218 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8219 A C statement to output to the stdio stream @var{stream} assembler code
8220 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8221 to have the value of the tree node @var{decl_of_value}. This macro will
8222 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8223 the tree nodes are available.
8226 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8227 correct for most systems.
8230 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8231 A C statement that evaluates to true if the assembler code which defines
8232 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8233 of the tree node @var{decl_of_value} should be emitted near the end of the
8234 current compilation unit. The default is to not defer output of defines.
8235 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8236 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8239 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8240 A C statement to output to the stdio stream @var{stream} assembler code
8241 which defines (equates) the weak symbol @var{name} to have the value
8242 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8243 an undefined weak symbol.
8245 Define this macro if the target only supports weak aliases; define
8246 @code{ASM_OUTPUT_DEF} instead if possible.
8249 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8250 Define this macro to override the default assembler names used for
8251 Objective-C methods.
8253 The default name is a unique method number followed by the name of the
8254 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8255 the category is also included in the assembler name (e.g.@:
8258 These names are safe on most systems, but make debugging difficult since
8259 the method's selector is not present in the name. Therefore, particular
8260 systems define other ways of computing names.
8262 @var{buf} is an expression of type @code{char *} which gives you a
8263 buffer in which to store the name; its length is as long as
8264 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8265 50 characters extra.
8267 The argument @var{is_inst} specifies whether the method is an instance
8268 method or a class method; @var{class_name} is the name of the class;
8269 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8270 in a category); and @var{sel_name} is the name of the selector.
8272 On systems where the assembler can handle quoted names, you can use this
8273 macro to provide more human-readable names.
8276 @node Initialization
8277 @subsection How Initialization Functions Are Handled
8278 @cindex initialization routines
8279 @cindex termination routines
8280 @cindex constructors, output of
8281 @cindex destructors, output of
8283 The compiled code for certain languages includes @dfn{constructors}
8284 (also called @dfn{initialization routines})---functions to initialize
8285 data in the program when the program is started. These functions need
8286 to be called before the program is ``started''---that is to say, before
8287 @code{main} is called.
8289 Compiling some languages generates @dfn{destructors} (also called
8290 @dfn{termination routines}) that should be called when the program
8293 To make the initialization and termination functions work, the compiler
8294 must output something in the assembler code to cause those functions to
8295 be called at the appropriate time. When you port the compiler to a new
8296 system, you need to specify how to do this.
8298 There are two major ways that GCC currently supports the execution of
8299 initialization and termination functions. Each way has two variants.
8300 Much of the structure is common to all four variations.
8302 @findex __CTOR_LIST__
8303 @findex __DTOR_LIST__
8304 The linker must build two lists of these functions---a list of
8305 initialization functions, called @code{__CTOR_LIST__}, and a list of
8306 termination functions, called @code{__DTOR_LIST__}.
8308 Each list always begins with an ignored function pointer (which may hold
8309 0, @minus{}1, or a count of the function pointers after it, depending on
8310 the environment). This is followed by a series of zero or more function
8311 pointers to constructors (or destructors), followed by a function
8312 pointer containing zero.
8314 Depending on the operating system and its executable file format, either
8315 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8316 time and exit time. Constructors are called in reverse order of the
8317 list; destructors in forward order.
8319 The best way to handle static constructors works only for object file
8320 formats which provide arbitrarily-named sections. A section is set
8321 aside for a list of constructors, and another for a list of destructors.
8322 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8323 object file that defines an initialization function also puts a word in
8324 the constructor section to point to that function. The linker
8325 accumulates all these words into one contiguous @samp{.ctors} section.
8326 Termination functions are handled similarly.
8328 This method will be chosen as the default by @file{target-def.h} if
8329 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8330 support arbitrary sections, but does support special designated
8331 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8332 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8334 When arbitrary sections are available, there are two variants, depending
8335 upon how the code in @file{crtstuff.c} is called. On systems that
8336 support a @dfn{.init} section which is executed at program startup,
8337 parts of @file{crtstuff.c} are compiled into that section. The
8338 program is linked by the @command{gcc} driver like this:
8341 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8344 The prologue of a function (@code{__init}) appears in the @code{.init}
8345 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8346 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8347 files are provided by the operating system or by the GNU C library, but
8348 are provided by GCC for a few targets.
8350 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8351 compiled from @file{crtstuff.c}. They contain, among other things, code
8352 fragments within the @code{.init} and @code{.fini} sections that branch
8353 to routines in the @code{.text} section. The linker will pull all parts
8354 of a section together, which results in a complete @code{__init} function
8355 that invokes the routines we need at startup.
8357 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8360 If no init section is available, when GCC compiles any function called
8361 @code{main} (or more accurately, any function designated as a program
8362 entry point by the language front end calling @code{expand_main_function}),
8363 it inserts a procedure call to @code{__main} as the first executable code
8364 after the function prologue. The @code{__main} function is defined
8365 in @file{libgcc2.c} and runs the global constructors.
8367 In file formats that don't support arbitrary sections, there are again
8368 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8369 and an `a.out' format must be used. In this case,
8370 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8371 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8372 and with the address of the void function containing the initialization
8373 code as its value. The GNU linker recognizes this as a request to add
8374 the value to a @dfn{set}; the values are accumulated, and are eventually
8375 placed in the executable as a vector in the format described above, with
8376 a leading (ignored) count and a trailing zero element.
8377 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8378 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8379 the compilation of @code{main} to call @code{__main} as above, starting
8380 the initialization process.
8382 The last variant uses neither arbitrary sections nor the GNU linker.
8383 This is preferable when you want to do dynamic linking and when using
8384 file formats which the GNU linker does not support, such as `ECOFF'@. In
8385 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8386 termination functions are recognized simply by their names. This requires
8387 an extra program in the linkage step, called @command{collect2}. This program
8388 pretends to be the linker, for use with GCC; it does its job by running
8389 the ordinary linker, but also arranges to include the vectors of
8390 initialization and termination functions. These functions are called
8391 via @code{__main} as described above. In order to use this method,
8392 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8395 The following section describes the specific macros that control and
8396 customize the handling of initialization and termination functions.
8399 @node Macros for Initialization
8400 @subsection Macros Controlling Initialization Routines
8402 Here are the macros that control how the compiler handles initialization
8403 and termination functions:
8405 @defmac INIT_SECTION_ASM_OP
8406 If defined, a C string constant, including spacing, for the assembler
8407 operation to identify the following data as initialization code. If not
8408 defined, GCC will assume such a section does not exist. When you are
8409 using special sections for initialization and termination functions, this
8410 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8411 run the initialization functions.
8414 @defmac HAS_INIT_SECTION
8415 If defined, @code{main} will not call @code{__main} as described above.
8416 This macro should be defined for systems that control start-up code
8417 on a symbol-by-symbol basis, such as OSF/1, and should not
8418 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8421 @defmac LD_INIT_SWITCH
8422 If defined, a C string constant for a switch that tells the linker that
8423 the following symbol is an initialization routine.
8426 @defmac LD_FINI_SWITCH
8427 If defined, a C string constant for a switch that tells the linker that
8428 the following symbol is a finalization routine.
8431 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8432 If defined, a C statement that will write a function that can be
8433 automatically called when a shared library is loaded. The function
8434 should call @var{func}, which takes no arguments. If not defined, and
8435 the object format requires an explicit initialization function, then a
8436 function called @code{_GLOBAL__DI} will be generated.
8438 This function and the following one are used by collect2 when linking a
8439 shared library that needs constructors or destructors, or has DWARF2
8440 exception tables embedded in the code.
8443 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8444 If defined, a C statement that will write a function that can be
8445 automatically called when a shared library is unloaded. The function
8446 should call @var{func}, which takes no arguments. If not defined, and
8447 the object format requires an explicit finalization function, then a
8448 function called @code{_GLOBAL__DD} will be generated.
8451 @defmac INVOKE__main
8452 If defined, @code{main} will call @code{__main} despite the presence of
8453 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8454 where the init section is not actually run automatically, but is still
8455 useful for collecting the lists of constructors and destructors.
8458 @defmac SUPPORTS_INIT_PRIORITY
8459 If nonzero, the C++ @code{init_priority} attribute is supported and the
8460 compiler should emit instructions to control the order of initialization
8461 of objects. If zero, the compiler will issue an error message upon
8462 encountering an @code{init_priority} attribute.
8465 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8466 This value is true if the target supports some ``native'' method of
8467 collecting constructors and destructors to be run at startup and exit.
8468 It is false if we must use @command{collect2}.
8471 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8472 If defined, a function that outputs assembler code to arrange to call
8473 the function referenced by @var{symbol} at initialization time.
8475 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8476 no arguments and with no return value. If the target supports initialization
8477 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8478 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8480 If this macro is not defined by the target, a suitable default will
8481 be chosen if (1) the target supports arbitrary section names, (2) the
8482 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8486 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8487 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8488 functions rather than initialization functions.
8491 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8492 generated for the generated object file will have static linkage.
8494 If your system uses @command{collect2} as the means of processing
8495 constructors, then that program normally uses @command{nm} to scan
8496 an object file for constructor functions to be called.
8498 On certain kinds of systems, you can define this macro to make
8499 @command{collect2} work faster (and, in some cases, make it work at all):
8501 @defmac OBJECT_FORMAT_COFF
8502 Define this macro if the system uses COFF (Common Object File Format)
8503 object files, so that @command{collect2} can assume this format and scan
8504 object files directly for dynamic constructor/destructor functions.
8506 This macro is effective only in a native compiler; @command{collect2} as
8507 part of a cross compiler always uses @command{nm} for the target machine.
8510 @defmac REAL_NM_FILE_NAME
8511 Define this macro as a C string constant containing the file name to use
8512 to execute @command{nm}. The default is to search the path normally for
8517 @command{collect2} calls @command{nm} to scan object files for static
8518 constructors and destructors and LTO info. By default, @option{-n} is
8519 passed. Define @code{NM_FLAGS} to a C string constant if other options
8520 are needed to get the same output format as GNU @command{nm -n}
8524 If your system supports shared libraries and has a program to list the
8525 dynamic dependencies of a given library or executable, you can define
8526 these macros to enable support for running initialization and
8527 termination functions in shared libraries:
8530 Define this macro to a C string constant containing the name of the program
8531 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8534 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8535 Define this macro to be C code that extracts filenames from the output
8536 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8537 of type @code{char *} that points to the beginning of a line of output
8538 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8539 code must advance @var{ptr} to the beginning of the filename on that
8540 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8543 @defmac SHLIB_SUFFIX
8544 Define this macro to a C string constant containing the default shared
8545 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8546 strips version information after this suffix when generating global
8547 constructor and destructor names. This define is only needed on targets
8548 that use @command{collect2} to process constructors and destructors.
8551 @node Instruction Output
8552 @subsection Output of Assembler Instructions
8554 @c prevent bad page break with this line
8555 This describes assembler instruction output.
8557 @defmac REGISTER_NAMES
8558 A C initializer containing the assembler's names for the machine
8559 registers, each one as a C string constant. This is what translates
8560 register numbers in the compiler into assembler language.
8563 @defmac ADDITIONAL_REGISTER_NAMES
8564 If defined, a C initializer for an array of structures containing a name
8565 and a register number. This macro defines additional names for hard
8566 registers, thus allowing the @code{asm} option in declarations to refer
8567 to registers using alternate names.
8570 @defmac OVERLAPPING_REGISTER_NAMES
8571 If defined, a C initializer for an array of structures containing a
8572 name, a register number and a count of the number of consecutive
8573 machine registers the name overlaps. This macro defines additional
8574 names for hard registers, thus allowing the @code{asm} option in
8575 declarations to refer to registers using alternate names. Unlike
8576 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8577 register name implies multiple underlying registers.
8579 This macro should be used when it is important that a clobber in an
8580 @code{asm} statement clobbers all the underlying values implied by the
8581 register name. For example, on ARM, clobbering the double-precision
8582 VFP register ``d0'' implies clobbering both single-precision registers
8586 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8587 Define this macro if you are using an unusual assembler that
8588 requires different names for the machine instructions.
8590 The definition is a C statement or statements which output an
8591 assembler instruction opcode to the stdio stream @var{stream}. The
8592 macro-operand @var{ptr} is a variable of type @code{char *} which
8593 points to the opcode name in its ``internal'' form---the form that is
8594 written in the machine description. The definition should output the
8595 opcode name to @var{stream}, performing any translation you desire, and
8596 increment the variable @var{ptr} to point at the end of the opcode
8597 so that it will not be output twice.
8599 In fact, your macro definition may process less than the entire opcode
8600 name, or more than the opcode name; but if you want to process text
8601 that includes @samp{%}-sequences to substitute operands, you must take
8602 care of the substitution yourself. Just be sure to increment
8603 @var{ptr} over whatever text should not be output normally.
8605 @findex recog_data.operand
8606 If you need to look at the operand values, they can be found as the
8607 elements of @code{recog_data.operand}.
8609 If the macro definition does nothing, the instruction is output
8613 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8614 If defined, a C statement to be executed just prior to the output of
8615 assembler code for @var{insn}, to modify the extracted operands so
8616 they will be output differently.
8618 Here the argument @var{opvec} is the vector containing the operands
8619 extracted from @var{insn}, and @var{noperands} is the number of
8620 elements of the vector which contain meaningful data for this insn.
8621 The contents of this vector are what will be used to convert the insn
8622 template into assembler code, so you can change the assembler output
8623 by changing the contents of the vector.
8625 This macro is useful when various assembler syntaxes share a single
8626 file of instruction patterns; by defining this macro differently, you
8627 can cause a large class of instructions to be output differently (such
8628 as with rearranged operands). Naturally, variations in assembler
8629 syntax affecting individual insn patterns ought to be handled by
8630 writing conditional output routines in those patterns.
8632 If this macro is not defined, it is equivalent to a null statement.
8635 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx @var{insn}, rtx *@var{opvec}, int @var{noperands})
8636 If defined, this target hook is a function which is executed just after the
8637 output of assembler code for @var{insn}, to change the mode of the assembler
8640 Here the argument @var{opvec} is the vector containing the operands
8641 extracted from @var{insn}, and @var{noperands} is the number of
8642 elements of the vector which contain meaningful data for this insn.
8643 The contents of this vector are what was used to convert the insn
8644 template into assembler code, so you can change the assembler mode
8645 by checking the contents of the vector.
8648 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8649 A C compound statement to output to stdio stream @var{stream} the
8650 assembler syntax for an instruction operand @var{x}. @var{x} is an
8653 @var{code} is a value that can be used to specify one of several ways
8654 of printing the operand. It is used when identical operands must be
8655 printed differently depending on the context. @var{code} comes from
8656 the @samp{%} specification that was used to request printing of the
8657 operand. If the specification was just @samp{%@var{digit}} then
8658 @var{code} is 0; if the specification was @samp{%@var{ltr}
8659 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8662 If @var{x} is a register, this macro should print the register's name.
8663 The names can be found in an array @code{reg_names} whose type is
8664 @code{char *[]}. @code{reg_names} is initialized from
8665 @code{REGISTER_NAMES}.
8667 When the machine description has a specification @samp{%@var{punct}}
8668 (a @samp{%} followed by a punctuation character), this macro is called
8669 with a null pointer for @var{x} and the punctuation character for
8673 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8674 A C expression which evaluates to true if @var{code} is a valid
8675 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8676 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8677 punctuation characters (except for the standard one, @samp{%}) are used
8681 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8682 A C compound statement to output to stdio stream @var{stream} the
8683 assembler syntax for an instruction operand that is a memory reference
8684 whose address is @var{x}. @var{x} is an RTL expression.
8686 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8687 On some machines, the syntax for a symbolic address depends on the
8688 section that the address refers to. On these machines, define the hook
8689 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8690 @code{symbol_ref}, and then check for it here. @xref{Assembler
8694 @findex dbr_sequence_length
8695 @defmac DBR_OUTPUT_SEQEND (@var{file})
8696 A C statement, to be executed after all slot-filler instructions have
8697 been output. If necessary, call @code{dbr_sequence_length} to
8698 determine the number of slots filled in a sequence (zero if not
8699 currently outputting a sequence), to decide how many no-ops to output,
8702 Don't define this macro if it has nothing to do, but it is helpful in
8703 reading assembly output if the extent of the delay sequence is made
8704 explicit (e.g.@: with white space).
8707 @findex final_sequence
8708 Note that output routines for instructions with delay slots must be
8709 prepared to deal with not being output as part of a sequence
8710 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8711 found.) The variable @code{final_sequence} is null when not
8712 processing a sequence, otherwise it contains the @code{sequence} rtx
8716 @defmac REGISTER_PREFIX
8717 @defmacx LOCAL_LABEL_PREFIX
8718 @defmacx USER_LABEL_PREFIX
8719 @defmacx IMMEDIATE_PREFIX
8720 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8721 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8722 @file{final.c}). These are useful when a single @file{md} file must
8723 support multiple assembler formats. In that case, the various @file{tm.h}
8724 files can define these macros differently.
8727 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8728 If defined this macro should expand to a series of @code{case}
8729 statements which will be parsed inside the @code{switch} statement of
8730 the @code{asm_fprintf} function. This allows targets to define extra
8731 printf formats which may useful when generating their assembler
8732 statements. Note that uppercase letters are reserved for future
8733 generic extensions to asm_fprintf, and so are not available to target
8734 specific code. The output file is given by the parameter @var{file}.
8735 The varargs input pointer is @var{argptr} and the rest of the format
8736 string, starting the character after the one that is being switched
8737 upon, is pointed to by @var{format}.
8740 @defmac ASSEMBLER_DIALECT
8741 If your target supports multiple dialects of assembler language (such as
8742 different opcodes), define this macro as a C expression that gives the
8743 numeric index of the assembler language dialect to use, with zero as the
8746 If this macro is defined, you may use constructs of the form
8748 @samp{@{option0|option1|option2@dots{}@}}
8751 in the output templates of patterns (@pxref{Output Template}) or in the
8752 first argument of @code{asm_fprintf}. This construct outputs
8753 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8754 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8755 within these strings retain their usual meaning. If there are fewer
8756 alternatives within the braces than the value of
8757 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
8758 to print curly braces or @samp{|} character in assembler output directly,
8759 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
8761 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8762 @samp{@}} do not have any special meaning when used in templates or
8763 operands to @code{asm_fprintf}.
8765 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8766 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8767 the variations in assembler language syntax with that mechanism. Define
8768 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8769 if the syntax variant are larger and involve such things as different
8770 opcodes or operand order.
8773 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8774 A C expression to output to @var{stream} some assembler code
8775 which will push hard register number @var{regno} onto the stack.
8776 The code need not be optimal, since this macro is used only when
8780 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8781 A C expression to output to @var{stream} some assembler code
8782 which will pop hard register number @var{regno} off of the stack.
8783 The code need not be optimal, since this macro is used only when
8787 @node Dispatch Tables
8788 @subsection Output of Dispatch Tables
8790 @c prevent bad page break with this line
8791 This concerns dispatch tables.
8793 @cindex dispatch table
8794 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8795 A C statement to output to the stdio stream @var{stream} an assembler
8796 pseudo-instruction to generate a difference between two labels.
8797 @var{value} and @var{rel} are the numbers of two internal labels. The
8798 definitions of these labels are output using
8799 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8800 way here. For example,
8803 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8804 @var{value}, @var{rel})
8807 You must provide this macro on machines where the addresses in a
8808 dispatch table are relative to the table's own address. If defined, GCC
8809 will also use this macro on all machines when producing PIC@.
8810 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8811 mode and flags can be read.
8814 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8815 This macro should be provided on machines where the addresses
8816 in a dispatch table are absolute.
8818 The definition should be a C statement to output to the stdio stream
8819 @var{stream} an assembler pseudo-instruction to generate a reference to
8820 a label. @var{value} is the number of an internal label whose
8821 definition is output using @code{(*targetm.asm_out.internal_label)}.
8825 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8829 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8830 Define this if the label before a jump-table needs to be output
8831 specially. The first three arguments are the same as for
8832 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8833 jump-table which follows (a @code{jump_table_data} containing an
8834 @code{addr_vec} or @code{addr_diff_vec}).
8836 This feature is used on system V to output a @code{swbeg} statement
8839 If this macro is not defined, these labels are output with
8840 @code{(*targetm.asm_out.internal_label)}.
8843 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8844 Define this if something special must be output at the end of a
8845 jump-table. The definition should be a C statement to be executed
8846 after the assembler code for the table is written. It should write
8847 the appropriate code to stdio stream @var{stream}. The argument
8848 @var{table} is the jump-table insn, and @var{num} is the label-number
8849 of the preceding label.
8851 If this macro is not defined, nothing special is output at the end of
8855 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
8856 This target hook emits a label at the beginning of each FDE@. It
8857 should be defined on targets where FDEs need special labels, and it
8858 should write the appropriate label, for the FDE associated with the
8859 function declaration @var{decl}, to the stdio stream @var{stream}.
8860 The third argument, @var{for_eh}, is a boolean: true if this is for an
8861 exception table. The fourth argument, @var{empty}, is a boolean:
8862 true if this is a placeholder label for an omitted FDE@.
8864 The default is that FDEs are not given nonlocal labels.
8867 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
8868 This target hook emits a label at the beginning of the exception table.
8869 It should be defined on targets where it is desirable for the table
8870 to be broken up according to function.
8872 The default is that no label is emitted.
8875 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
8876 If the target implements @code{TARGET_ASM_UNWIND_EMIT}, this hook may be used to emit a directive to install a personality hook into the unwind info. This hook should not be used if dwarf2 unwind info is used.
8879 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx @var{insn})
8880 This target hook emits assembly directives required to unwind the
8881 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8882 returns @code{UI_TARGET}.
8885 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8886 True if the @code{TARGET_ASM_UNWIND_EMIT} hook should be called before the assembly for @var{insn} has been emitted, false if the hook should be called afterward.
8889 @node Exception Region Output
8890 @subsection Assembler Commands for Exception Regions
8892 @c prevent bad page break with this line
8894 This describes commands marking the start and the end of an exception
8897 @defmac EH_FRAME_SECTION_NAME
8898 If defined, a C string constant for the name of the section containing
8899 exception handling frame unwind information. If not defined, GCC will
8900 provide a default definition if the target supports named sections.
8901 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8903 You should define this symbol if your target supports DWARF 2 frame
8904 unwind information and the default definition does not work.
8907 @defmac EH_FRAME_IN_DATA_SECTION
8908 If defined, DWARF 2 frame unwind information will be placed in the
8909 data section even though the target supports named sections. This
8910 might be necessary, for instance, if the system linker does garbage
8911 collection and sections cannot be marked as not to be collected.
8913 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8917 @defmac EH_TABLES_CAN_BE_READ_ONLY
8918 Define this macro to 1 if your target is such that no frame unwind
8919 information encoding used with non-PIC code will ever require a
8920 runtime relocation, but the linker may not support merging read-only
8921 and read-write sections into a single read-write section.
8924 @defmac MASK_RETURN_ADDR
8925 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8926 that it does not contain any extraneous set bits in it.
8929 @defmac DWARF2_UNWIND_INFO
8930 Define this macro to 0 if your target supports DWARF 2 frame unwind
8931 information, but it does not yet work with exception handling.
8932 Otherwise, if your target supports this information (if it defines
8933 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
8934 GCC will provide a default definition of 1.
8937 @deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
8938 This hook defines the mechanism that will be used for exception handling
8939 by the target. If the target has ABI specified unwind tables, the hook
8940 should return @code{UI_TARGET}. If the target is to use the
8941 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8942 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8943 information, the hook should return @code{UI_DWARF2}.
8945 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8946 This may end up simplifying other parts of target-specific code. The
8947 default implementation of this hook never returns @code{UI_NONE}.
8949 Note that the value returned by this hook should be constant. It should
8950 not depend on anything except the command-line switches described by
8951 @var{opts}. In particular, the
8952 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8953 macros and builtin functions related to exception handling are set up
8954 depending on this setting.
8956 The default implementation of the hook first honors the
8957 @option{--enable-sjlj-exceptions} configure option, then
8958 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
8959 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
8960 must define this hook so that @var{opts} is used correctly.
8963 @deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8964 This variable should be set to @code{true} if the target ABI requires unwinding
8965 tables even when exceptions are not used. It must not be modified by
8966 command-line option processing.
8969 @defmac DONT_USE_BUILTIN_SETJMP
8970 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8971 should use the @code{setjmp}/@code{longjmp} functions from the C library
8972 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8975 @defmac JMP_BUF_SIZE
8976 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
8977 defined. Define this macro if the default size of @code{jmp_buf} buffer
8978 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
8979 is not large enough, or if it is much too large.
8980 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
8983 @defmac DWARF_CIE_DATA_ALIGNMENT
8984 This macro need only be defined if the target might save registers in the
8985 function prologue at an offset to the stack pointer that is not aligned to
8986 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8987 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8988 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8989 the target supports DWARF 2 frame unwind information.
8992 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8993 Contains the value true if the target should add a zero word onto the
8994 end of a Dwarf-2 frame info section when used for exception handling.
8995 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8999 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
9000 Given a register, this hook should return a parallel of registers to
9001 represent where to find the register pieces. Define this hook if the
9002 register and its mode are represented in Dwarf in non-contiguous
9003 locations, or if the register should be represented in more than one
9004 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
9005 If not defined, the default is to return @code{NULL_RTX}.
9008 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
9009 If some registers are represented in Dwarf-2 unwind information in
9010 multiple pieces, define this hook to fill in information about the
9011 sizes of those pieces in the table used by the unwinder at runtime.
9012 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
9013 filling in a single size corresponding to each hard register;
9014 @var{address} is the address of the table.
9017 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
9018 This hook is used to output a reference from a frame unwinding table to
9019 the type_info object identified by @var{sym}. It should return @code{true}
9020 if the reference was output. Returning @code{false} will cause the
9021 reference to be output using the normal Dwarf2 routines.
9024 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
9025 This flag should be set to @code{true} on targets that use an ARM EABI
9026 based unwinding library, and @code{false} on other targets. This effects
9027 the format of unwinding tables, and how the unwinder in entered after
9028 running a cleanup. The default is @code{false}.
9031 @node Alignment Output
9032 @subsection Assembler Commands for Alignment
9034 @c prevent bad page break with this line
9035 This describes commands for alignment.
9037 @defmac JUMP_ALIGN (@var{label})
9038 The alignment (log base 2) to put in front of @var{label}, which is
9039 a common destination of jumps and has no fallthru incoming edge.
9041 This macro need not be defined if you don't want any special alignment
9042 to be done at such a time. Most machine descriptions do not currently
9045 Unless it's necessary to inspect the @var{label} parameter, it is better
9046 to set the variable @var{align_jumps} in the target's
9047 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9048 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
9051 @deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx @var{label})
9052 The maximum number of bytes to skip before @var{label} when applying
9053 @code{JUMP_ALIGN}. This works only if
9054 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9057 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
9058 The alignment (log base 2) to put in front of @var{label}, which follows
9061 This macro need not be defined if you don't want any special alignment
9062 to be done at such a time. Most machine descriptions do not currently
9066 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx @var{label})
9067 The maximum number of bytes to skip before @var{label} when applying
9068 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
9069 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9072 @defmac LOOP_ALIGN (@var{label})
9073 The alignment (log base 2) to put in front of @var{label} that heads
9074 a frequently executed basic block (usually the header of a loop).
9076 This macro need not be defined if you don't want any special alignment
9077 to be done at such a time. Most machine descriptions do not currently
9080 Unless it's necessary to inspect the @var{label} parameter, it is better
9081 to set the variable @code{align_loops} in the target's
9082 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9083 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
9086 @deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx @var{label})
9087 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
9088 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
9092 @defmac LABEL_ALIGN (@var{label})
9093 The alignment (log base 2) to put in front of @var{label}.
9094 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
9095 the maximum of the specified values is used.
9097 Unless it's necessary to inspect the @var{label} parameter, it is better
9098 to set the variable @code{align_labels} in the target's
9099 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9100 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9103 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx @var{label})
9104 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
9105 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
9109 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9110 A C statement to output to the stdio stream @var{stream} an assembler
9111 instruction to advance the location counter by @var{nbytes} bytes.
9112 Those bytes should be zero when loaded. @var{nbytes} will be a C
9113 expression of type @code{unsigned HOST_WIDE_INT}.
9116 @defmac ASM_NO_SKIP_IN_TEXT
9117 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9118 text section because it fails to put zeros in the bytes that are skipped.
9119 This is true on many Unix systems, where the pseudo--op to skip bytes
9120 produces no-op instructions rather than zeros when used in the text
9124 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9125 A C statement to output to the stdio stream @var{stream} an assembler
9126 command to advance the location counter to a multiple of 2 to the
9127 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9130 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9131 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9132 for padding, if necessary.
9135 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9136 A C statement to output to the stdio stream @var{stream} an assembler
9137 command to advance the location counter to a multiple of 2 to the
9138 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9139 satisfy the alignment request. @var{power} and @var{max_skip} will be
9140 a C expression of type @code{int}.
9144 @node Debugging Info
9145 @section Controlling Debugging Information Format
9147 @c prevent bad page break with this line
9148 This describes how to specify debugging information.
9151 * All Debuggers:: Macros that affect all debugging formats uniformly.
9152 * DBX Options:: Macros enabling specific options in DBX format.
9153 * DBX Hooks:: Hook macros for varying DBX format.
9154 * File Names and DBX:: Macros controlling output of file names in DBX format.
9155 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9156 * VMS Debug:: Macros for VMS debug format.
9160 @subsection Macros Affecting All Debugging Formats
9162 @c prevent bad page break with this line
9163 These macros affect all debugging formats.
9165 @defmac DBX_REGISTER_NUMBER (@var{regno})
9166 A C expression that returns the DBX register number for the compiler
9167 register number @var{regno}. In the default macro provided, the value
9168 of this expression will be @var{regno} itself. But sometimes there are
9169 some registers that the compiler knows about and DBX does not, or vice
9170 versa. In such cases, some register may need to have one number in the
9171 compiler and another for DBX@.
9173 If two registers have consecutive numbers inside GCC, and they can be
9174 used as a pair to hold a multiword value, then they @emph{must} have
9175 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9176 Otherwise, debuggers will be unable to access such a pair, because they
9177 expect register pairs to be consecutive in their own numbering scheme.
9179 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9180 does not preserve register pairs, then what you must do instead is
9181 redefine the actual register numbering scheme.
9184 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9185 A C expression that returns the integer offset value for an automatic
9186 variable having address @var{x} (an RTL expression). The default
9187 computation assumes that @var{x} is based on the frame-pointer and
9188 gives the offset from the frame-pointer. This is required for targets
9189 that produce debugging output for DBX or COFF-style debugging output
9190 for SDB and allow the frame-pointer to be eliminated when the
9191 @option{-g} options is used.
9194 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9195 A C expression that returns the integer offset value for an argument
9196 having address @var{x} (an RTL expression). The nominal offset is
9200 @defmac PREFERRED_DEBUGGING_TYPE
9201 A C expression that returns the type of debugging output GCC should
9202 produce when the user specifies just @option{-g}. Define
9203 this if you have arranged for GCC to support more than one format of
9204 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9205 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9206 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9208 When the user specifies @option{-ggdb}, GCC normally also uses the
9209 value of this macro to select the debugging output format, but with two
9210 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9211 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9212 defined, GCC uses @code{DBX_DEBUG}.
9214 The value of this macro only affects the default debugging output; the
9215 user can always get a specific type of output by using @option{-gstabs},
9216 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9220 @subsection Specific Options for DBX Output
9222 @c prevent bad page break with this line
9223 These are specific options for DBX output.
9225 @defmac DBX_DEBUGGING_INFO
9226 Define this macro if GCC should produce debugging output for DBX
9227 in response to the @option{-g} option.
9230 @defmac XCOFF_DEBUGGING_INFO
9231 Define this macro if GCC should produce XCOFF format debugging output
9232 in response to the @option{-g} option. This is a variant of DBX format.
9235 @defmac DEFAULT_GDB_EXTENSIONS
9236 Define this macro to control whether GCC should by default generate
9237 GDB's extended version of DBX debugging information (assuming DBX-format
9238 debugging information is enabled at all). If you don't define the
9239 macro, the default is 1: always generate the extended information
9240 if there is any occasion to.
9243 @defmac DEBUG_SYMS_TEXT
9244 Define this macro if all @code{.stabs} commands should be output while
9245 in the text section.
9248 @defmac ASM_STABS_OP
9249 A C string constant, including spacing, naming the assembler pseudo op to
9250 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9251 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9252 applies only to DBX debugging information format.
9255 @defmac ASM_STABD_OP
9256 A C string constant, including spacing, naming the assembler pseudo op to
9257 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9258 value is the current location. If you don't define this macro,
9259 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9263 @defmac ASM_STABN_OP
9264 A C string constant, including spacing, naming the assembler pseudo op to
9265 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9266 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9267 macro applies only to DBX debugging information format.
9270 @defmac DBX_NO_XREFS
9271 Define this macro if DBX on your system does not support the construct
9272 @samp{xs@var{tagname}}. On some systems, this construct is used to
9273 describe a forward reference to a structure named @var{tagname}.
9274 On other systems, this construct is not supported at all.
9277 @defmac DBX_CONTIN_LENGTH
9278 A symbol name in DBX-format debugging information is normally
9279 continued (split into two separate @code{.stabs} directives) when it
9280 exceeds a certain length (by default, 80 characters). On some
9281 operating systems, DBX requires this splitting; on others, splitting
9282 must not be done. You can inhibit splitting by defining this macro
9283 with the value zero. You can override the default splitting-length by
9284 defining this macro as an expression for the length you desire.
9287 @defmac DBX_CONTIN_CHAR
9288 Normally continuation is indicated by adding a @samp{\} character to
9289 the end of a @code{.stabs} string when a continuation follows. To use
9290 a different character instead, define this macro as a character
9291 constant for the character you want to use. Do not define this macro
9292 if backslash is correct for your system.
9295 @defmac DBX_STATIC_STAB_DATA_SECTION
9296 Define this macro if it is necessary to go to the data section before
9297 outputting the @samp{.stabs} pseudo-op for a non-global static
9301 @defmac DBX_TYPE_DECL_STABS_CODE
9302 The value to use in the ``code'' field of the @code{.stabs} directive
9303 for a typedef. The default is @code{N_LSYM}.
9306 @defmac DBX_STATIC_CONST_VAR_CODE
9307 The value to use in the ``code'' field of the @code{.stabs} directive
9308 for a static variable located in the text section. DBX format does not
9309 provide any ``right'' way to do this. The default is @code{N_FUN}.
9312 @defmac DBX_REGPARM_STABS_CODE
9313 The value to use in the ``code'' field of the @code{.stabs} directive
9314 for a parameter passed in registers. DBX format does not provide any
9315 ``right'' way to do this. The default is @code{N_RSYM}.
9318 @defmac DBX_REGPARM_STABS_LETTER
9319 The letter to use in DBX symbol data to identify a symbol as a parameter
9320 passed in registers. DBX format does not customarily provide any way to
9321 do this. The default is @code{'P'}.
9324 @defmac DBX_FUNCTION_FIRST
9325 Define this macro if the DBX information for a function and its
9326 arguments should precede the assembler code for the function. Normally,
9327 in DBX format, the debugging information entirely follows the assembler
9331 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9332 Define this macro, with value 1, if the value of a symbol describing
9333 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9334 relative to the start of the enclosing function. Normally, GCC uses
9335 an absolute address.
9338 @defmac DBX_LINES_FUNCTION_RELATIVE
9339 Define this macro, with value 1, if the value of a symbol indicating
9340 the current line number (@code{N_SLINE}) should be relative to the
9341 start of the enclosing function. Normally, GCC uses an absolute address.
9344 @defmac DBX_USE_BINCL
9345 Define this macro if GCC should generate @code{N_BINCL} and
9346 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9347 macro also directs GCC to output a type number as a pair of a file
9348 number and a type number within the file. Normally, GCC does not
9349 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9350 number for a type number.
9354 @subsection Open-Ended Hooks for DBX Format
9356 @c prevent bad page break with this line
9357 These are hooks for DBX format.
9359 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9360 A C statement to output DBX debugging information before code for line
9361 number @var{line} of the current source file to the stdio stream
9362 @var{stream}. @var{counter} is the number of time the macro was
9363 invoked, including the current invocation; it is intended to generate
9364 unique labels in the assembly output.
9366 This macro should not be defined if the default output is correct, or
9367 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9370 @defmac NO_DBX_FUNCTION_END
9371 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9372 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9373 On those machines, define this macro to turn this feature off without
9374 disturbing the rest of the gdb extensions.
9377 @defmac NO_DBX_BNSYM_ENSYM
9378 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9379 extension construct. On those machines, define this macro to turn this
9380 feature off without disturbing the rest of the gdb extensions.
9383 @node File Names and DBX
9384 @subsection File Names in DBX Format
9386 @c prevent bad page break with this line
9387 This describes file names in DBX format.
9389 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9390 A C statement to output DBX debugging information to the stdio stream
9391 @var{stream}, which indicates that file @var{name} is the main source
9392 file---the file specified as the input file for compilation.
9393 This macro is called only once, at the beginning of compilation.
9395 This macro need not be defined if the standard form of output
9396 for DBX debugging information is appropriate.
9398 It may be necessary to refer to a label equal to the beginning of the
9399 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9400 to do so. If you do this, you must also set the variable
9401 @var{used_ltext_label_name} to @code{true}.
9404 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9405 Define this macro, with value 1, if GCC should not emit an indication
9406 of the current directory for compilation and current source language at
9407 the beginning of the file.
9410 @defmac NO_DBX_GCC_MARKER
9411 Define this macro, with value 1, if GCC should not emit an indication
9412 that this object file was compiled by GCC@. The default is to emit
9413 an @code{N_OPT} stab at the beginning of every source file, with
9414 @samp{gcc2_compiled.} for the string and value 0.
9417 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9418 A C statement to output DBX debugging information at the end of
9419 compilation of the main source file @var{name}. Output should be
9420 written to the stdio stream @var{stream}.
9422 If you don't define this macro, nothing special is output at the end
9423 of compilation, which is correct for most machines.
9426 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9427 Define this macro @emph{instead of} defining
9428 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9429 the end of compilation is an @code{N_SO} stab with an empty string,
9430 whose value is the highest absolute text address in the file.
9435 @subsection Macros for SDB and DWARF Output
9437 @c prevent bad page break with this line
9438 Here are macros for SDB and DWARF output.
9440 @defmac SDB_DEBUGGING_INFO
9441 Define this macro if GCC should produce COFF-style debugging output
9442 for SDB in response to the @option{-g} option.
9445 @defmac DWARF2_DEBUGGING_INFO
9446 Define this macro if GCC should produce dwarf version 2 format
9447 debugging output in response to the @option{-g} option.
9449 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
9450 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9451 be emitted for each function. Instead of an integer return the enum
9452 value for the @code{DW_CC_} tag.
9455 To support optional call frame debugging information, you must also
9456 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9457 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9458 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9459 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9462 @defmac DWARF2_FRAME_INFO
9463 Define this macro to a nonzero value if GCC should always output
9464 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9465 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9466 exceptions are enabled, GCC will output this information not matter
9467 how you define @code{DWARF2_FRAME_INFO}.
9470 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
9471 This hook defines the mechanism that will be used for describing frame
9472 unwind information to the debugger. Normally the hook will return
9473 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9474 return @code{UI_NONE} otherwise.
9476 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9477 is disabled in order to always output DWARF 2 frame information.
9479 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9480 This will suppress generation of the normal debug frame unwind information.
9483 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9484 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9485 line debug info sections. This will result in much more compact line number
9486 tables, and hence is desirable if it works.
9489 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
9490 True if the @code{.debug_pubtypes} and @code{.debug_pubnames} sections should be emitted. These sections are not used on most platforms, and in particular GDB does not use them.
9493 @deftypevr {Target Hook} bool TARGET_FORCE_AT_COMP_DIR
9494 True if the @code{DW_AT_comp_dir} attribute should be emitted for each compilation unit. This attribute is required for the darwin linker to emit debug information.
9497 @deftypevr {Target Hook} bool TARGET_DELAY_SCHED2
9498 True if sched2 is not to be run at its normal place. This usually means it will be run as part of machine-specific reorg.
9501 @deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK
9502 True if vartrack is not to be run at its normal place. This usually means it will be run as part of machine-specific reorg.
9505 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9506 A C statement to issue assembly directives that create a difference
9507 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9510 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9511 A C statement to issue assembly directives that create a difference
9512 between the two given labels in system defined units, e.g. instruction
9513 slots on IA64 VMS, using an integer of the given size.
9516 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9517 A C statement to issue assembly directives that create a
9518 section-relative reference to the given @var{label}, using an integer of the
9519 given @var{size}. The label is known to be defined in the given @var{section}.
9522 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9523 A C statement to issue assembly directives that create a self-relative
9524 reference to the given @var{label}, using an integer of the given @var{size}.
9527 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9528 A C statement to issue assembly directives that create a reference to
9529 the DWARF table identifier @var{label} from the current section. This
9530 is used on some systems to avoid garbage collecting a DWARF table which
9531 is referenced by a function.
9534 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
9535 If defined, this target hook is a function which outputs a DTP-relative
9536 reference to the given TLS symbol of the specified size.
9539 @defmac PUT_SDB_@dots{}
9540 Define these macros to override the assembler syntax for the special
9541 SDB assembler directives. See @file{sdbout.c} for a list of these
9542 macros and their arguments. If the standard syntax is used, you need
9543 not define them yourself.
9547 Some assemblers do not support a semicolon as a delimiter, even between
9548 SDB assembler directives. In that case, define this macro to be the
9549 delimiter to use (usually @samp{\n}). It is not necessary to define
9550 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9554 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9555 Define this macro to allow references to unknown structure,
9556 union, or enumeration tags to be emitted. Standard COFF does not
9557 allow handling of unknown references, MIPS ECOFF has support for
9561 @defmac SDB_ALLOW_FORWARD_REFERENCES
9562 Define this macro to allow references to structure, union, or
9563 enumeration tags that have not yet been seen to be handled. Some
9564 assemblers choke if forward tags are used, while some require it.
9567 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9568 A C statement to output SDB debugging information before code for line
9569 number @var{line} of the current source file to the stdio stream
9570 @var{stream}. The default is to emit an @code{.ln} directive.
9575 @subsection Macros for VMS Debug Format
9577 @c prevent bad page break with this line
9578 Here are macros for VMS debug format.
9580 @defmac VMS_DEBUGGING_INFO
9581 Define this macro if GCC should produce debugging output for VMS
9582 in response to the @option{-g} option. The default behavior for VMS
9583 is to generate minimal debug info for a traceback in the absence of
9584 @option{-g} unless explicitly overridden with @option{-g0}. This
9585 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9586 @code{TARGET_OPTION_OVERRIDE}.
9589 @node Floating Point
9590 @section Cross Compilation and Floating Point
9591 @cindex cross compilation and floating point
9592 @cindex floating point and cross compilation
9594 While all modern machines use twos-complement representation for integers,
9595 there are a variety of representations for floating point numbers. This
9596 means that in a cross-compiler the representation of floating point numbers
9597 in the compiled program may be different from that used in the machine
9598 doing the compilation.
9600 Because different representation systems may offer different amounts of
9601 range and precision, all floating point constants must be represented in
9602 the target machine's format. Therefore, the cross compiler cannot
9603 safely use the host machine's floating point arithmetic; it must emulate
9604 the target's arithmetic. To ensure consistency, GCC always uses
9605 emulation to work with floating point values, even when the host and
9606 target floating point formats are identical.
9608 The following macros are provided by @file{real.h} for the compiler to
9609 use. All parts of the compiler which generate or optimize
9610 floating-point calculations must use these macros. They may evaluate
9611 their operands more than once, so operands must not have side effects.
9613 @defmac REAL_VALUE_TYPE
9614 The C data type to be used to hold a floating point value in the target
9615 machine's format. Typically this is a @code{struct} containing an
9616 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9620 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9621 Compares for equality the two values, @var{x} and @var{y}. If the target
9622 floating point format supports negative zeroes and/or NaNs,
9623 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9624 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9627 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9628 Tests whether @var{x} is less than @var{y}.
9631 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9632 Truncates @var{x} to a signed integer, rounding toward zero.
9635 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9636 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9637 @var{x} is negative, returns zero.
9640 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9641 Converts @var{string} into a floating point number in the target machine's
9642 representation for mode @var{mode}. This routine can handle both
9643 decimal and hexadecimal floating point constants, using the syntax
9644 defined by the C language for both.
9647 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9648 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9651 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9652 Determines whether @var{x} represents infinity (positive or negative).
9655 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9656 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9659 @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})
9660 Calculates an arithmetic operation on the two floating point values
9661 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9664 The operation to be performed is specified by @var{code}. Only the
9665 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9666 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9668 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9669 target's floating point format cannot represent infinity, it will call
9670 @code{abort}. Callers should check for this situation first, using
9671 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9674 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9675 Returns the negative of the floating point value @var{x}.
9678 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9679 Returns the absolute value of @var{x}.
9682 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9683 Converts a floating point value @var{x} into a double-precision integer
9684 which is then stored into @var{low} and @var{high}. If the value is not
9685 integral, it is truncated.
9688 @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})
9689 Converts a double-precision integer found in @var{low} and @var{high},
9690 into a floating point value which is then stored into @var{x}. The
9691 value is truncated to fit in mode @var{mode}.
9694 @node Mode Switching
9695 @section Mode Switching Instructions
9696 @cindex mode switching
9697 The following macros control mode switching optimizations:
9699 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9700 Define this macro if the port needs extra instructions inserted for mode
9701 switching in an optimizing compilation.
9703 For an example, the SH4 can perform both single and double precision
9704 floating point operations, but to perform a single precision operation,
9705 the FPSCR PR bit has to be cleared, while for a double precision
9706 operation, this bit has to be set. Changing the PR bit requires a general
9707 purpose register as a scratch register, hence these FPSCR sets have to
9708 be inserted before reload, i.e.@: you can't put this into instruction emitting
9709 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9711 You can have multiple entities that are mode-switched, and select at run time
9712 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9713 return nonzero for any @var{entity} that needs mode-switching.
9714 If you define this macro, you also have to define
9715 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9716 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9717 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9721 @defmac NUM_MODES_FOR_MODE_SWITCHING
9722 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9723 initializer for an array of integers. Each initializer element
9724 N refers to an entity that needs mode switching, and specifies the number
9725 of different modes that might need to be set for this entity.
9726 The position of the initializer in the initializer---starting counting at
9727 zero---determines the integer that is used to refer to the mode-switched
9729 In macros that take mode arguments / yield a mode result, modes are
9730 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9731 switch is needed / supplied.
9734 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9735 @var{entity} is an integer specifying a mode-switched entity. If
9736 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9737 return an integer value not larger than the corresponding element in
9738 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9739 be switched into prior to the execution of @var{insn}.
9742 @defmac MODE_AFTER (@var{entity}, @var{mode}, @var{insn})
9743 @var{entity} is an integer specifying a mode-switched entity. If
9744 this macro is defined, it is evaluated for every @var{insn} during
9745 mode switching. It determines the mode that an insn results in (if
9746 different from the incoming mode).
9749 @defmac MODE_ENTRY (@var{entity})
9750 If this macro is defined, it is evaluated for every @var{entity} that needs
9751 mode switching. It should evaluate to an integer, which is a mode that
9752 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9753 is defined then @code{MODE_EXIT} must be defined.
9756 @defmac MODE_EXIT (@var{entity})
9757 If this macro is defined, it is evaluated for every @var{entity} that needs
9758 mode switching. It should evaluate to an integer, which is a mode that
9759 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9760 is defined then @code{MODE_ENTRY} must be defined.
9763 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9764 This macro specifies the order in which modes for @var{entity} are processed.
9765 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9766 lowest. The value of the macro should be an integer designating a mode
9767 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9768 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9769 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9772 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9773 Generate one or more insns to set @var{entity} to @var{mode}.
9774 @var{hard_reg_live} is the set of hard registers live at the point where
9775 the insn(s) are to be inserted.
9778 @node Target Attributes
9779 @section Defining target-specific uses of @code{__attribute__}
9780 @cindex target attributes
9781 @cindex machine attributes
9782 @cindex attributes, target-specific
9784 Target-specific attributes may be defined for functions, data and types.
9785 These are described using the following target hooks; they also need to
9786 be documented in @file{extend.texi}.
9788 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9789 If defined, this target hook points to an array of @samp{struct
9790 attribute_spec} (defined in @file{tree.h}) specifying the machine
9791 specific attributes for this target and some of the restrictions on the
9792 entities to which these attributes are applied and the arguments they
9796 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
9797 If defined, this target hook is a function which returns true if the
9798 machine-specific attribute named @var{name} expects an identifier
9799 given as its first argument to be passed on as a plain identifier, not
9800 subjected to name lookup. If this is not defined, the default is
9801 false for all machine-specific attributes.
9804 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
9805 If defined, this target hook is a function which returns zero if the attributes on
9806 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9807 and two if they are nearly compatible (which causes a warning to be
9808 generated). If this is not defined, machine-specific attributes are
9809 supposed always to be compatible.
9812 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9813 If defined, this target hook is a function which assigns default attributes to
9814 the newly defined @var{type}.
9817 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9818 Define this target hook if the merging of type attributes needs special
9819 handling. If defined, the result is a list of the combined
9820 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9821 that @code{comptypes} has already been called and returned 1. This
9822 function may call @code{merge_attributes} to handle machine-independent
9826 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9827 Define this target hook if the merging of decl attributes needs special
9828 handling. If defined, the result is a list of the combined
9829 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9830 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9831 when this is needed are when one attribute overrides another, or when an
9832 attribute is nullified by a subsequent definition. This function may
9833 call @code{merge_attributes} to handle machine-independent merging.
9835 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9836 If the only target-specific handling you require is @samp{dllimport}
9837 for Microsoft Windows targets, you should define the macro
9838 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9839 will then define a function called
9840 @code{merge_dllimport_decl_attributes} which can then be defined as
9841 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9842 add @code{handle_dll_attribute} in the attribute table for your port
9843 to perform initial processing of the @samp{dllimport} and
9844 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9845 @file{i386/i386.c}, for example.
9848 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
9849 @var{decl} is a variable or function with @code{__attribute__((dllimport))} specified. Use this hook if the target needs to add extra validation checks to @code{handle_dll_attribute}.
9852 @defmac TARGET_DECLSPEC
9853 Define this macro to a nonzero value if you want to treat
9854 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9855 default, this behavior is enabled only for targets that define
9856 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9857 of @code{__declspec} is via a built-in macro, but you should not rely
9858 on this implementation detail.
9861 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9862 Define this target hook if you want to be able to add attributes to a decl
9863 when it is being created. This is normally useful for back ends which
9864 wish to implement a pragma by using the attributes which correspond to
9865 the pragma's effect. The @var{node} argument is the decl which is being
9866 created. The @var{attr_ptr} argument is a pointer to the attribute list
9867 for this decl. The list itself should not be modified, since it may be
9868 shared with other decls, but attributes may be chained on the head of
9869 the list and @code{*@var{attr_ptr}} modified to point to the new
9870 attributes, or a copy of the list may be made if further changes are
9874 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
9876 This target hook returns @code{true} if it is OK to inline @var{fndecl}
9877 into the current function, despite its having target-specific
9878 attributes, @code{false} otherwise. By default, if a function has a
9879 target specific attribute attached to it, it will not be inlined.
9882 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9883 This hook is called to parse @code{attribute(target("..."))}, which
9884 allows setting target-specific options on individual functions.
9885 These function-specific options may differ
9886 from the options specified on the command line. The hook should return
9887 @code{true} if the options are valid.
9889 The hook should set the @code{DECL_FUNCTION_SPECIFIC_TARGET} field in
9890 the function declaration to hold a pointer to a target-specific
9891 @code{struct cl_target_option} structure.
9894 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr}, struct gcc_options *@var{opts})
9895 This hook is called to save any additional target-specific information
9896 in the @code{struct cl_target_option} structure for function-specific
9897 options from the @code{struct gcc_options} structure.
9898 @xref{Option file format}.
9901 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct gcc_options *@var{opts}, struct cl_target_option *@var{ptr})
9902 This hook is called to restore any additional target-specific
9903 information in the @code{struct cl_target_option} structure for
9904 function-specific options to the @code{struct gcc_options} structure.
9907 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
9908 This hook is called to print any additional target-specific
9909 information in the @code{struct cl_target_option} structure for
9910 function-specific options.
9913 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
9914 This target hook parses the options for @code{#pragma GCC target}, which
9915 sets the target-specific options for functions that occur later in the
9916 input stream. The options accepted should be the same as those handled by the
9917 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9920 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
9921 Sometimes certain combinations of command options do not make sense on
9922 a particular target machine. You can override the hook
9923 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9924 once just after all the command options have been parsed.
9926 Don't use this hook to turn on various extra optimizations for
9927 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9929 If you need to do something whenever the optimization level is
9930 changed via the optimize attribute or pragma, see
9931 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9934 @deftypefn {Target Hook} bool TARGET_OPTION_FUNCTION_VERSIONS (tree @var{decl1}, tree @var{decl2})
9935 This target hook returns @code{true} if @var{DECL1} and @var{DECL2} are
9936 versions of the same function. @var{DECL1} and @var{DECL2} are function
9937 versions if and only if they have the same function signature and
9938 different target specific attributes, that is, they are compiled for
9939 different target machines.
9942 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
9943 This target hook returns @code{false} if the @var{caller} function
9944 cannot inline @var{callee}, based on target specific information. By
9945 default, inlining is not allowed if the callee function has function
9946 specific target options and the caller does not use the same options.
9950 @section Emulating TLS
9951 @cindex Emulated TLS
9953 For targets whose psABI does not provide Thread Local Storage via
9954 specific relocations and instruction sequences, an emulation layer is
9955 used. A set of target hooks allows this emulation layer to be
9956 configured for the requirements of a particular target. For instance
9957 the psABI may in fact specify TLS support in terms of an emulation
9960 The emulation layer works by creating a control object for every TLS
9961 object. To access the TLS object, a lookup function is provided
9962 which, when given the address of the control object, will return the
9963 address of the current thread's instance of the TLS object.
9965 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9966 Contains the name of the helper function that uses a TLS control
9967 object to locate a TLS instance. The default causes libgcc's
9968 emulated TLS helper function to be used.
9971 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9972 Contains the name of the helper function that should be used at
9973 program startup to register TLS objects that are implicitly
9974 initialized to zero. If this is @code{NULL}, all TLS objects will
9975 have explicit initializers. The default causes libgcc's emulated TLS
9976 registration function to be used.
9979 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9980 Contains the name of the section in which TLS control variables should
9981 be placed. The default of @code{NULL} allows these to be placed in
9985 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9986 Contains the name of the section in which TLS initializers should be
9987 placed. The default of @code{NULL} allows these to be placed in any
9991 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9992 Contains the prefix to be prepended to TLS control variable names.
9993 The default of @code{NULL} uses a target-specific prefix.
9996 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9997 Contains the prefix to be prepended to TLS initializer objects. The
9998 default of @code{NULL} uses a target-specific prefix.
10001 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
10002 Specifies a function that generates the FIELD_DECLs for a TLS control
10003 object type. @var{type} is the RECORD_TYPE the fields are for and
10004 @var{name} should be filled with the structure tag, if the default of
10005 @code{__emutls_object} is unsuitable. The default creates a type suitable
10006 for libgcc's emulated TLS function.
10009 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
10010 Specifies a function that generates the CONSTRUCTOR to initialize a
10011 TLS control object. @var{var} is the TLS control object, @var{decl}
10012 is the TLS object and @var{tmpl_addr} is the address of the
10013 initializer. The default initializes libgcc's emulated TLS control object.
10016 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
10017 Specifies whether the alignment of TLS control variable objects is
10018 fixed and should not be increased as some backends may do to optimize
10019 single objects. The default is false.
10022 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
10023 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
10024 may be used to describe emulated TLS control objects.
10027 @node MIPS Coprocessors
10028 @section Defining coprocessor specifics for MIPS targets.
10029 @cindex MIPS coprocessor-definition macros
10031 The MIPS specification allows MIPS implementations to have as many as 4
10032 coprocessors, each with as many as 32 private registers. GCC supports
10033 accessing these registers and transferring values between the registers
10034 and memory using asm-ized variables. For example:
10037 register unsigned int cp0count asm ("c0r1");
10043 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
10044 names may be added as described below, or the default names may be
10045 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
10047 Coprocessor registers are assumed to be epilogue-used; sets to them will
10048 be preserved even if it does not appear that the register is used again
10049 later in the function.
10051 Another note: according to the MIPS spec, coprocessor 1 (if present) is
10052 the FPU@. One accesses COP1 registers through standard mips
10053 floating-point support; they are not included in this mechanism.
10055 There is one macro used in defining the MIPS coprocessor interface which
10056 you may want to override in subtargets; it is described below.
10059 @section Parameters for Precompiled Header Validity Checking
10060 @cindex parameters, precompiled headers
10062 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
10063 This hook returns a pointer to the data needed by
10064 @code{TARGET_PCH_VALID_P} and sets
10065 @samp{*@var{sz}} to the size of the data in bytes.
10068 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
10069 This hook checks whether the options used to create a PCH file are
10070 compatible with the current settings. It returns @code{NULL}
10071 if so and a suitable error message if not. Error messages will
10072 be presented to the user and must be localized using @samp{_(@var{msg})}.
10074 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
10075 when the PCH file was created and @var{sz} is the size of that data in bytes.
10076 It's safe to assume that the data was created by the same version of the
10077 compiler, so no format checking is needed.
10079 The default definition of @code{default_pch_valid_p} should be
10080 suitable for most targets.
10083 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
10084 If this hook is nonnull, the default implementation of
10085 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
10086 of @code{target_flags}. @var{pch_flags} specifies the value that
10087 @code{target_flags} had when the PCH file was created. The return
10088 value is the same as for @code{TARGET_PCH_VALID_P}.
10091 @deftypefn {Target Hook} void TARGET_PREPARE_PCH_SAVE (void)
10092 Called before writing out a PCH file. If the target has some
10093 garbage-collected data that needs to be in a particular state on PCH loads,
10094 it can use this hook to enforce that state. Very few targets need
10095 to do anything here.
10099 @section C++ ABI parameters
10100 @cindex parameters, c++ abi
10102 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10103 Define this hook to override the integer type used for guard variables.
10104 These are used to implement one-time construction of static objects. The
10105 default is long_long_integer_type_node.
10108 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10109 This hook determines how guard variables are used. It should return
10110 @code{false} (the default) if the first byte should be used. A return value of
10111 @code{true} indicates that only the least significant bit should be used.
10114 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10115 This hook returns the size of the cookie to use when allocating an array
10116 whose elements have the indicated @var{type}. Assumes that it is already
10117 known that a cookie is needed. The default is
10118 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10119 IA64/Generic C++ ABI@.
10122 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10123 This hook should return @code{true} if the element size should be stored in
10124 array cookies. The default is to return @code{false}.
10127 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10128 If defined by a backend this hook allows the decision made to export
10129 class @var{type} to be overruled. Upon entry @var{import_export}
10130 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10131 to be imported and 0 otherwise. This function should return the
10132 modified value and perform any other actions necessary to support the
10133 backend's targeted operating system.
10136 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10137 This hook should return @code{true} if constructors and destructors return
10138 the address of the object created/destroyed. The default is to return
10142 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10143 This hook returns true if the key method for a class (i.e., the method
10144 which, if defined in the current translation unit, causes the virtual
10145 table to be emitted) may be an inline function. Under the standard
10146 Itanium C++ ABI the key method may be an inline function so long as
10147 the function is not declared inline in the class definition. Under
10148 some variants of the ABI, an inline function can never be the key
10149 method. The default is to return @code{true}.
10152 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10153 @var{decl} is a virtual table, virtual table table, typeinfo object, or other similar implicit class data object that will be emitted with external linkage in this translation unit. No ELF visibility has been explicitly specified. If the target needs to specify a visibility other than that of the containing class, use this hook to set @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
10156 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10157 This hook returns true (the default) if virtual tables and other
10158 similar implicit class data objects are always COMDAT if they have
10159 external linkage. If this hook returns false, then class data for
10160 classes whose virtual table will be emitted in only one translation
10161 unit will not be COMDAT.
10164 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10165 This hook returns true (the default) if the RTTI information for
10166 the basic types which is defined in the C++ runtime should always
10167 be COMDAT, false if it should not be COMDAT.
10170 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10171 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10172 should be used to register static destructors when @option{-fuse-cxa-atexit}
10173 is in effect. The default is to return false to use @code{__cxa_atexit}.
10176 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10177 This hook returns true if the target @code{atexit} function can be used
10178 in the same manner as @code{__cxa_atexit} to register C++ static
10179 destructors. This requires that @code{atexit}-registered functions in
10180 shared libraries are run in the correct order when the libraries are
10181 unloaded. The default is to return false.
10184 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10185 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been defined. Use this hook to make adjustments to the class (eg, tweak visibility or perform any other required target modifications).
10188 @deftypefn {Target Hook} tree TARGET_CXX_DECL_MANGLING_CONTEXT (const_tree @var{decl})
10189 Return target-specific mangling context of @var{decl} or @code{NULL_TREE}.
10192 @node Named Address Spaces
10193 @section Adding support for named address spaces
10194 @cindex named address spaces
10196 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10197 standards committee, @cite{Programming Languages - C - Extensions to
10198 support embedded processors}, specifies a syntax for embedded
10199 processors to specify alternate address spaces. You can configure a
10200 GCC port to support section 5.1 of the draft report to add support for
10201 address spaces other than the default address space. These address
10202 spaces are new keywords that are similar to the @code{volatile} and
10203 @code{const} type attributes.
10205 Pointers to named address spaces can have a different size than
10206 pointers to the generic address space.
10208 For example, the SPU port uses the @code{__ea} address space to refer
10209 to memory in the host processor, rather than memory local to the SPU
10210 processor. Access to memory in the @code{__ea} address space involves
10211 issuing DMA operations to move data between the host processor and the
10212 local processor memory address space. Pointers in the @code{__ea}
10213 address space are either 32 bits or 64 bits based on the
10214 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10217 Internally, address spaces are represented as a small integer in the
10218 range 0 to 15 with address space 0 being reserved for the generic
10221 To register a named address space qualifier keyword with the C front end,
10222 the target may call the @code{c_register_addr_space} routine. For example,
10223 the SPU port uses the following to declare @code{__ea} as the keyword for
10224 named address space #1:
10226 #define ADDR_SPACE_EA 1
10227 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10230 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10231 Define this to return the machine mode to use for pointers to
10232 @var{address_space} if the target supports named address spaces.
10233 The default version of this hook returns @code{ptr_mode} for the
10234 generic address space only.
10237 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10238 Define this to return the machine mode to use for addresses in
10239 @var{address_space} if the target supports named address spaces.
10240 The default version of this hook returns @code{Pmode} for the
10241 generic address space only.
10244 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (enum machine_mode @var{mode}, addr_space_t @var{as})
10245 Define this to return nonzero if the port can handle pointers
10246 with machine mode @var{mode} to address space @var{as}. This target
10247 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10248 except that it includes explicit named address space support. The default
10249 version of this hook returns true for the modes returned by either the
10250 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10251 target hooks for the given address space.
10254 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{exp}, bool @var{strict}, addr_space_t @var{as})
10255 Define this to return true if @var{exp} is a valid address for mode
10256 @var{mode} in the named address space @var{as}. The @var{strict}
10257 parameter says whether strict addressing is in effect after reload has
10258 finished. This target hook is the same as the
10259 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10260 explicit named address space support.
10263 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode}, addr_space_t @var{as})
10264 Define this to modify an invalid address @var{x} to be a valid address
10265 with mode @var{mode} in the named address space @var{as}. This target
10266 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10267 except that it includes explicit named address space support.
10270 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{subset}, addr_space_t @var{superset})
10271 Define this to return whether the @var{subset} named address space is
10272 contained within the @var{superset} named address space. Pointers to
10273 a named address space that is a subset of another named address space
10274 will be converted automatically without a cast if used together in
10275 arithmetic operations. Pointers to a superset address space can be
10276 converted to pointers to a subset address space via explicit casts.
10279 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10280 Define this to convert the pointer expression represented by the RTL
10281 @var{op} with type @var{from_type} that points to a named address
10282 space to a new pointer expression with type @var{to_type} that points
10283 to a different named address space. When this hook it called, it is
10284 guaranteed that one of the two address spaces is a subset of the other,
10285 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10289 @section Miscellaneous Parameters
10290 @cindex parameters, miscellaneous
10292 @c prevent bad page break with this line
10293 Here are several miscellaneous parameters.
10295 @defmac HAS_LONG_COND_BRANCH
10296 Define this boolean macro to indicate whether or not your architecture
10297 has conditional branches that can span all of memory. It is used in
10298 conjunction with an optimization that partitions hot and cold basic
10299 blocks into separate sections of the executable. If this macro is
10300 set to false, gcc will convert any conditional branches that attempt
10301 to cross between sections into unconditional branches or indirect jumps.
10304 @defmac HAS_LONG_UNCOND_BRANCH
10305 Define this boolean macro to indicate whether or not your architecture
10306 has unconditional branches that can span all of memory. It is used in
10307 conjunction with an optimization that partitions hot and cold basic
10308 blocks into separate sections of the executable. If this macro is
10309 set to false, gcc will convert any unconditional branches that attempt
10310 to cross between sections into indirect jumps.
10313 @defmac CASE_VECTOR_MODE
10314 An alias for a machine mode name. This is the machine mode that
10315 elements of a jump-table should have.
10318 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10319 Optional: return the preferred mode for an @code{addr_diff_vec}
10320 when the minimum and maximum offset are known. If you define this,
10321 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10322 To make this work, you also have to define @code{INSN_ALIGN} and
10323 make the alignment for @code{addr_diff_vec} explicit.
10324 The @var{body} argument is provided so that the offset_unsigned and scale
10325 flags can be updated.
10328 @defmac CASE_VECTOR_PC_RELATIVE
10329 Define this macro to be a C expression to indicate when jump-tables
10330 should contain relative addresses. You need not define this macro if
10331 jump-tables never contain relative addresses, or jump-tables should
10332 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10336 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10337 This function return the smallest number of different values for which it
10338 is best to use a jump-table instead of a tree of conditional branches.
10339 The default is four for machines with a @code{casesi} instruction and
10340 five otherwise. This is best for most machines.
10343 @defmac WORD_REGISTER_OPERATIONS
10344 Define this macro if operations between registers with integral mode
10345 smaller than a word are always performed on the entire register.
10346 Most RISC machines have this property and most CISC machines do not.
10349 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10350 Define this macro to be a C expression indicating when insns that read
10351 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10352 bits outside of @var{mem_mode} to be either the sign-extension or the
10353 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10354 of @var{mem_mode} for which the
10355 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10356 @code{UNKNOWN} for other modes.
10358 This macro is not called with @var{mem_mode} non-integral or with a width
10359 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10360 value in this case. Do not define this macro if it would always return
10361 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10362 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10364 You may return a non-@code{UNKNOWN} value even if for some hard registers
10365 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10366 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10367 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10368 integral mode larger than this but not larger than @code{word_mode}.
10370 You must return @code{UNKNOWN} if for some hard registers that allow this
10371 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10372 @code{word_mode}, but that they can change to another integral mode that
10373 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10376 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10377 Define this macro if loading short immediate values into registers sign
10381 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
10382 When @option{-ffast-math} is in effect, GCC tries to optimize
10383 divisions by the same divisor, by turning them into multiplications by
10384 the reciprocal. This target hook specifies the minimum number of divisions
10385 that should be there for GCC to perform the optimization for a variable
10386 of mode @var{mode}. The default implementation returns 3 if the machine
10387 has an instruction for the division, and 2 if it does not.
10391 The maximum number of bytes that a single instruction can move quickly
10392 between memory and registers or between two memory locations.
10395 @defmac MAX_MOVE_MAX
10396 The maximum number of bytes that a single instruction can move quickly
10397 between memory and registers or between two memory locations. If this
10398 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10399 constant value that is the largest value that @code{MOVE_MAX} can have
10403 @defmac SHIFT_COUNT_TRUNCATED
10404 A C expression that is nonzero if on this machine the number of bits
10405 actually used for the count of a shift operation is equal to the number
10406 of bits needed to represent the size of the object being shifted. When
10407 this macro is nonzero, the compiler will assume that it is safe to omit
10408 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10409 truncates the count of a shift operation. On machines that have
10410 instructions that act on bit-fields at variable positions, which may
10411 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10412 also enables deletion of truncations of the values that serve as
10413 arguments to bit-field instructions.
10415 If both types of instructions truncate the count (for shifts) and
10416 position (for bit-field operations), or if no variable-position bit-field
10417 instructions exist, you should define this macro.
10419 However, on some machines, such as the 80386 and the 680x0, truncation
10420 only applies to shift operations and not the (real or pretended)
10421 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10422 such machines. Instead, add patterns to the @file{md} file that include
10423 the implied truncation of the shift instructions.
10425 You need not define this macro if it would always have the value of zero.
10428 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10429 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
10430 This function describes how the standard shift patterns for @var{mode}
10431 deal with shifts by negative amounts or by more than the width of the mode.
10432 @xref{shift patterns}.
10434 On many machines, the shift patterns will apply a mask @var{m} to the
10435 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10436 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10437 this is true for mode @var{mode}, the function should return @var{m},
10438 otherwise it should return 0. A return value of 0 indicates that no
10439 particular behavior is guaranteed.
10441 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10442 @emph{not} apply to general shift rtxes; it applies only to instructions
10443 that are generated by the named shift patterns.
10445 The default implementation of this function returns
10446 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10447 and 0 otherwise. This definition is always safe, but if
10448 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10449 nevertheless truncate the shift count, you may get better code
10453 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10454 A C expression which is nonzero if on this machine it is safe to
10455 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10456 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10457 operating on it as if it had only @var{outprec} bits.
10459 On many machines, this expression can be 1.
10461 @c rearranged this, removed the phrase "it is reported that". this was
10462 @c to fix an overfull hbox. --mew 10feb93
10463 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10464 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10465 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10466 such cases may improve things.
10469 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
10470 The representation of an integral mode can be such that the values
10471 are always extended to a wider integral mode. Return
10472 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10473 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10474 otherwise. (Currently, none of the targets use zero-extended
10475 representation this way so unlike @code{LOAD_EXTEND_OP},
10476 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10477 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10478 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10479 widest integral mode and currently we take advantage of this fact.)
10481 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10482 value even if the extension is not performed on certain hard registers
10483 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10484 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10486 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10487 describe two related properties. If you define
10488 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10489 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10492 In order to enforce the representation of @code{mode},
10493 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10497 @defmac STORE_FLAG_VALUE
10498 A C expression describing the value returned by a comparison operator
10499 with an integral mode and stored by a store-flag instruction
10500 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10501 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10502 comparison operators whose results have a @code{MODE_INT} mode.
10504 A value of 1 or @minus{}1 means that the instruction implementing the
10505 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10506 and 0 when the comparison is false. Otherwise, the value indicates
10507 which bits of the result are guaranteed to be 1 when the comparison is
10508 true. This value is interpreted in the mode of the comparison
10509 operation, which is given by the mode of the first operand in the
10510 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10511 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10514 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10515 generate code that depends only on the specified bits. It can also
10516 replace comparison operators with equivalent operations if they cause
10517 the required bits to be set, even if the remaining bits are undefined.
10518 For example, on a machine whose comparison operators return an
10519 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10520 @samp{0x80000000}, saying that just the sign bit is relevant, the
10524 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10528 can be converted to
10531 (ashift:SI @var{x} (const_int @var{n}))
10535 where @var{n} is the appropriate shift count to move the bit being
10536 tested into the sign bit.
10538 There is no way to describe a machine that always sets the low-order bit
10539 for a true value, but does not guarantee the value of any other bits,
10540 but we do not know of any machine that has such an instruction. If you
10541 are trying to port GCC to such a machine, include an instruction to
10542 perform a logical-and of the result with 1 in the pattern for the
10543 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10545 Often, a machine will have multiple instructions that obtain a value
10546 from a comparison (or the condition codes). Here are rules to guide the
10547 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10552 Use the shortest sequence that yields a valid definition for
10553 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10554 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10555 comparison operators to do so because there may be opportunities to
10556 combine the normalization with other operations.
10559 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10560 slightly preferred on machines with expensive jumps and 1 preferred on
10564 As a second choice, choose a value of @samp{0x80000001} if instructions
10565 exist that set both the sign and low-order bits but do not define the
10569 Otherwise, use a value of @samp{0x80000000}.
10572 Many machines can produce both the value chosen for
10573 @code{STORE_FLAG_VALUE} and its negation in the same number of
10574 instructions. On those machines, you should also define a pattern for
10575 those cases, e.g., one matching
10578 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10581 Some machines can also perform @code{and} or @code{plus} operations on
10582 condition code values with less instructions than the corresponding
10583 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10584 machines, define the appropriate patterns. Use the names @code{incscc}
10585 and @code{decscc}, respectively, for the patterns which perform
10586 @code{plus} or @code{minus} operations on condition code values. See
10587 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10588 find such instruction sequences on other machines.
10590 If this macro is not defined, the default value, 1, is used. You need
10591 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10592 instructions, or if the value generated by these instructions is 1.
10595 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10596 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10597 returned when comparison operators with floating-point results are true.
10598 Define this macro on machines that have comparison operations that return
10599 floating-point values. If there are no such operations, do not define
10603 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10604 A C expression that gives a rtx representing the nonzero true element
10605 for vector comparisons. The returned rtx should be valid for the inner
10606 mode of @var{mode} which is guaranteed to be a vector mode. Define
10607 this macro on machines that have vector comparison operations that
10608 return a vector result. If there are no such operations, do not define
10609 this macro. Typically, this macro is defined as @code{const1_rtx} or
10610 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10611 the compiler optimizing such vector comparison operations for the
10615 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10616 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10617 A C expression that indicates whether the architecture defines a value
10618 for @code{clz} or @code{ctz} with a zero operand.
10619 A result of @code{0} indicates the value is undefined.
10620 If the value is defined for only the RTL expression, the macro should
10621 evaluate to @code{1}; if the value applies also to the corresponding optab
10622 entry (which is normally the case if it expands directly into
10623 the corresponding RTL), then the macro should evaluate to @code{2}.
10624 In the cases where the value is defined, @var{value} should be set to
10627 If this macro is not defined, the value of @code{clz} or
10628 @code{ctz} at zero is assumed to be undefined.
10630 This macro must be defined if the target's expansion for @code{ffs}
10631 relies on a particular value to get correct results. Otherwise it
10632 is not necessary, though it may be used to optimize some corner cases, and
10633 to provide a default expansion for the @code{ffs} optab.
10635 Note that regardless of this macro the ``definedness'' of @code{clz}
10636 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10637 visible to the user. Thus one may be free to adjust the value at will
10638 to match the target expansion of these operations without fear of
10643 An alias for the machine mode for pointers. On most machines, define
10644 this to be the integer mode corresponding to the width of a hardware
10645 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10646 On some machines you must define this to be one of the partial integer
10647 modes, such as @code{PSImode}.
10649 The width of @code{Pmode} must be at least as large as the value of
10650 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10651 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10655 @defmac FUNCTION_MODE
10656 An alias for the machine mode used for memory references to functions
10657 being called, in @code{call} RTL expressions. On most CISC machines,
10658 where an instruction can begin at any byte address, this should be
10659 @code{QImode}. On most RISC machines, where all instructions have fixed
10660 size and alignment, this should be a mode with the same size and alignment
10661 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10664 @defmac STDC_0_IN_SYSTEM_HEADERS
10665 In normal operation, the preprocessor expands @code{__STDC__} to the
10666 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10667 hosts, like Solaris, the system compiler uses a different convention,
10668 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10669 strict conformance to the C Standard.
10671 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10672 convention when processing system header files, but when processing user
10673 files @code{__STDC__} will always expand to 1.
10676 @deftypefn {C Target Hook} {const char *} TARGET_C_PREINCLUDE (void)
10677 Define this hook to return the name of a header file to be included at the start of all compilations, as if it had been included with @code{#include <@var{file}>}. If this hook returns @code{NULL}, or is not defined, or the header is not found, or if the user specifies @option{-ffreestanding} or @option{-nostdinc}, no header is included.
10679 This hook can be used together with a header provided by the system C library to implement ISO C requirements for certain macros to be predefined that describe properties of the whole implementation rather than just the compiler.
10682 @deftypefn {C Target Hook} bool TARGET_CXX_IMPLICIT_EXTERN_C (const char*@var{})
10683 Define this hook to add target-specific C++ implicit extern C functions. If this function returns true for the name of a file-scope function, that function implicitly gets extern "C" linkage rather than whatever language linkage the declaration would normally have. An example of such function is WinMain on Win32 targets.
10686 @defmac NO_IMPLICIT_EXTERN_C
10687 Define this macro if the system header files support C++ as well as C@.
10688 This macro inhibits the usual method of using system header files in
10689 C++, which is to pretend that the file's contents are enclosed in
10690 @samp{extern "C" @{@dots{}@}}.
10695 @defmac REGISTER_TARGET_PRAGMAS ()
10696 Define this macro if you want to implement any target-specific pragmas.
10697 If defined, it is a C expression which makes a series of calls to
10698 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10699 for each pragma. The macro may also do any
10700 setup required for the pragmas.
10702 The primary reason to define this macro is to provide compatibility with
10703 other compilers for the same target. In general, we discourage
10704 definition of target-specific pragmas for GCC@.
10706 If the pragma can be implemented by attributes then you should consider
10707 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10709 Preprocessor macros that appear on pragma lines are not expanded. All
10710 @samp{#pragma} directives that do not match any registered pragma are
10711 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10714 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10715 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10717 Each call to @code{c_register_pragma} or
10718 @code{c_register_pragma_with_expansion} establishes one pragma. The
10719 @var{callback} routine will be called when the preprocessor encounters a
10723 #pragma [@var{space}] @var{name} @dots{}
10726 @var{space} is the case-sensitive namespace of the pragma, or
10727 @code{NULL} to put the pragma in the global namespace. The callback
10728 routine receives @var{pfile} as its first argument, which can be passed
10729 on to cpplib's functions if necessary. You can lex tokens after the
10730 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10731 callback will be silently ignored. The end of the line is indicated by
10732 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10733 arguments of pragmas registered with
10734 @code{c_register_pragma_with_expansion} but not on the arguments of
10735 pragmas registered with @code{c_register_pragma}.
10737 Note that the use of @code{pragma_lex} is specific to the C and C++
10738 compilers. It will not work in the Java or Fortran compilers, or any
10739 other language compilers for that matter. Thus if @code{pragma_lex} is going
10740 to be called from target-specific code, it must only be done so when
10741 building the C and C++ compilers. This can be done by defining the
10742 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10743 target entry in the @file{config.gcc} file. These variables should name
10744 the target-specific, language-specific object file which contains the
10745 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10746 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10747 how to build this object file.
10750 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10751 Define this macro if macros should be expanded in the
10752 arguments of @samp{#pragma pack}.
10755 @defmac TARGET_DEFAULT_PACK_STRUCT
10756 If your target requires a structure packing default other than 0 (meaning
10757 the machine default), define this macro to the necessary value (in bytes).
10758 This must be a value that would also be valid to use with
10759 @samp{#pragma pack()} (that is, a small power of two).
10762 @defmac DOLLARS_IN_IDENTIFIERS
10763 Define this macro to control use of the character @samp{$} in
10764 identifier names for the C family of languages. 0 means @samp{$} is
10765 not allowed by default; 1 means it is allowed. 1 is the default;
10766 there is no need to define this macro in that case.
10769 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10770 Define this macro as a C expression that is nonzero if it is safe for the
10771 delay slot scheduler to place instructions in the delay slot of @var{insn},
10772 even if they appear to use a resource set or clobbered in @var{insn}.
10773 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10774 every @code{call_insn} has this behavior. On machines where some @code{insn}
10775 or @code{jump_insn} is really a function call and hence has this behavior,
10776 you should define this macro.
10778 You need not define this macro if it would always return zero.
10781 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10782 Define this macro as a C expression that is nonzero if it is safe for the
10783 delay slot scheduler to place instructions in the delay slot of @var{insn},
10784 even if they appear to set or clobber a resource referenced in @var{insn}.
10785 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10786 some @code{insn} or @code{jump_insn} is really a function call and its operands
10787 are registers whose use is actually in the subroutine it calls, you should
10788 define this macro. Doing so allows the delay slot scheduler to move
10789 instructions which copy arguments into the argument registers into the delay
10790 slot of @var{insn}.
10792 You need not define this macro if it would always return zero.
10795 @defmac MULTIPLE_SYMBOL_SPACES
10796 Define this macro as a C expression that is nonzero if, in some cases,
10797 global symbols from one translation unit may not be bound to undefined
10798 symbols in another translation unit without user intervention. For
10799 instance, under Microsoft Windows symbols must be explicitly imported
10800 from shared libraries (DLLs).
10802 You need not define this macro if it would always evaluate to zero.
10805 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10806 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10807 any hard regs the port wishes to automatically clobber for an asm.
10808 It should return the result of the last @code{tree_cons} used to add a
10809 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10810 corresponding parameters to the asm and may be inspected to avoid
10811 clobbering a register that is an input or output of the asm. You can use
10812 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10813 for overlap with regards to asm-declared registers.
10816 @defmac MATH_LIBRARY
10817 Define this macro as a C string constant for the linker argument to link
10818 in the system math library, minus the initial @samp{"-l"}, or
10819 @samp{""} if the target does not have a
10820 separate math library.
10822 You need only define this macro if the default of @samp{"m"} is wrong.
10825 @defmac LIBRARY_PATH_ENV
10826 Define this macro as a C string constant for the environment variable that
10827 specifies where the linker should look for libraries.
10829 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10833 @defmac TARGET_POSIX_IO
10834 Define this macro if the target supports the following POSIX@ file
10835 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10836 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10837 to use file locking when exiting a program, which avoids race conditions
10838 if the program has forked. It will also create directories at run-time
10839 for cross-profiling.
10842 @defmac MAX_CONDITIONAL_EXECUTE
10844 A C expression for the maximum number of instructions to execute via
10845 conditional execution instructions instead of a branch. A value of
10846 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10847 1 if it does use cc0.
10850 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10851 Used if the target needs to perform machine-dependent modifications on the
10852 conditionals used for turning basic blocks into conditionally executed code.
10853 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10854 contains information about the currently processed blocks. @var{true_expr}
10855 and @var{false_expr} are the tests that are used for converting the
10856 then-block and the else-block, respectively. Set either @var{true_expr} or
10857 @var{false_expr} to a null pointer if the tests cannot be converted.
10860 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10861 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10862 if-statements into conditions combined by @code{and} and @code{or} operations.
10863 @var{bb} contains the basic block that contains the test that is currently
10864 being processed and about to be turned into a condition.
10867 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10868 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10869 be converted to conditional execution format. @var{ce_info} points to
10870 a data structure, @code{struct ce_if_block}, which contains information
10871 about the currently processed blocks.
10874 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10875 A C expression to perform any final machine dependent modifications in
10876 converting code to conditional execution. The involved basic blocks
10877 can be found in the @code{struct ce_if_block} structure that is pointed
10878 to by @var{ce_info}.
10881 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10882 A C expression to cancel any machine dependent modifications in
10883 converting code to conditional execution. The involved basic blocks
10884 can be found in the @code{struct ce_if_block} structure that is pointed
10885 to by @var{ce_info}.
10888 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
10889 A C expression to initialize any machine specific data for if-conversion
10890 of the if-block in the @code{struct ce_if_block} structure that is pointed
10891 to by @var{ce_info}.
10894 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
10895 If non-null, this hook performs a target-specific pass over the
10896 instruction stream. The compiler will run it at all optimization levels,
10897 just before the point at which it normally does delayed-branch scheduling.
10899 The exact purpose of the hook varies from target to target. Some use
10900 it to do transformations that are necessary for correctness, such as
10901 laying out in-function constant pools or avoiding hardware hazards.
10902 Others use it as an opportunity to do some machine-dependent optimizations.
10904 You need not implement the hook if it has nothing to do. The default
10905 definition is null.
10908 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
10909 Define this hook if you have any machine-specific built-in functions
10910 that need to be defined. It should be a function that performs the
10913 Machine specific built-in functions can be useful to expand special machine
10914 instructions that would otherwise not normally be generated because
10915 they have no equivalent in the source language (for example, SIMD vector
10916 instructions or prefetch instructions).
10918 To create a built-in function, call the function
10919 @code{lang_hooks.builtin_function}
10920 which is defined by the language front end. You can use any type nodes set
10921 up by @code{build_common_tree_nodes};
10922 only language front ends that use those two functions will call
10923 @samp{TARGET_INIT_BUILTINS}.
10926 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
10927 Define this hook if you have any machine-specific built-in functions
10928 that need to be defined. It should be a function that returns the
10929 builtin function declaration for the builtin function code @var{code}.
10930 If there is no such builtin and it cannot be initialized at this time
10931 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10932 If @var{code} is out of range the function should return
10933 @code{error_mark_node}.
10936 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10938 Expand a call to a machine specific built-in function that was set up by
10939 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10940 function call; the result should go to @var{target} if that is
10941 convenient, and have mode @var{mode} if that is convenient.
10942 @var{subtarget} may be used as the target for computing one of
10943 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10944 ignored. This function should return the result of the call to the
10948 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
10949 Select a replacement for a machine specific built-in function that
10950 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10951 @emph{before} regular type checking, and so allows the target to
10952 implement a crude form of function overloading. @var{fndecl} is the
10953 declaration of the built-in function. @var{arglist} is the list of
10954 arguments passed to the built-in function. The result is a
10955 complete expression that implements the operation, usually
10956 another @code{CALL_EXPR}.
10957 @var{arglist} really has type @samp{VEC(tree,gc)*}
10960 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
10961 Fold a call to a machine specific built-in function that was set up by
10962 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10963 built-in function. @var{n_args} is the number of arguments passed to
10964 the function; the arguments themselves are pointed to by @var{argp}.
10965 The result is another tree, valid for both GIMPLE and GENERIC,
10966 containing a simplified expression for the call's result. If
10967 @var{ignore} is true the value will be ignored.
10970 @deftypefn {Target Hook} bool TARGET_GIMPLE_FOLD_BUILTIN (gimple_stmt_iterator *@var{gsi})
10971 Fold a call to a machine specific built-in function that was set up
10972 by @samp{TARGET_INIT_BUILTINS}. @var{gsi} points to the gimple
10973 statement holding the function call. Returns true if any change
10974 was made to the GIMPLE stream.
10977 @deftypefn {Target Hook} int TARGET_COMPARE_VERSION_PRIORITY (tree @var{decl1}, tree @var{decl2})
10978 This hook is used to compare the target attributes in two functions to
10979 determine which function's features get higher priority. This is used
10980 during function multi-versioning to figure out the order in which two
10981 versions must be dispatched. A function version with a higher priority
10982 is checked for dispatching earlier. @var{decl1} and @var{decl2} are
10983 the two function decls that will be compared.
10986 @deftypefn {Target Hook} tree TARGET_GET_FUNCTION_VERSIONS_DISPATCHER (void *@var{decl})
10987 This hook is used to get the dispatcher function for a set of function
10988 versions. The dispatcher function is called to invoke the right function
10989 version at run-time. @var{decl} is one version from a set of semantically
10990 identical versions.
10993 @deftypefn {Target Hook} tree TARGET_GENERATE_VERSION_DISPATCHER_BODY (void *@var{arg})
10994 This hook is used to generate the dispatcher logic to invoke the right
10995 function version at run-time for a given set of function versions.
10996 @var{arg} points to the callgraph node of the dispatcher function whose
10997 body must be generated.
11000 @deftypefn {Target Hook} bool TARGET_CAN_USE_DOLOOP_P (double_int @var{iterations}, double_int @var{iterations_max}, unsigned int @var{loop_depth}, bool @var{entered_at_top})
11001 Return true if it is possible to use low-overhead loops (@code{doloop_end}
11002 and @code{doloop_begin}) for a particular loop. @var{iterations} gives the
11003 exact number of iterations, or 0 if not known. @var{iterations_max} gives
11004 the maximum number of iterations, or 0 if not known. @var{loop_depth} is
11005 the nesting depth of the loop, with 1 for innermost loops, 2 for loops that
11006 contain innermost loops, and so on. @var{entered_at_top} is true if the
11007 loop is only entered from the top.
11009 This hook is only used if @code{doloop_end} is available. The default
11010 implementation returns true. You can use @code{can_use_doloop_if_innermost}
11011 if the loop must be the innermost, and if there are no other restrictions.
11014 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const_rtx @var{insn})
11016 Take an instruction in @var{insn} and return NULL if it is valid within a
11017 low-overhead loop, otherwise return a string explaining why doloop
11018 could not be applied.
11020 Many targets use special registers for low-overhead looping. For any
11021 instruction that clobbers these this function should return a string indicating
11022 the reason why the doloop could not be applied.
11023 By default, the RTL loop optimizer does not use a present doloop pattern for
11024 loops containing function calls or branch on table instructions.
11027 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_COMBINED_INSN (rtx @var{insn})
11028 Take an instruction in @var{insn} and return @code{false} if the instruction is not appropriate as a combination of two or more instructions. The default is to accept all instructions.
11031 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
11033 Take a branch insn in @var{branch1} and another in @var{branch2}.
11034 Return true if redirecting @var{branch1} to the destination of
11035 @var{branch2} is possible.
11037 On some targets, branches may have a limited range. Optimizing the
11038 filling of delay slots can result in branches being redirected, and this
11039 may in turn cause a branch offset to overflow.
11042 @deftypefn {Target Hook} bool TARGET_CAN_FOLLOW_JUMP (const_rtx @var{follower}, const_rtx @var{followee})
11043 FOLLOWER and FOLLOWEE are JUMP_INSN instructions; return true if FOLLOWER may be modified to follow FOLLOWEE; false, if it can't. For example, on some targets, certain kinds of branches can't be made to follow through a hot/cold partitioning.
11046 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
11047 This target hook returns @code{true} if @var{x} is considered to be commutative.
11048 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
11049 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
11050 of the enclosing rtl, if known, otherwise it is UNKNOWN.
11053 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
11055 When the initial value of a hard register has been copied in a pseudo
11056 register, it is often not necessary to actually allocate another register
11057 to this pseudo register, because the original hard register or a stack slot
11058 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
11059 is called at the start of register allocation once for each hard register
11060 that had its initial value copied by using
11061 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
11062 Possible values are @code{NULL_RTX}, if you don't want
11063 to do any special allocation, a @code{REG} rtx---that would typically be
11064 the hard register itself, if it is known not to be clobbered---or a
11066 If you are returning a @code{MEM}, this is only a hint for the allocator;
11067 it might decide to use another register anyways.
11068 You may use @code{current_function_is_leaf} or
11069 @code{REG_N_SETS} in the hook to determine if the hard
11070 register in question will not be clobbered.
11071 The default value of this hook is @code{NULL}, which disables any special
11075 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
11076 This target hook returns nonzero if @var{x}, an @code{unspec} or
11077 @code{unspec_volatile} operation, might cause a trap. Targets can use
11078 this hook to enhance precision of analysis for @code{unspec} and
11079 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
11080 to analyze inner elements of @var{x} in which case @var{flags} should be
11084 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
11085 The compiler invokes this hook whenever it changes its current function
11086 context (@code{cfun}). You can define this function if
11087 the back end needs to perform any initialization or reset actions on a
11088 per-function basis. For example, it may be used to implement function
11089 attributes that affect register usage or code generation patterns.
11090 The argument @var{decl} is the declaration for the new function context,
11091 and may be null to indicate that the compiler has left a function context
11092 and is returning to processing at the top level.
11093 The default hook function does nothing.
11095 GCC sets @code{cfun} to a dummy function context during initialization of
11096 some parts of the back end. The hook function is not invoked in this
11097 situation; you need not worry about the hook being invoked recursively,
11098 or when the back end is in a partially-initialized state.
11099 @code{cfun} might be @code{NULL} to indicate processing at top level,
11100 outside of any function scope.
11103 @defmac TARGET_OBJECT_SUFFIX
11104 Define this macro to be a C string representing the suffix for object
11105 files on your target machine. If you do not define this macro, GCC will
11106 use @samp{.o} as the suffix for object files.
11109 @defmac TARGET_EXECUTABLE_SUFFIX
11110 Define this macro to be a C string representing the suffix to be
11111 automatically added to executable files on your target machine. If you
11112 do not define this macro, GCC will use the null string as the suffix for
11116 @defmac COLLECT_EXPORT_LIST
11117 If defined, @code{collect2} will scan the individual object files
11118 specified on its command line and create an export list for the linker.
11119 Define this macro for systems like AIX, where the linker discards
11120 object files that are not referenced from @code{main} and uses export
11124 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11125 Define this macro to a C expression representing a variant of the
11126 method call @var{mdecl}, if Java Native Interface (JNI) methods
11127 must be invoked differently from other methods on your target.
11128 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11129 the @code{stdcall} calling convention and this macro is then
11130 defined as this expression:
11133 build_type_attribute_variant (@var{mdecl},
11135 (get_identifier ("stdcall"),
11140 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11141 This target hook returns @code{true} past the point in which new jump
11142 instructions could be created. On machines that require a register for
11143 every jump such as the SHmedia ISA of SH5, this point would typically be
11144 reload, so this target hook should be defined to a function such as:
11148 cannot_modify_jumps_past_reload_p ()
11150 return (reload_completed || reload_in_progress);
11155 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11156 This target hook returns a register class for which branch target register
11157 optimizations should be applied. All registers in this class should be
11158 usable interchangeably. After reload, registers in this class will be
11159 re-allocated and loads will be hoisted out of loops and be subjected
11160 to inter-block scheduling.
11163 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11164 Branch target register optimization will by default exclude callee-saved
11166 that are not already live during the current function; if this target hook
11167 returns true, they will be included. The target code must than make sure
11168 that all target registers in the class returned by
11169 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11170 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11171 epilogues have already been generated. Note, even if you only return
11172 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11173 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11174 to reserve space for caller-saved target registers.
11177 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11178 This target hook returns true if the target supports conditional execution.
11179 This target hook is required only when the target has several different
11180 modes and they have different conditional execution capability, such as ARM.
11183 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop})
11184 This target hook returns a new value for the number of times @var{loop}
11185 should be unrolled. The parameter @var{nunroll} is the number of times
11186 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11187 the loop, which is going to be checked for unrolling. This target hook
11188 is required only when the target has special constraints like maximum
11189 number of memory accesses.
11192 @defmac POWI_MAX_MULTS
11193 If defined, this macro is interpreted as a signed integer C expression
11194 that specifies the maximum number of floating point multiplications
11195 that should be emitted when expanding exponentiation by an integer
11196 constant inline. When this value is defined, exponentiation requiring
11197 more than this number of multiplications is implemented by calling the
11198 system library's @code{pow}, @code{powf} or @code{powl} routines.
11199 The default value places no upper bound on the multiplication count.
11202 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11203 This target hook should register any extra include files for the
11204 target. The parameter @var{stdinc} indicates if normal include files
11205 are present. The parameter @var{sysroot} is the system root directory.
11206 The parameter @var{iprefix} is the prefix for the gcc directory.
11209 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11210 This target hook should register any extra include files for the
11211 target before any standard headers. The parameter @var{stdinc}
11212 indicates if normal include files are present. The parameter
11213 @var{sysroot} is the system root directory. The parameter
11214 @var{iprefix} is the prefix for the gcc directory.
11217 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11218 This target hook should register special include paths for the target.
11219 The parameter @var{path} is the include to register. On Darwin
11220 systems, this is used for Framework includes, which have semantics
11221 that are different from @option{-I}.
11224 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11225 This target macro returns @code{true} if it is safe to use a local alias
11226 for a virtual function @var{fndecl} when constructing thunks,
11227 @code{false} otherwise. By default, the macro returns @code{true} for all
11228 functions, if a target supports aliases (i.e.@: defines
11229 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11232 @defmac TARGET_FORMAT_TYPES
11233 If defined, this macro is the name of a global variable containing
11234 target-specific format checking information for the @option{-Wformat}
11235 option. The default is to have no target-specific format checks.
11238 @defmac TARGET_N_FORMAT_TYPES
11239 If defined, this macro is the number of entries in
11240 @code{TARGET_FORMAT_TYPES}.
11243 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11244 If defined, this macro is the name of a global variable containing
11245 target-specific format overrides for the @option{-Wformat} option. The
11246 default is to have no target-specific format overrides. If defined,
11247 @code{TARGET_FORMAT_TYPES} must be defined, too.
11250 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11251 If defined, this macro specifies the number of entries in
11252 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11255 @defmac TARGET_OVERRIDES_FORMAT_INIT
11256 If defined, this macro specifies the optional initialization
11257 routine for target specific customizations of the system printf
11258 and scanf formatter settings.
11261 @deftypevr {Target Hook} bool TARGET_RELAXED_ORDERING
11262 If set to @code{true}, means that the target's memory model does not
11263 guarantee that loads which do not depend on one another will access
11264 main memory in the order of the instruction stream; if ordering is
11265 important, an explicit memory barrier must be used. This is true of
11266 many recent processors which implement a policy of ``relaxed,''
11267 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11268 and ia64. The default is @code{false}.
11271 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11272 If defined, this macro returns the diagnostic message when it is
11273 illegal to pass argument @var{val} to function @var{funcdecl}
11274 with prototype @var{typelist}.
11277 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11278 If defined, this macro returns the diagnostic message when it is
11279 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11280 if validity should be determined by the front end.
11283 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11284 If defined, this macro returns the diagnostic message when it is
11285 invalid to apply operation @var{op} (where unary plus is denoted by
11286 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11287 if validity should be determined by the front end.
11290 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11291 If defined, this macro returns the diagnostic message when it is
11292 invalid to apply operation @var{op} to operands of types @var{type1}
11293 and @var{type2}, or @code{NULL} if validity should be determined by
11297 @deftypefn {Target Hook} {const char *} TARGET_INVALID_PARAMETER_TYPE (const_tree @var{type})
11298 If defined, this macro returns the diagnostic message when it is
11299 invalid for functions to include parameters of type @var{type},
11300 or @code{NULL} if validity should be determined by
11301 the front end. This is currently used only by the C and C++ front ends.
11304 @deftypefn {Target Hook} {const char *} TARGET_INVALID_RETURN_TYPE (const_tree @var{type})
11305 If defined, this macro returns the diagnostic message when it is
11306 invalid for functions to have return type @var{type},
11307 or @code{NULL} if validity should be determined by
11308 the front end. This is currently used only by the C and C++ front ends.
11311 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11312 If defined, this target hook returns the type to which values of
11313 @var{type} should be promoted when they appear in expressions,
11314 analogous to the integer promotions, or @code{NULL_TREE} to use the
11315 front end's normal promotion rules. This hook is useful when there are
11316 target-specific types with special promotion rules.
11317 This is currently used only by the C and C++ front ends.
11320 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11321 If defined, this hook returns the result of converting @var{expr} to
11322 @var{type}. It should return the converted expression,
11323 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11324 This hook is useful when there are target-specific types with special
11326 This is currently used only by the C and C++ front ends.
11329 @defmac TARGET_USE_JCR_SECTION
11330 This macro determines whether to use the JCR section to register Java
11331 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11332 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11336 This macro determines the size of the objective C jump buffer for the
11337 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11340 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11341 Define this macro if any target-specific attributes need to be attached
11342 to the functions in @file{libgcc} that provide low-level support for
11343 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11344 and the associated definitions of those functions.
11347 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11348 Define this macro to update the current function stack boundary if
11352 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11353 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11354 different argument pointer register is needed to access the function's
11355 argument list due to stack realignment. Return @code{NULL} if no DRAP
11359 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11360 When optimization is disabled, this hook indicates whether or not
11361 arguments should be allocated to stack slots. Normally, GCC allocates
11362 stacks slots for arguments when not optimizing in order to make
11363 debugging easier. However, when a function is declared with
11364 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11365 cannot safely move arguments from the registers in which they are passed
11366 to the stack. Therefore, this hook should return true in general, but
11367 false for naked functions. The default implementation always returns true.
11370 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11371 On some architectures it can take multiple instructions to synthesize
11372 a constant. If there is another constant already in a register that
11373 is close enough in value then it is preferable that the new constant
11374 is computed from this register using immediate addition or
11375 subtraction. We accomplish this through CSE. Besides the value of
11376 the constant we also add a lower and an upper constant anchor to the
11377 available expressions. These are then queried when encountering new
11378 constants. The anchors are computed by rounding the constant up and
11379 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11380 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11381 accepted by immediate-add plus one. We currently assume that the
11382 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11383 MIPS, where add-immediate takes a 16-bit signed value,
11384 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11385 is zero, which disables this optimization.
11388 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_ASAN_SHADOW_OFFSET (void)
11389 Return the offset bitwise ored into shifted address to get corresponding
11390 Address Sanitizer shadow memory address. NULL if Address Sanitizer is not
11391 supported by the target.
11394 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_MEMMODEL_CHECK (unsigned HOST_WIDE_INT @var{val})
11395 Validate target specific memory model mask bits. When NULL no target specific
11396 memory model bits are allowed.
11399 @deftypevr {Target Hook} {unsigned char} TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
11400 This value should be set if the result written by @code{atomic_test_and_set} is not exactly 1, i.e. the @code{bool} @code{true}.
11403 @deftypefn {Target Hook} bool TARGET_HAS_IFUNC_P (void)
11404 It returns true if the target supports GNU indirect functions.
11405 The support includes the assembler, linker and dynamic linker.
11406 The default value of this hook is based on target's libc.
11409 @deftypefn {Target Hook} {unsigned int} TARGET_ATOMIC_ALIGN_FOR_MODE (enum machine_mode @var{mode})
11410 If defined, this function returns an appropriate alignment in bits for an atomic object of machine_mode @var{mode}. If 0 is returned then the default alignment for the specified mode is used.
11413 @deftypefn {Target Hook} void TARGET_ATOMIC_ASSIGN_EXPAND_FENV (tree *@var{hold}, tree *@var{clear}, tree *@var{update})
11414 ISO C11 requires atomic compound assignments that may raise floating-point exceptions to raise exceptions corresponding to the arithmetic operation whose result was successfully stored in a compare-and-exchange sequence. This requires code equivalent to calls to @code{feholdexcept}, @code{feclearexcept} and @code{feupdateenv} to be generated at appropriate points in the compare-and-exchange sequence. This hook should set @code{*@var{hold}} to an expression equivalent to the call to @code{feholdexcept}, @code{*@var{clear}} to an expression equivalent to the call to @code{feclearexcept} and @code{*@var{update}} to an expression equivalent to the call to @code{feupdateenv}. The three expressions are @code{NULL_TREE} on entry to the hook and may be left as @code{NULL_TREE} if no code is required in a particular place. The default implementation leaves all three expressions as @code{NULL_TREE}. The @code{__atomic_feraiseexcept} function from @code{libatomic} may be of use as part of the code generated in @code{*@var{update}}.